A hydrogen fuel cell engine rapid start-up preheating system

Abstract

The present invention discloses a hydrogen fuel cell engine rapid start preheating system, which relates to the field of battery engine technology, establishes an initial thermodynamic prediction model, establishes a dynamic purge mechanism, obtains a second prediction model, establishes a state correction mechanism, obtains a third prediction model, integrates and establishes a weight distribution mechanism, obtains optimization results, and outputs control instructions. The present invention uses a dynamic purge and waste heat recovery mechanism to adjust the purge parameters and heating power according to real-time working conditions, thereby significantly reducing the energy consumption of the hydrogen fuel cell engine during low-temperature start-up. At the same time, the synergistic effect of humidity monitoring and high-frequency purge effectively reduces flooding failures and improves reliability. In addition, hydrogen retransmission is realized, further improving energy utilization efficiency, and providing strong support for the widespread application of hydrogen fuel cell engines in the field of new energy vehicles.

Description

Technical Field

[0001] The present invention relates to the technical field, and in particular to a rapid start-up and preheating system for a hydrogen fuel cell engine. Background Art

[0002] As a core power source for new energy vehicles, hydrogen fuel cell engines have long been a hot topic of research, optimizing their performance and improving their reliability. Starting a hydrogen fuel cell engine in low-temperature environments can be slow and energy-intensive due to factors such as low internal cell temperatures, slow hydrogen diffusion, and high membrane electrode hydration. Furthermore, low temperatures can easily lead to flooding, where moisture accumulates within the cell, affecting proper hydrogen diffusion and performance.

[0003] Currently, a Chinese invention patent application numbered CN202311724829.1 discloses an engine intake air preheating method, system, device, and engine. The method includes: collecting ambient temperature and atmospheric pressure values ​​after the engine starts to be powered; if the ambient temperature is less than 0 degrees Celsius, heating the air entering the engine through a heating rod, and injecting fuel onto the surface of the heating rod based on a pre-constructed MAP; after starting the engine, collecting the engine speed, and completing the engine intake air preheating if the engine speed is higher than a first set speed. This method is used to address the existing defects of poor performance of diesel engines during cold starts in plateau environments, and can achieve a minimum intake air preheating power that is adaptively adjusted with the diesel engine speed, altitude, and atmospheric temperature during the diesel engine startup process. This supplements the deficiencies of existing design methods and is conducive to improving the environmental adaptability of diesel engines.

[0004] The above technology is difficult to ensure the adequacy and effectiveness of anode hydrogen purge, and it is difficult to achieve hydrogen retransmission and waste heat recovery, which affects the preheating effect and causes excessive energy consumption. Summary of the Invention

[0005] The technical problem solved by the present invention is that the existing technology is difficult to ensure the sufficiency and effectiveness of anode hydrogen purge, and it is difficult to achieve hydrogen retransmission and waste heat recovery, which affects the preheating effect and causes excessive energy consumption.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] A hydrogen fuel cell engine rapid start-up and preheating system includes a model building module, a dynamic purge module, a multi-physics field monitoring module, a signal optimization module, and a strategy execution module:

[0008] The model building module is used to collect historical low-temperature startup data and establish an initial thermodynamic prediction model to output a predicted value of the fuel cell core temperature;

[0009] The dynamic purge module is used to collect membrane electrode state data, build a dynamic purge mechanism based on relative humidity and temperature change rate, and embed it into the initial thermodynamic prediction model to correct the temperature prediction value, generate a second prediction model, and output a corrected temperature sequence;

[0010] The multi-physics field monitoring module is used to collect temperature field distribution data, mechanical stress data, and near-infrared spectrum humidity data, extract multi-physics features, build a state correction mechanism, fuse the corrected temperature sequence with the physical field data, perform error correction processing, establish a third prediction model based on the corrected temperature sequence output by the second prediction model, and output a further optimized temperature prediction result;

[0011] The signal optimization module is used to integrate the output results of the third prediction model with the real-time purge parameters, establish a weight distribution mechanism, and output a comprehensive optimization result based on the optimization strategy;

[0012] The strategy execution module is used to output control instructions according to the comprehensive optimization results.

[0013] Preferably, the model building module includes a historical data collection unit and an initial model building unit;

[0014] The historical data acquisition unit is used to collect historical low-temperature startup data, wherein the historical low-temperature startup data includes ambient temperature, stack core temperature rise curve, first purge frequency, first purge pressure and membrane electrode hydration degree;

[0015] The initial model establishment unit is used to establish an initial thermodynamic prediction model according to the first purge frequency and the first purge pressure. The input data of the initial thermodynamic prediction model are the ambient temperature and the hydrogen pressure. The stack core temperature rise curve is the training target of the initial thermodynamic prediction model. The hydrogen pressure is obtained by mapping the first purge frequency and the first purge pressure. The mathematical expression is:

[0016] ;

[0017] in, is the hydrogen pressure, is the unit conversion coefficient, is the first purge frequency, is the first purge pressure, is the compensation constant;

[0018] The output data of the initial thermodynamic prediction model is the predicted value of the core temperature of the fuel cell stack.

[0019] Preferably, the logic of the initial thermodynamic prediction model is:

[0020] The historical low-temperature startup data is divided into a training set, a validation set, and a test set according to a preset ratio. The training set is used to adjust the parameters of the long short-term memory network model, and the optimal parameter combination is selected through the validation set. The model expression of the initial thermodynamic prediction model is:

[0021] ;

[0022] in, is the predicted value of the stack core temperature, is the ambient temperature, is the time variable;

[0023] Output the predicted value of the fuel cell core temperature.

[0024] Preferably, the dynamic purge module includes a purge data acquisition unit and a purge mechanism establishment unit;

[0025] The purge data acquisition unit is used to collect real-time hydrogen purge parameters and membrane electrode status data, the real-time hydrogen purge parameters include purge frequency and purge pressure, the membrane electrode status data include relative humidity and temperature change rate, the relative humidity is calculated based on the membrane electrode hydration degree and membrane surface temperature, there is a linear relationship or power function relationship between the relative humidity and the membrane electrode hydration degree, and the relative humidity is calculated using an empirical mapping relationship between the membrane surface temperature and the membrane electrode hydration degree;

[0026] The purge mechanism establishment unit is used to establish a dynamic purge mechanism based on the membrane electrode state data, embed the dynamic purge mechanism into the initial thermodynamic prediction model, replace the first purge frequency and first purge pressure in the initial thermodynamic prediction model with the second purge frequency and purge pressure output by the dynamic purge mechanism, and output a second prediction model, which outputs a corrected temperature prediction sequence.

[0027] Preferably, the logic of the dynamic purge mechanism is:

[0028] When the ambient temperature is lower than a preset first ambient temperature, the purge pressure is set to a first purge pressure value;

[0029] When the ambient temperature is between the first ambient temperature and zero degrees, the purge pressure is set to a second purge pressure value;

[0030] When the ambient temperature is greater than or equal to zero degrees, the purge pressure is set to the third purge pressure value;

[0031] The first purge pressure value is greater than the second purge pressure value and greater than the third purge pressure value.

[0032] Preferably, the multi-physics field monitoring module includes a sensor network unit and a state correction unit;

[0033] The sensor network unit collects temperature field distribution data, mechanical stress data and near-infrared spectrum humidity data according to the micro fiber Bragg grating array inside the battery stack, and combines the temperature field distribution data, mechanical stress data and near-infrared spectrum humidity data and outputs them as physical field data;

[0034] The state correction unit is used to extract the characteristics of multi-physical field data, receive the corrected temperature prediction sequence output by the second prediction model as a reference input, establish a state correction mechanism, and establish a third prediction model based on the state correction mechanism and the corrected temperature prediction sequence output by the second prediction model.

[0035] Preferably, the state correction unit performs feature extraction on the multi-physical field data, and the feature extraction includes statistical feature extraction, frequency domain feature extraction and physical coupling feature extraction;

[0036] Establish a state correction mechanism, the logic of which is:

[0037] Establish abnormal state judgment rules:

[0038] If the near-infrared spectrum humidity data is greater than the preset first humidity data and the temperature field distribution data is less than zero degrees, a high-frequency purge instruction is triggered;

[0039] If the mechanical stress data is greater than the preset first stress data, it is determined to be a structural damage risk and the safety shutdown process is initiated;

[0040] Establish a correction strategy mathematical model, the correction strategy mathematical model includes:

[0041] Purge frequency correction mechanism:

[0042] ;

[0043] in, is the corrected purge frequency, is the proportionality coefficient, is the humidity gradient modulus;

[0044] Heating power compensation mechanism:

[0045] If the temperature field distribution data is less than the area of ​​the preset first temperature data, calculate the additional heating power:

[0046] ;

[0047] in, For additional heating power, is the compensation coefficient, The local temperature of the area where the temperature field distribution data is less than the preset first temperature data;

[0048] The state correction mechanism further receives the corrected temperature prediction sequence output by the second prediction model as a reference input of the state correction mechanism, and performs error correction processing after fusing it with the physical field data;

[0049] The corrected purge frequency and additional heating power are used as new input nodes and injected into the second prediction model through the residual connection:

[0050] ;

[0051] in, is the corrected temperature prediction sequence, is the purge frequency correction factor, is the heating power correction factor;

[0052] The original input and the correction are fused through a fully connected layer, and the network weights are updated using an adaptive moment estimation algorithm to form a third prediction model, which outputs a further optimized temperature prediction result.

[0053] Preferably, the signal optimization module includes a logic integration unit and an optimization result acquisition unit;

[0054] The logic integration unit is used to integrate the logic of the third prediction model and establish a weight distribution mechanism;

[0055] The optimization result acquisition unit is used to input real-time purge parameters and multi-physical field data into the third prediction model, accept the further optimized temperature prediction result output by the third prediction model, and combine the optimized purge frequency and the optimized heating power to output the optimized result.

[0056] Preferably, the logic of the weight distribution mechanism is:

[0057] The purge demand weight is calculated according to the dynamic purge mechanism. The mathematical expression of the purge demand weight is:

[0058] ;

[0059] in, is the purge demand weight, is the purge frequency weight coefficient, is the purge pressure weight coefficient, is the purge pressure, is the second purge frequency;

[0060] If the purge demand weight is greater than a preset purge demand weight threshold, reducing the heating power by a first percentage and outputting the optimized heating power;

[0061] If the purge demand weight is less than or equal to the preset purge weight threshold, the heating power is maintained;

[0062] The optimized heating power and the optimized purge frequency are combined with the further optimized temperature prediction result to form a comprehensive optimization result.

[0063] Preferably, the policy execution module includes an instruction generation unit and an execution unit;

[0064] The instruction generation unit is used to generate control instructions according to the comprehensive optimization results;

[0065] The execution unit is used to send control instructions to the catalytic burner, the purge valve and the stack controller.

[0066] Beneficial effects of the present invention: The present invention uses a dynamic purge mechanism to adjust the purge parameters and heating power according to real-time operating conditions, thereby significantly reducing the energy consumption of low-temperature start-up of the hydrogen fuel cell engine. At the same time, the synergistic effect of humidity monitoring and high-frequency purge effectively reduces flooding failures and improves reliability, providing strong support for the widespread application of hydrogen fuel cell engines in the field of new energy vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A basic flow chart of a hydrogen fuel cell engine rapid start-up and preheating system provided in accordance with one embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0069] Example, see Figure 1 , provides a hydrogen fuel cell engine rapid start-up preheating system, including a model building module, a dynamic purge module, a multi-physics field monitoring module, a signal optimization module and a strategy execution module:

[0070] The model building module is used to collect historical low-temperature startup data and establish an initial thermodynamic prediction model to output the predicted value of the fuel cell core temperature.

[0071] The dynamic purge module is used to collect membrane electrode status data, build a dynamic purge mechanism based on relative humidity and temperature change rate, and embed it into the initial thermodynamic prediction model to correct the temperature prediction value, generate a second prediction model, and output the corrected temperature sequence.

[0072] The multi-physics field monitoring module is used to collect temperature field distribution data, mechanical stress data and near-infrared spectral humidity data, extract multi-physics features, build a state correction mechanism, fuse the corrected temperature sequence with the physical field data, and perform error correction processing. Based on the corrected temperature sequence output by the second prediction model, a third prediction model is established to output further optimized temperature prediction results.

[0073] The signal optimization module is used to integrate the output results of the third prediction model with the real-time purge parameters, establish a weight distribution mechanism, and output comprehensive optimization results based on the optimization strategy.

[0074] The strategy execution module is used to output control instructions based on the comprehensive optimization results.

[0075] The model building module includes a historical data collection unit and an initial model building unit.

[0076] The historical data acquisition unit is used to collect historical low-temperature startup data, which includes ambient temperature, stack core temperature rise curve, first purge frequency, first purge pressure and membrane electrode hydration degree.

[0077] The historical data acquisition unit is responsible for comprehensively collecting the startup data of the hydrogen fuel cell engine in a low-temperature environment. These data provide a solid foundation for the establishment of the initial thermodynamic prediction model, ensuring the accuracy and reliability of the model.

[0078] The initial model establishment unit is used to establish an initial thermodynamic prediction model based on the first purge frequency and the first purge pressure. The input data of the initial thermodynamic prediction model are the ambient temperature and the hydrogen pressure. The core temperature rise curve of the fuel cell stack is the training target of the initial thermodynamic prediction model. The hydrogen pressure is obtained by mapping the first purge frequency and the first purge pressure. The mathematical expression is:

[0079] ;

[0080] in, is the hydrogen pressure, is the unit conversion coefficient, is the first purge frequency, is the first purge pressure, is the compensation constant.

[0081] The output data of the initial thermodynamic prediction model is the predicted value of the core temperature of the fuel cell stack.

[0082] The core temperature rise curve of the fuel cell stack is the target output variable in the model training phase, which represents the trajectory of the change of the core temperature of the fuel cell stack over time. It serves as the supervision signal of the initial thermodynamic prediction model. By learning the mapping relationship between the input variables and the temperature rise curve, the prediction function of the future temperature is realized.

[0083] The logic of the initial thermodynamic prediction model is:

[0084] The historical low-temperature startup data is divided into a training set, a validation set, and a test set according to a preset ratio. The training set is used to adjust the parameters of the long short-term memory network model, and the optimal parameter combination is selected through the validation set. The model expression of the initial thermodynamic prediction model is:

[0085] ;

[0086] in, is the predicted value of the stack core temperature, is the ambient temperature, is the time variable.

[0087] Output the predicted value of the fuel cell core temperature.

[0088] The initial model building unit utilizes historical low-temperature startup data and, through a long-short-term memory network, constructs an initial thermodynamic prediction model capable of predicting the core temperature of the fuel cell stack. Based on the inputs of ambient temperature and hydrogen pressure, this model outputs a temperature prediction sequence for the first time period in the future, providing strong data support for the engine warm-up strategy. Model parameters are adjusted using the training set, and the optimal parameter combination is selected using the validation set, ensuring the model's prediction accuracy and generalization capabilities.

[0089] The model building module integrates historical data collection with initial model building to construct an accurate initial thermodynamic prediction model based on extensive historical low-temperature startup data. This model can predict the core temperature of the fuel cell stack within the first time period, providing a scientific basis for rapid startup and preheating of hydrogen fuel cell engines, thereby improving startup efficiency and reducing energy consumption.

[0090] The dynamic purge module includes a purge data acquisition unit and a purge mechanism establishment unit.

[0091] The purge data acquisition unit is used to collect real-time hydrogen purge parameters and membrane electrode status data. The real-time hydrogen purge parameters include purge frequency and purge pressure. The membrane electrode status data includes relative humidity and temperature change rate. The relative humidity is calculated from the membrane electrode hydration degree and membrane surface temperature. There is a linear relationship or power function relationship between the relative humidity and the membrane electrode hydration degree. The relative humidity is calculated using an empirical mapping relationship between the membrane surface temperature and the membrane electrode hydration degree.

[0092] The purge data acquisition unit accurately collects hydrogen purge parameters and membrane electrode status data in real time. This data is the foundation for establishing a dynamic purge mechanism and is key information for optimizing the engine startup process. Continuous and stable data acquisition provides strong data support for establishing and adjusting the dynamic purge mechanism.

[0093] The purge mechanism establishment unit is used to establish a dynamic purge mechanism based on the membrane electrode state data, embed the dynamic purge mechanism into the initial thermodynamic prediction model, replace the first purge frequency and first purge pressure in the initial thermodynamic prediction model with the second purge frequency and purge pressure output by the dynamic purge mechanism, and output the second prediction model. The second prediction model outputs a corrected temperature prediction sequence.

[0094] The logic of the dynamic purge mechanism is:

[0095] When the ambient temperature is lower than a preset first ambient temperature, the purge pressure is set to a first purge pressure value.

[0096] When the ambient temperature is between the first ambient temperature and zero degrees, the purge pressure is set to a second purge pressure value.

[0097] When the ambient temperature is greater than or equal to zero degrees, the purge pressure is set to a third purge pressure value.

[0098] The first purge pressure value is greater than the second purge pressure value and greater than the third purge pressure value.

[0099] The purge mechanism establishment unit establishes a dynamic purge mechanism based on collected membrane electrode status data and changes in ambient temperature. This mechanism automatically adjusts the purge frequency and pressure according to ambient temperature, minimizing energy consumption while ensuring engine startup speed. By embedding the dynamic purge mechanism into the initial thermodynamic prediction model, more refined control of the engine startup process is achieved, improving the model's prediction accuracy and practicality. Furthermore, the unit continuously optimizes and adjusts the dynamic purge mechanism based on actual needs to adapt to engine startup requirements under different operating conditions and environmental conditions.

[0100] The dynamic purge module collects real-time hydrogen purge parameters and membrane electrode status data, and establishes a dynamic purge mechanism based on this data, enabling intelligent control of the hydrogen fuel cell engine's purge process. The module automatically adjusts purge frequency and pressure based on ambient temperature changes, optimizing purge effectiveness, increasing engine startup speed, and reducing unnecessary energy consumption. By embedding these dynamic purge strategies into the initial thermodynamic prediction model, the model's accuracy and practicality are further enhanced, providing a more refined control strategy for engine preheating and startup.

[0101] The multi-physics field monitoring module includes a sensor network unit and a state correction unit.

[0102] The sensor network unit collects temperature field distribution data, mechanical stress data and near-infrared spectrum humidity data based on the micro fiber Bragg grating array inside the fuel cell stack, and merges the temperature field distribution data, mechanical stress data and near-infrared spectrum humidity data into physical field data.

[0103] The sensor network unit utilizes high-precision sensors such as a micro-fiber Bragg grating array to collect real-time data on the temperature distribution, mechanical stress, and near-infrared spectral humidity within the fuel cell stack. This data covers key physical field information related to fuel cell operation, providing a rich and accurate foundation for subsequent state correction and predictive modeling. By optimizing sensor layout and acquisition strategies, the unit ensures comprehensive and real-time data, providing a strong foundation for intelligent engine monitoring.

[0104] The state correction unit is used to extract the characteristics of multi-physical field data, receive the corrected temperature prediction sequence output by the second prediction model as a reference input, establish a state correction mechanism, and establish a third prediction model based on the state correction mechanism and the corrected temperature prediction sequence output by the second prediction model.

[0105] The state correction unit performs feature extraction on multi-physical field data, and the feature extraction includes statistical feature extraction, frequency domain feature extraction and physical coupling feature extraction.

[0106] Establish a status correction mechanism. The logic of the status correction mechanism is:

[0107] Establish abnormal state judgment rules:

[0108] If the near-infrared spectrum humidity data is greater than the preset first humidity data and the temperature field distribution data is less than zero degrees, a high-frequency purge instruction is triggered.

[0109] If the mechanical stress data is greater than the preset first stress data, it is determined to be a structural damage risk and the safety shutdown process is initiated.

[0110] Establish a mathematical model for the correction strategy, which includes:

[0111] Purge frequency correction mechanism:

[0112] ;

[0113] in, is the corrected purge frequency, is the proportionality coefficient, is the humidity gradient modulus.

[0114] Heating power compensation mechanism:

[0115] If the temperature field distribution data is less than the area of ​​the preset first temperature data, calculate the additional heating power:

[0116] ;

[0117] in, For additional heating power, is the compensation coefficient, It is the local temperature of the area where the temperature field distribution data is less than the preset first temperature data.

[0118] The state correction mechanism further receives the corrected temperature prediction sequence output by the second prediction model as a reference input of the state correction mechanism, and performs error correction processing after fusing it with the physical field data.

[0119] The corrected purge frequency and additional heating power are used as new input nodes and injected into the second prediction model through the residual connection:

[0120] ;

[0121] in, is the corrected temperature prediction sequence, is the purge frequency correction factor, is the heating power correction factor.

[0122] The original input and the correction are fused through a fully connected layer, and the network weights are updated using an adaptive moment estimation algorithm to form a third prediction model, which outputs a further optimized temperature prediction result.

[0123] The state correction unit performs in-depth feature extraction on multi-physics field data, including statistical features, frequency domain features, and physical coupling features, thereby comprehensively capturing the operating status of the fuel cell stack. By establishing abnormal state judgment rules and correction strategy mathematical models, the unit can intelligently identify abnormal states of the fuel cell stack and trigger corresponding countermeasures, such as high-frequency purge instructions or safe shutdown processes. In addition, the corrected purge frequency and additional heating power are injected into the prediction model as new input nodes, the original input and correction information are fused through the fully connected layer, and the network weights are updated using an adaptive moment estimation algorithm, further improving the prediction accuracy and adaptability of the model. These measures jointly ensure the safe and stable operation of the engine and optimize its performance.

[0124] The multi-physics field monitoring module integrates a sensor network and a state correction unit to enable real-time monitoring and intelligent correction of the temperature field, mechanical stress, and humidity within the hydrogen fuel cell stack. This module not only improves the accuracy and comprehensiveness of data acquisition but also effectively identifies and responds to abnormal stack conditions through feature extraction and state correction mechanisms, ensuring safe and stable engine operation. Furthermore, embedding correction strategies into the predictive model significantly improves the model's predictive accuracy and adaptability, providing strong support for optimized control during engine preheating, startup, and operation.

[0125] The signal optimization module includes a logic integration unit and an optimization result acquisition unit.

[0126] The logic integration unit is used to integrate the logic of the third prediction model and establish a weight distribution mechanism.

[0127] The logic of the weight distribution mechanism is:

[0128] The purge demand weight is calculated based on the dynamic purge mechanism. The mathematical expression of the purge demand weight is:

[0129] ;

[0130] in, is the purge demand weight, is the purge frequency weight coefficient, is the purge pressure weight coefficient, is the purge pressure, is the second purge frequency.

[0131] If the purge requirement weight is greater than a preset purge requirement weight threshold, the heating power is reduced by a first percentage and output as the optimized heating power.

[0132] If the purge demand weight is less than or equal to the preset purge weight threshold, the heating power is maintained.

[0133] The optimized heating power and the optimized purge frequency are combined with the further optimized temperature prediction results to form a comprehensive optimization result.

[0134] The logic integration unit integrates the logic of the third prediction model and establishes a weight allocation mechanism based on the dynamic purge mechanism. This mechanism intelligently analyzes the purge demand weight and dynamically adjusts the heating power based on real-time operating conditions. When the purge demand weight exceeds a preset threshold, the heating power is automatically reduced to prioritize the purge demand. When the purge demand weight is less than or equal to the preset threshold, the heating power is maintained unchanged. This intelligent weight allocation strategy ensures safe engine operation while improving energy efficiency.

[0135] The optimization result acquisition unit is used to input real-time purge parameters and multi-physical field data into the third prediction model, accept the further optimized temperature prediction results output by the third prediction model, and combine the optimized purge frequency and the optimized heating power to output the optimized results.

[0136] The optimization result acquisition unit inputs real-time purge parameters and multi-physics field data into a third prediction model. Through precise calculations, the model outputs optimized purge frequency and heating power. These optimization results not only consider the actual needs of the current operating conditions but also fully integrate historical data and forecast information, ensuring precise and forward-looking control. By combining the optimized purge frequency and heating power into an output, this unit provides powerful data support for intelligent engine control, helping to improve overall engine performance and operating efficiency.

[0137] The signal optimization module integrates the logic of the third prediction model and establishes a scientific weighting mechanism. This intelligently analyzes real-time purge parameters and multi-physics field data, and outputs optimized purge frequency and heating power accordingly. This module not only improves the accuracy of purge and heating control but also effectively balances purge demand and heating energy consumption, ensuring efficient and safe hydrogen fuel cell engine startup and operation.

[0138] The policy execution module includes an instruction generation unit and an execution unit.

[0139] The instruction generation unit is used to generate control instructions according to the comprehensive optimization results.

[0140] The Command Generation Unit is a core component of the Strategy Execution Module. It is responsible for intelligently generating control commands based on the optimization results. These commands not only accurately reflect the optimal control strategy for the current operating conditions, but also fully consider engine safety and operating efficiency. Through the Command Generation Unit, the optimization results are converted into specific, executable control commands, providing clear guidance for subsequent execution.

[0141] The execution unit is used to send control instructions to the catalytic burner, purge valve and stack controller.

[0142] The execution unit is responsible for accurately transmitting the control commands generated by the command generation unit to the catalytic burner, purge valve, and stack controller. Through a high-speed, stable communication channel, the execution unit ensures the timely transmission and effective execution of control commands. This not only improves the system's responsiveness but also enhances control accuracy and stability. During execution, the execution unit also features fault detection and self-protection capabilities, enabling timely action in the event of anomalies to ensure safe engine operation.

[0143] The strategy execution module, the terminal link of the system, is responsible for converting optimization results into specific control instructions and accurately executing these instructions to achieve intelligent regulation of the catalytic burner, purge valve, and stack controller. This module ensures the effective implementation of the optimization strategy, improves engine operating efficiency and safety, and enhances the system's responsiveness and stability.

[0144] This system utilizes a dynamic purge module, which uses real-time hydrogen purge parameters and membrane electrode status data to establish a dynamic purge mechanism. This mechanism dynamically adjusts the purge frequency and pressure based on ambient temperature and membrane electrode status, minimizing unnecessary energy consumption while ensuring startup speed. Furthermore, the system utilizes a multi-physics monitoring module and a signal optimization module to fine-tune purge parameters and heating power, further reducing energy consumption. The system also incorporates waste heat recovery, utilizing waste heat from the catalytic combustor to preheat the battery cells, improving energy efficiency. The system uses a sensor network within the multi-physics monitoring module to collect real-time humidity data within the stack. When humidity exceeds a preset threshold, the system triggers a high-frequency purge command to promptly remove moisture from the battery cells and prevent flooding. This synergistic effect of humidity monitoring and high-frequency purge significantly improves system reliability and safety. During the dynamic purge process, the system not only optimizes purge parameters but also enables hydrogen recirculation. By precisely controlling the opening and closing of the purge valve, the system recirculates unreacted hydrogen back into the battery cells, reducing hydrogen waste. At the same time, the system also makes full use of the waste heat generated by the catalytic burner and uses it to preheat the battery, thereby improving the overall energy utilization efficiency.

[0145] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium may be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0146] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A hydrogen fuel cell engine rapid start-up preheating system, characterized in that: Includes model building module, dynamic purge module, multi-physics field monitoring module, signal optimization module and strategy execution module: The model building module is used to collect historical low-temperature startup data and establish an initial thermodynamic prediction model to output a predicted value of the fuel cell core temperature; The dynamic purge module is used to collect membrane electrode state data, build a dynamic purge mechanism based on relative humidity and temperature change rate, and embed it into the initial thermodynamic prediction model to correct the temperature prediction value, generate a second prediction model, and output a corrected temperature sequence; The multi-physics field monitoring module is used to collect temperature field distribution data, mechanical stress data, and near-infrared spectrum humidity data, extract multi-physics features, build a state correction mechanism, fuse the corrected temperature sequence with the physical field data, perform error correction processing, establish a third prediction model based on the corrected temperature sequence output by the second prediction model, and output a further optimized temperature prediction result; The signal optimization module is used to integrate the output results of the third prediction model with the real-time purge parameters, establish a weight distribution mechanism, and output a comprehensive optimization result based on the optimization strategy; The strategy execution module is used to output control instructions according to the comprehensive optimization results; The model building module includes a historical data collection unit and an initial model building unit; The historical data acquisition unit is used to collect historical low-temperature startup data, wherein the historical low-temperature startup data includes ambient temperature, stack core temperature rise curve, first purge frequency, first purge pressure and membrane electrode hydration degree; The initial model establishment unit is used to establish an initial thermodynamic prediction model according to the first purge frequency and the first purge pressure. The input data of the initial thermodynamic prediction model are the ambient temperature and the hydrogen pressure. The stack core temperature rise curve is the training target of the initial thermodynamic prediction model. The hydrogen pressure is obtained by mapping the first purge frequency and the first purge pressure. The mathematical expression is: ； in, is the hydrogen pressure, is the unit conversion coefficient, is the first purge frequency, is the first purge pressure, is the compensation constant; The output data of the initial thermodynamic prediction model is the predicted value of the core temperature of the fuel cell stack; The logic of the initial thermodynamic prediction model is: The historical low-temperature startup data is divided into a training set, a validation set, and a test set according to a preset ratio. The training set is used to adjust the parameters of the long short-term memory network model, and the optimal parameter combination is selected through the validation set. The model expression of the initial thermodynamic prediction model is: ； in, is the predicted value of the stack core temperature, is the ambient temperature, is the time variable; Output the predicted value of the fuel cell core temperature; The logic of the dynamic purge mechanism is: When the ambient temperature is lower than a preset first ambient temperature, the purge pressure is set to a first purge pressure value; When the ambient temperature is between the first ambient temperature and zero degrees, the purge pressure is set to a second purge pressure value; When the ambient temperature is greater than or equal to zero degrees, the purge pressure is set to the third purge pressure value; The first purge pressure value is greater than the second purge pressure value and greater than the third purge pressure value; The state correction unit performs feature extraction on the multi-physical field data, wherein the feature extraction includes statistical feature extraction, frequency domain feature extraction and physical coupling feature extraction; Establish a state correction mechanism, the logic of which is: Establish abnormal state judgment rules: If the near-infrared spectrum humidity data is greater than the preset first humidity data and the temperature field distribution data is less than zero degrees, a high-frequency purge instruction is triggered; If the mechanical stress data is greater than the preset first stress data, it is determined to be a structural damage risk and the safety shutdown process is initiated; Establish a correction strategy mathematical model, the correction strategy mathematical model includes: Purge frequency correction mechanism: ； in, is the corrected purge frequency, is the proportionality coefficient, is the humidity gradient modulus; Heating power compensation mechanism: If the temperature field distribution data is less than the area of ​​the preset first temperature data, calculate the additional heating power: ； in, For additional heating power, is the compensation coefficient, The local temperature of the area where the temperature field distribution data is less than the preset first temperature data; The state correction mechanism further receives the corrected temperature prediction sequence output by the second prediction model as a reference input of the state correction mechanism, and performs error correction processing after fusing it with the physical field data; The corrected purge frequency and additional heating power are used as new input nodes and injected into the second prediction model through the residual connection: ； in, is the corrected temperature prediction sequence, is the purge frequency correction factor, is the heating power correction factor; The original input and the correction are fused through a fully connected layer, and the network weights are updated using an adaptive moment estimation algorithm to form a third prediction model, which outputs a further optimized temperature prediction result; The logic of the weight distribution mechanism is: The purge demand weight is calculated according to the dynamic purge mechanism. The mathematical expression of the purge demand weight is: ； in, is the purge demand weight, is the purge frequency weight coefficient, is the purge pressure weight coefficient, is the purge pressure, is the second purge frequency; If the purge demand weight is greater than a preset purge demand weight threshold, reducing the heating power by a first percentage and outputting the optimized heating power; If the purge demand weight is less than or equal to the preset purge weight threshold, the heating power is maintained; The optimized heating power and the optimized purge frequency are combined with the further optimized temperature prediction result to form a comprehensive optimization result.

2. A hydrogen fuel cell engine rapid start-up preheating system according to claim 1, characterized in that: The dynamic purge module includes a purge data acquisition unit and a purge mechanism establishment unit; The purge data acquisition unit is used to collect real-time hydrogen purge parameters and membrane electrode status data, the real-time hydrogen purge parameters include purge frequency and purge pressure, the membrane electrode status data include relative humidity and temperature change rate, the relative humidity is calculated based on the membrane electrode hydration degree and membrane surface temperature, there is a linear relationship or power function relationship between the relative humidity and the membrane electrode hydration degree, and the relative humidity is calculated using an empirical mapping relationship between the membrane surface temperature and the membrane electrode hydration degree; The purge mechanism establishment unit is used to establish a dynamic purge mechanism based on the membrane electrode state data, embed the dynamic purge mechanism into the initial thermodynamic prediction model, replace the first purge frequency and first purge pressure in the initial thermodynamic prediction model with the second purge frequency and purge pressure output by the dynamic purge mechanism, and output a second prediction model, which outputs a corrected temperature prediction sequence.

3. A hydrogen fuel cell engine rapid start-up and preheating system according to claim 1, characterized in that: The multi-physics field monitoring module includes a sensor network unit and a state correction unit; The sensor network unit collects temperature field distribution data, mechanical stress data and near-infrared spectrum humidity data according to the micro fiber Bragg grating array inside the battery stack, and combines the temperature field distribution data, mechanical stress data and near-infrared spectrum humidity data and outputs them as physical field data; The state correction unit is used to extract the characteristics of multi-physical field data, receive the corrected temperature prediction sequence output by the second prediction model as a reference input, establish a state correction mechanism, and establish a third prediction model based on the state correction mechanism and the corrected temperature prediction sequence output by the second prediction model.

4. A hydrogen fuel cell engine rapid start-up and preheating system as claimed in claim 1, characterized in that: The signal optimization module includes a logic integration unit and an optimization result acquisition unit; The logic integration unit is used to integrate the logic of the third prediction model and establish a weight distribution mechanism; The optimization result acquisition unit is used to input real-time purge parameters and multi-physical field data into the third prediction model, accept the further optimized temperature prediction result output by the third prediction model, and combine the optimized purge frequency and the optimized heating power to output the optimized result.

5. A hydrogen fuel cell engine rapid start-up and preheating system as claimed in claim 1, characterized in that: The policy execution module includes an instruction generation unit and an execution unit; The instruction generation unit is used to generate control instructions according to the comprehensive optimization results; The execution unit is used to send control instructions to the catalytic burner, the purge valve and the stack controller.

Images