A turbine inter-stage flame fuel cell hybrid control system based on hysteresis regulation and aging compensation

Abstract

The application discloses a turbine inter-stage flame fuel cell hybrid control system based on hysteresis regulation and aging compensation, which comprises, in sequence along the airflow direction, an air inlet, a compressor, a combustion chamber, a high-pressure turbine, a rich combustion chamber, a fuel cell stack, a lean combustion chamber, a low-pressure turbine and a tail nozzle; and simultaneously comprises a fuel tank, a fuel pump, a first shunt valve, a second shunt valve and an intelligent control unit. The application aims to deeply integrate advanced hysteresis regulation and disturbance compensation algorithms into the hardware architecture of the turbine inter-stage flame fuel cell. By constructing a non-disturbance temperature model and an aging equation for the turbine environment, the opening degree of the fuel shunt and the working parameters of the stack are corrected in real time, so that efficient cooperation of power generation and propulsion and long-life operation are realized.

Description

Technical Field

[0001] This invention belongs to the field of battery hybrid control system technology, and specifically relates to a turbine-stage flame fuel cell hybrid control system based on hysteresis regulation and aging compensation. Background Technology

[0002] With the increasing demand for aviation electrification, hybrid power systems integrating fuel cells and gas turbines have become a research hotspot. In particular, directly driving solid oxide fuel cells (SOFCs) using the high-temperature environment between turbine stages can eliminate the need for heavy preheaters and utilize waste heat for thrust boosting. Meanwhile, for the control of such complex nonlinear systems, model-based predictive control and intelligent algorithm optimization (such as ant colony optimization and fuzzy PID) are gradually replacing traditional single control methods.

[0003] Existing technology (application number CN 202511547599.5) proposes a hybrid system integrating a flame fuel cell between turbine stages, comprising a fuel-rich combustor (for hydrogen production), an SOFC stack, and a lean combustor (for afterburner-driven turbine), using a splitter to regulate fuel flow. However, its hardware architecture lacks a sophisticated closed-loop control strategy. Changes in turbine operating conditions (such as climb and dive) can instantaneously alter the airflow parameters entering the fuel-rich combustor. If only a simple splitter is used for adjustment, it will cause drastic fluctuations in the SOFC inlet temperature, which will not only affect power generation efficiency but may even damage the stack due to thermal stress.

[0004] Another existing technology (application number CN 202511408883.4) proposes a hydrogen fuel cell auxiliary power control system. It utilizes a "simulation module" to construct a disturbance-free model, a "compensation module" to calculate environmental disturbance weights, and a "hysteresis control model" to tune the temperature. However, this strategy is primarily designed for general-purpose hydrogen fuel cells and does not consider the unique environment where the oxidant flow and temperature between turbine stages are constrained by the compressor and high-pressure turbine speeds. Furthermore, the existing technology does not integrate battery aging prediction with turbine thrust maintenance. When battery aging leads to a decrease in conversion efficiency, the distribution ratio needs to be dynamically adjusted to avoid insufficient total thrust or exhaust overheating. Summary of the Invention

[0005] To overcome the above technical problems, the present invention aims to provide a turbine-stage flame fuel cell hybrid control system based on hysteresis control and aging compensation. This system aims to deeply integrate advanced hysteresis control and disturbance compensation algorithms into the turbine-stage flame fuel cell hardware architecture. By constructing an interference-free temperature model and aging equation for the turbine environment, the opening degree of the fuel splitter and the stack operating parameters are corrected in real time, achieving efficient coordination of power generation and propulsion, and long-life operation.

[0006] A hybrid control system for inter-stage flame fuel cell based on hysteresis regulation and aging compensation includes, in sequence along the airflow direction, an intake duct, a compressor, a combustion chamber, a high-pressure turbine, a fuel-rich combustion chamber, a fuel cell stack, a lean combustion chamber, a low-pressure turbine, and a tail nozzle.

[0007] It also includes a fuel storage tank, a fuel pump, a first diversion valve (to regulate total fuel distribution), a second diversion valve (to regulate oxidant distribution), and an intelligent control unit.

[0008] Outside air enters the compressor through the intake duct and is pressurized. It then enters the main combustion chamber to mix and burn with the fuel. The resulting high-temperature, high-pressure gas first drives the high-pressure turbine to perform work. The interstage gas discharged from the high-pressure turbine is split into two paths upon entering the second split valve:

[0009] The main flow path enters the rich combustion chamber, while the bypass flow path leads directly to the lean combustion chamber. The products from the rich combustion chamber enter the anode of the fuel cell stack for electrochemical power generation. The exhaust gas from the stack and the bypass flow converge in the lean combustion chamber, undergo secondary combustion, expand and drive the low-pressure turbine, and finally exit through the tail nozzle.

[0010] Fuel is drawn from the fuel storage tank by the fuel pump and enters the first diversion valve. The first diversion valve injects the total fuel into the rich combustion chamber and the lean combustion chamber according to the command ratio.

[0011] The intelligent control unit includes a data acquisition layer, a simulation prediction layer, a compensation and correction layer, and an execution control layer.

[0012] The data acquisition layer is connected to sensor signals arranged in the air intake, the high-pressure turbine outlet and both ends of the fuel cell stack, and is used to collect physical environment and state parameters.

[0013] The simulation prediction layer is used to output theoretical baseline data based on the embedded fluid dynamics model;

[0014] The compensation and correction layer is connected to the data acquisition layer and the simulation prediction layer respectively, and is used to calculate the corrected target instruction based on the difference between the physical environment and state parameters and the theoretical benchmark data.

[0015] The execution control layer is electrically connected to the compensation correction layer, the first diverter valve, and the second diverter valve, respectively, and is used to convert the target command into a drive voltage signal to adjust the opening degree of the first diverter valve and the second diverter valve.

[0016] The data acquisition layer has temperature, pressure and speed sensors installed in the air intake, high-pressure turbine outlet and both ends of the fuel cell stack to convert physical environmental parameters into electrical signals in real time.

[0017] The simulation prediction layer (simulation module) is embedded in the control chip. It operates based on an ideal fluid dynamics model, is not affected by sudden external disturbances, and outputs theoretical benchmark data.

[0018] The compensation and correction layer (compensation module) receives the actual data from the acquisition layer and the theoretical data from the simulation layer, calculates the difference and assigns weights, and outputs the corrected target command.

[0019] The execution control layer converts the modified target command into a specific control voltage (such as a PWM signal) for the opening degree of the first and second diverter valves, directly driving the physical valves to operate.

[0020] An execution method for a turbine-stage flame fuel cell hybrid control system based on hysteresis regulation and aging compensation includes the following steps;

[0021] Step 1: Establish an interference-free temperature model through the simulation prediction layer of the intelligent control unit to set a theoretical operating benchmark for the system under ideal physical conditions;

[0022] Step 2: Based on the theoretical operating benchmark, environmental parameters are collected through the data acquisition layer of the intelligent control unit, and the environmental parameters are compensated by the compensation and correction layer of the intelligent control unit to correct them to real-time target values ​​that conform to the current flight altitude and atmospheric conditions.

[0023] Step 3: Based on the real-time target value, the compensation and correction layer of the intelligent control unit introduces an aging equation to compensate for the performance drift caused by long-term operation, ensuring that the corrected final instruction can reflect the true health status of the system and obtain accurate final instructions.

[0024] Step 4: After obtaining the precise final instruction, the execution control layer of the intelligent control unit optimizes and tunes the process using an ant colony algorithm, driving the physical actuator to overcome the physical lag of fluid transmission and achieve millisecond-level tracking of the instruction.

[0025] Step one specifically involves:

[0026] The intelligent control unit's simulation and prediction layer presets relevant parameters, including simulated high-pressure turbine outlet temperature, simulated gas turbine speed, and simulated fuel distribution ratio; it uses fluid simulation software to simulate the theoretical temperatures at the outlet of the fuel-rich combustion chamber and the inlet of the fuel stack under ideal conditions (no sudden changes in external airflow, no equipment aging). The calculation expression uses a time-dependent exponential decay function.

[0027] To reflect thermal inertia:

[0028]

[0029] in, Key parameters include high-pressure turbine speed and fuel flow rate; The heat capacity constant of the equipment is determined experimentally. The duration of the system under this condition is given by the equation, which characterizes the natural hysteresis of the system temperature with respect to time and key parameters in the absence of external disturbances.

[0030] Step two specifically involves:

[0031] The data acquisition layer of the intelligent control unit collects actual flight data through environmental sensors distributed on the fuselage and internal operating condition sensors, including the actual ambient temperature and humidity outside the fuselage (affecting air density), the actual high-pressure turbine exhaust pressure (collected through the turbine outlet pitot tube), and the actual fuel cell stack output voltage (collected through the fuel cell stack terminal voltmeter).

[0032] Define the error quantity: Let the actual collected temperature be... Calculate the deviation

[0033]

[0034] Multiple regression analysis: using temperature deviation As the dependent variable, environmental parameters (such as air pressure corresponding to flight altitude) are used as the dependent variable. ,humidity Using as independent variables, establish a linear superposition matrix; the system calculates the weight coefficients of each interference factor. Thus we get:

[0035]

[0036] This process generates the first temperature model. This refers to the corrected target temperature that the system should achieve under the current flight environment, thereby accurately identifying anomalies caused by system malfunctions or non-environmental factors.

[0037] Step three specifically involves:

[0038] The intelligent control unit's compensation and correction layer predicts the performance degradation of the fuel cell stack and combustion chamber through aging equations; the calculation logic is as follows: the system monitors the power conversion efficiency V of the fuel cell stack in real time; and records the second-order difference (i.e., the decay acceleration) of the historical efficiency V. A non-linear function of time intervals is used to construct the aging factor:

[0039]

[0040] Dynamic adjustment mechanism: When the control unit detects When the value exceeds the safety threshold (i.e., when the expected conversion efficiency declines irreversibly due to aging), the system determines that the fuel cell stack alone cannot meet the aircraft's overall electrical or thrust requirements. At this point, the control unit automatically sends an increase signal to the first diversion valve, increasing the fuel allocation from 50% to 60% for the rich combustion chamber, or directly increasing the afterburner injection rate for the lean combustion chamber. This compensation mechanism sacrifices a very small amount of fuel economy in exchange for the complete recovery of power or thrust losses caused by aging, ensuring the aircraft's safety envelope.

[0041] Step four specifically involves:

[0042] The execution control layer of the intelligent control unit constructs a hysteresis control model for the opening control of the first and second flow divider valves; due to the significant hysteresis in the airflow regulation between turbine stages, the ant colony algorithm (ACO) is used to adjust the PID parameters (proportional coefficient). Integral coefficient Differential coefficients Perform closed-loop optimization and pre-compensation calculations;

[0043] (4.1) Closed-loop optimization calculation steps:

[0044] Step (4.1.1): Temperature difference criterion and optimization trigger:

[0045] The control unit collects the expected temperature of the fuel cell stack in real time according to the sampling period. Compared with actual temperature Calculate the current temperature difference:

[0046]

[0047] If |ΔT(t)| < 0.5℃, it indicates that the current system is in thermal equilibrium, and the current PID model parameters remain unchanged; if |ΔT(t)| ≥ 0.5℃, an alarm is triggered, the system immediately clears the tabu list of the ant colony algorithm, and starts adaptive optimization of the PID parameters;

[0048] Step (4.1.2): Constructing the parameter optimization space and fitness function:

[0049] The three parameters to be optimized ( The range of values ​​for ) is discretized into grid nodes;

[0050] Define the objective cost function J to measure the control effect (the objectives are to minimize overshoot and minimize settling time):

[0051]

[0052] in, The temperature overshoot of the system response. To adjust the settling time; These are preset weighting coefficients;

[0053] Define virtual ants fitness function The lower the cost, the higher the fitness:

[0054]

[0055] Step (4.1.3): Calculation of state transition probability (path selection):

[0056] Virtual Ants From nodes in parameter space Transfer to node (i.e., choose a certain) , or The probability of discrete values The calculation formula is as follows:

[0057]

[0058] in:

[0059] The path at time t The concentration of pheromones on the surface;

[0060] This is a heuristic function, typically taken as the reciprocal of the local error;

[0061] As a factor of pheromone importance, As a heuristic factor;

[0062] gather For ants The current set of neighboring nodes that can be selected;

[0063] Step (4.1.4): Global optimal pheromone update:

[0064] After all ants have completed one traversal of the parameter combinations, the fitness is selected. The highest globally optimal parameter set, and path pheromone. Update:

[0065]

[0066] in, ρ is the pheromone evaporation coefficient (0 < ρ < 1). Where is the pheromone constant. For the ants on the path in this iteration The incremental amount of pheromones released. The algorithm iterates to a set number of times or... After convergence, output the globally optimal ( , , Parameter set;

[0067] (4.2) Feedforward-based tuning operation:

[0068] This module provides predictive compensation and enhancement based on the PID optimization results obtained in step (4.1). The system uses the ant colony algorithm in step (4.1) to determine the optimal PID parameter set for the current operating conditions. , , Following this, due to the inherent lag in feedback control (PID), this module further superimposes feedforward control variables on top of the previously optimized parameters. Its core lies in utilizing the optimal dynamic characteristics obtained through the above optimization, combined with the prediction of flight control commands, to achieve early intervention on the valve opening, thus forming a complete hysteresis control system with both feedback steady-state accuracy and feedforward dynamic prediction together with step (4.1).

[0069] Specifically as follows:

[0070] Step (4.2.1): Prediction of feedforward interference:

[0071] When an acceleration command signal is received from the flight control system At that time, the simulation prediction layer calculates the impending sudden change in airflow rate at the high-pressure turbine based on the derivative of the command.

[0072]

[0073] Considering the system's thermal inertia and pipeline transport hysteresis, calculate the future... Expected temperature disturbance at time :

[0074]

[0075] in, This is the flow gain coefficient. The thermal disturbance coefficient, The time constant for airflow transport between turbine stages;

[0076] Step (4.2.2): Calculation of feedforward control variables:

[0077] To counteract anticipated temperature disturbances in advance, the system calculates and tunes the control components. :

[0078]

[0079] in, To compensate for the gain, This is the mutation rate damping coefficient. This formula ensures that the system generates valve action commands to suppress overshoot before the temperature rises.

[0080] Step (4.2.3): Calculation of the final valve opening control law:

[0081] The optimal PID control input obtained by the ant colony optimization algorithm is combined with the feedforward control input to calculate the total control opening command for the second flow divider valve (or the first flow divider valve). :

[0082] The system introduces an error prediction mechanism to reduce the pre-prediction error.

[0083]

[0084] The formula for the main valve opening control law is:

[0085]

[0086] The control module will include the above. This is converted into a corresponding PWM drive voltage, which directly drives the shunt. Through this tuning, the system successfully bypasses the inherent defect of conventional PID control that requires waiting for error accumulation, ensuring that the SOFC stack remains strictly constrained within the absolute safe range of 600-1000 ℃ even under sudden changes in operating conditions.

[0087] The beneficial effects of this invention are:

[0088] This invention combines a non-interference model with interference compensation to effectively filter airflow disturbances caused by changes in flight attitude and control temperature fluctuations within ±0.5 ℃. Specifically, the system uses the high-pressure turbine outlet state as a feedforward variable for predictive control, eliminating interference from ambient temperature, humidity, and air pressure. This physically avoids the fatal thermal stress tearing caused by alternating hot and cold airflows to the fragile components of the fuel cell stack under drastic operating conditions, significantly reducing damage to the fuel cell stack due to thermal fatigue.

[0089] This invention introduces an aging equation, which enables the system to automatically increase the fuel supply to the lean combustion chamber when the stack performance degrades (e.g., after thousands of hours of operation), ensuring that the thrust of the low-pressure turbine does not decrease with stack aging, thus achieving decoupling protection of "thrust-electricity".

[0090] This invention, through the tuning of a hysteresis control model, allows the system to adjust before temperature deviations occur, instead acting proactively based on prediction. Due to the physical delay in fluid transfer between turbine stages, this system employs an ant colony algorithm to adaptively optimize PID parameters. Upon sensing a flight control acceleration command, it directly drives the splitter opening to a preset optimal value. This is crucial for aero-engines, as flight safety does not allow for millisecond-level power interruptions or surges. Attached Figure Description

[0091] Figure 1 This is a diagram of the overall system architecture and control logic.

[0092] Figure 2 The flowchart shows the non-interference temperature model and compensation algorithm.

[0093] Figure 3 A schematic diagram of the hysteresis control logic optimized for the ant colony algorithm. Detailed Implementation

[0094] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0095] like Figure 1 The overall system architecture is shown in Figure 1.

[0096] The system mainly consists of physical hardware and intelligent control units;

[0097] 1. Physical hardware:

[0098] Along the airflow direction, the components are arranged in sequence: intake duct, compressor, combustion chamber, high-pressure turbine, rich combustion chamber, fuel cell stack, lean combustion chamber, low-pressure turbine, and exhaust nozzle. It includes a fuel tank, fuel pump, first split valve (regulating total fuel distribution), and second split valve (regulating oxidizer distribution).

[0099] Explanation of the connection relationships and functions of each component:

[0100] (1) Gas path connection: Outside air enters the compressor through the intake duct and is pressurized. It then enters the main combustion chamber and mixes with the fuel for combustion. The resulting high-temperature, high-pressure gas first drives the high-pressure turbine to do work (driving the compressor). The interstage gas discharged from the high-pressure turbine is divided into two paths when it enters the second split valve: the main flow path enters the rich combustion chamber, and the bypass flow path goes directly to the lean combustion chamber. The product of the rich combustion chamber (hydrogen-rich synthesis gas) enters the anode of the fuel cell stack for electrochemical power generation; the exhaust gas discharged from the stack and the bypass flow flow merge in the lean combustion chamber, undergo secondary combustion, expand and do work to drive the low-pressure turbine (driving the fan or output shaft), and finally exit through the tail nozzle.

[0101] (2) Fuel circuit connection: Fuel is drawn from the storage tank by the fuel pump and enters the first diversion valve. The first diversion valve injects the total fuel into the rich combustion chamber (for reforming to produce hydrogen) and the lean combustion chamber (for supplemental combustion and boost) according to the command ratio.

[0102] 2. Intelligent control unit:

[0103] It includes a data acquisition layer, a simulation prediction layer, a compensation and correction layer, and an execution control layer.

[0104] The interaction between the control unit and the hardware:

[0105] (1) Data acquisition layer: Temperature, pressure and speed sensors are arranged in the air intake, high-pressure turbine outlet and both ends of the fuel cell stack to convert physical environmental parameters into electrical signals in real time.

[0106] (2) Simulation prediction layer (simulation module): Embedded in the control chip, it operates based on an ideal fluid dynamics model, is not affected by sudden external interference, and outputs theoretical benchmark data.

[0107] (3) Compensation Correction Layer (Compensation Module): Receives the actual data from the acquisition layer and the theoretical data from the simulation layer, calculates the difference and assigns weights, and outputs the corrected target instruction.

[0108] (4) Execution control layer (control module): The modified target command is converted into a specific control voltage (such as PWM signal) for the opening degree of the first and second diversion valves, which directly drives the physical valves to operate.

[0109] like Figure 2 As shown, the control flow and principle are as follows:

[0110] This control flow is the core neural hub of the aforementioned system. Step one establishes a baseline, step two corrects for interference, step three compensates for aging, and step four executes dynamic regulation. These four steps work together to form a complete data flow closed loop. Its complete execution steps are as follows:

[0111] Step 1: Constructing a non-interference temperature model in the turbine stage environment:

[0112] The control unit's simulation module presets relevant parameters (simulating high-pressure turbine outlet temperature, simulating gas turbine speed, and simulating fuel distribution ratio). Fluid simulation software is used to simulate the theoretical temperatures at the outlet of the fuel-rich combustion chamber and the inlet of the fuel stack anode under ideal conditions (no sudden changes in external airflow, no equipment aging). The calculation expression uses a time-dependent exponential decay function.

[0113] To reflect thermal inertia:

[0114]

[0115] in, Key parameters include high-pressure turbine speed and fuel flow rate; The heat capacity constant of the equipment is determined experimentally. The duration of the system under this operating condition is given. This equation characterizes the natural hysteresis characteristics of the system temperature over time and changes in key parameters without external disturbances.

[0116] Step 2: Weight calculation and compensation of multidimensional interference parameters:

[0117] The compensation module collects actual flight data through environmental sensors distributed on the fuselage and internal operating condition sensors, including the actual ambient temperature and humidity outside the fuselage (which affect air density), the actual high-pressure turbine exhaust pressure (collected through the turbine outlet pitot tube), and the actual fuel cell stack output voltage (collected through the fuel cell stack terminal voltmeter).

[0118] Define the error quantity: Let the actual collected temperature be... Calculate the deviation

[0119]

[0120] Multiple regression analysis: using temperature deviation As the dependent variable, environmental parameters (such as air pressure corresponding to flight altitude) are used as the dependent variable. ,humidity Using () as independent variables, a linear superposition matrix is ​​established. The system calculates the weight coefficients of each interference factor. Thus we get:

[0121]

[0122] This process generates the first temperature model. This refers to the corrected target temperature that the system should achieve under the current flight environment, thereby accurately identifying anomalies caused by system malfunctions or non-environmental factors.

[0123] Step 3: Efficiency correction based on the aging equation:

[0124] The system predicts the performance degradation of the fuel cell stack and combustion chamber using aging equations.

[0125] Calculation logic: The system monitors the power conversion efficiency V of the fuel cell stack in real time. It records the second-order difference (i.e., the decay acceleration) of the historical efficiency V. A nonlinear function of time intervals is used to construct the aging factor.

[0126]

[0127] Dynamic adjustment mechanism: When the control unit detects When the value exceeds the safety threshold (i.e., when the expected conversion efficiency declines irreversibly due to aging), the system determines that the fuel cell stack alone cannot meet the aircraft's overall electrical or thrust requirements. At this point, the control unit automatically sends an increase signal to the first diversion valve, increasing the fuel allocation from 50% to 60% for the rich combustion chamber, or directly increasing the afterburner injection rate for the lean combustion chamber. This compensation mechanism sacrifices a very small amount of fuel economy in exchange for the complete recovery of power or thrust losses caused by aging, ensuring the aircraft's safety envelope.

[0128] Step 4: Hysteresis Control Based on Ant Colony Algorithm (see Appendix) Figure 3 )

[0129] The control module constructs a hysteresis control model for controlling the opening of the first and second flow divider valves. Due to the significant hysteresis in the inter-turbine airflow regulation, an ant colony optimization (ACO) algorithm is used to adjust the PID parameters (proportional coefficient). Integral coefficient Differential coefficients Perform closed-loop optimization and pre-compensation calculations.

[0130] (4.1) Closed-loop optimization calculation steps:

[0131] Step (4.1.1): Temperature difference criterion and optimization trigger:

[0132] The control unit collects the expected temperature of the fuel cell stack in real time according to the sampling period. Compared with actual temperature Calculate the current temperature difference:

[0133]

[0134] If |ΔT(t)| < 0.5℃, it indicates that the current system is in thermal equilibrium, and the current PID model parameters remain unchanged; if |ΔT(t)| ≥ 0.5℃, an alarm is triggered, the system immediately clears the tabu list of the ant colony algorithm, and starts adaptive optimization of the PID parameters.

[0135] Step (4.1.2): Constructing the parameter optimization space and fitness function:

[0136] The three parameters to be optimized ( The range of values ​​for ) is discretized into grid nodes.

[0137] Define the objective cost function J to measure the control effect (the objectives are to minimize overshoot and minimize settling time):

[0138]

[0139] in, The temperature overshoot of the system response. To adjust the settling time; These are the preset weighting coefficients.

[0140] Define virtual ants fitness function The lower the cost, the higher the fitness:

[0141]

[0142] Step (4.1.3): Calculation of state transition probability (path selection):

[0143] Virtual Ants From nodes in parameter space Transfer to node (i.e., choose a certain) , or The probability of discrete values The calculation formula is as follows:

[0144]

[0145] in:

[0146] The path at time t The concentration of pheromones on the surface;

[0147] This is a heuristic function, typically taken as the reciprocal of the local error;

[0148] As a factor of pheromone importance, As a heuristic factor;

[0149] gather For ants The set of neighboring nodes that can be selected at present.

[0150] Step (4.1.4): Global optimal pheromone update:

[0151] After all ants have completed one traversal of the parameter combinations, the fitness is selected. The highest globally optimal parameter set, and path pheromone. Update:

[0152]

[0153] in, ρ is the pheromone evaporation coefficient (0 < ρ < 1). Where is the pheromone constant. For the ants on the path in this iteration The incremental amount of pheromones released. The algorithm iterates to a set number of times or... After convergence, output the globally optimal ( , , ) Parameter set.

[0154] (4.2) Feedforward-based tuning operation:

[0155] This module performs predictive compensation and enhancement based on the PID optimization results obtained in step (4.1). Specifically:

[0156] Step (4.2.1): Prediction of feedforward interference:

[0157] When an acceleration command signal is received from the flight control system At that time, the simulation prediction layer calculates the impending sudden change in airflow rate at the high-pressure turbine based on the derivative of the command.

[0158]

[0159] Considering the system's thermal inertia and pipeline transport hysteresis, calculate the future... Expected temperature disturbance at time :

[0160]

[0161] in, This is the flow gain coefficient. The thermal disturbance coefficient, This is the time constant for airflow transport between turbine stages.

[0162] Step (4.2.2): Calculation of feedforward control variables:

[0163] To counteract anticipated temperature disturbances in advance, the system calculates and tunes the control components. :

[0164]

[0165] in, To compensate for the gain, This is the mutation rate damping coefficient. This formula ensures that the system generates valve action commands to suppress overshoot before the temperature rises.

[0166] Step (4.2.3): Calculation of the final valve opening control law:

[0167] The optimal PID control input obtained by the ant colony optimization algorithm is combined with the feedforward control input to calculate the total control opening command for the second flow divider valve (or the first flow divider valve). :

[0168] The system introduces an error prediction mechanism to reduce the pre-prediction error.

[0169]

[0170] The formula for the main valve opening control law is:

[0171]

[0172] The control module will include the above. This is converted into a corresponding PWM drive voltage, which directly drives the shunt. Through this tuning, the system successfully bypasses the inherent defect of conventional PID control that requires waiting for error accumulation, ensuring that the SOFC stack remains strictly constrained within the absolute safe range of 600-1000 ℃ even under sudden changes in operating conditions.

[0173] Example 1: Takeoff and landing control of a new generation hybrid-electric mainline passenger aircraft (MEA):

[0174] During takeoff and acceleration climb of a single-aisle passenger aircraft at full load, megawatt-level (MW) electrical energy needs to be extracted from the engine shaft to drive the distributed ducted fans. Traditional methods would lead to a rapid decrease in the main compressor surge margin. With this system, during takeoff, the control unit diverts high-pressure turbine stage exhaust gas into the SOFC (Solar-Fired Fuel Cell) for power generation via a second diversion valve. Under this drastic operating condition, the ant colony algorithm predicts airflow impact in milliseconds, adjusting the rich-fuel combustor diversion opening 1.5 seconds in advance, and performing strong afterburning in the lean-fuel combustor with the first diversion valve fully open to maintain maximum thrust. This not only prevents fuel cell stack thermal shock and engine surge but also significantly reduces NOx emissions during takeoff.

[0175] Example 2: Extreme cruise of a high-altitude long-endurance unmanned reconnaissance aircraft (HALE UAV):

[0176] This UAV conducted a two-week cruise mission at an altitude of over 20,000 meters. Due to the thin air (low pressure) and extremely low temperature (below -50°C) at high altitudes, traditional fuel cell stacks are highly susceptible to thermal imbalance. This system's compensation module monitors high-altitude air pressure and temperature in real time, initiating an interference suppression algorithm to correct the interference-free temperature model. When the UAV crosses different tropospheres, no manual intervention from the pilot is required; the system automatically adjusts the second diversion valve to control the mixing ratio of hot and cold airflow, maintaining optimal fuel cell stack efficiency. On the 12th day of the mission, the system detected natural performance degradation in the fuel cell stack (consistent with the aging equation's prediction). It then automatically increased the flow from the first diversion valve to the lean combustion chamber, increasing fuel supply and ensuring the UAV achieved full tactical maneuver thrust for a safe return. Simultaneously, the efficient response of the interstage SOFC reduced exhaust gas temperature, maintaining excellent low-infrared stealth characteristics.

[0177] Example 3: Ground-based distributed micro gas turbine cogeneration system:

[0178] In microgrid applications within data centers, a "rich-fuel reforming-SOFC power generation-lean-fuel power generation" cycle is constructed using hydrogen-blended natural gas pipelines, forming a micro gas-steam combined cycle (CCGT). When data center servers are fully loaded, causing severe fluctuations in grid load, the ant colony algorithm's hysteresis control module translates the logic for aviation airflow impacts into a response to sudden changes in grid load. The system achieves "tuning," adjusting the proportion of natural gas entering the fuel cell stack within seconds to quickly smooth out grid peaks; and at the end of the equipment's lifespan, aging equations are used to compensate for power output, extending unit replacement cycles and reducing operating costs.

Claims

1. A turbine-stage flame fuel cell hybrid control system based on hysteresis regulation and aging compensation, characterized in that, The components, arranged sequentially along the airflow direction, include the air intake, compressor, combustion chamber, high-pressure turbine, fuel-rich combustion chamber, fuel cell stack, lean combustion chamber, low-pressure turbine, and exhaust nozzle. It also includes a fuel storage tank, a fuel pump, a first diversion valve, a second diversion valve, and an intelligent control unit; Outside air enters the compressor through the intake duct and is pressurized. It then enters the main combustion chamber to mix and burn with the fuel. The resulting high-temperature, high-pressure gas first drives the high-pressure turbine to perform work. The interstage gas discharged from the high-pressure turbine is split into two paths upon entering the second split valve: The main flow path enters the rich combustion chamber, while the bypass flow path leads directly to the lean combustion chamber. The products from the rich combustion chamber enter the anode of the fuel cell stack for electrochemical power generation. The exhaust gas from the stack and the bypass flow converge in the lean combustion chamber, undergo secondary combustion, expand and drive the low-pressure turbine, and finally exit through the tail nozzle. Fuel is drawn from the fuel storage tank by the fuel pump and enters the first diversion valve. The first diversion valve injects the total fuel into the rich combustion chamber and the lean combustion chamber according to the command ratio. The intelligent control unit includes a data acquisition layer, a simulation prediction layer, a compensation and correction layer, and an execution control layer. The data acquisition layer is connected to sensor signals arranged in the air intake, the high-pressure turbine outlet and both ends of the fuel cell stack, and is used to collect physical environment and state parameters. The simulation prediction layer is used to output theoretical baseline data based on the embedded fluid dynamics model; The compensation and correction layer is connected to the data acquisition layer and the simulation prediction layer respectively, and is used to calculate the corrected target instruction based on the difference between the physical environment and state parameters and the theoretical benchmark data. The execution control layer is electrically connected to the compensation correction layer, the first diverter valve, and the second diverter valve, respectively, and is used to convert the target command into a drive voltage signal to adjust the opening degree of the first diverter valve and the second diverter valve.

2. The turbine-stage flame fuel cell hybrid control system based on hysteresis regulation and aging compensation according to claim 1, characterized in that, The data acquisition layer has temperature, pressure and speed sensors installed in the air intake, high-pressure turbine outlet and both ends of the fuel cell stack to convert physical environmental parameters into electrical signals in real time. The simulation prediction layer is embedded in the control chip, operates based on an ideal fluid dynamics model, is unaffected by sudden external disturbances, and outputs theoretical benchmark data; The compensation and correction layer receives the actual data from the acquisition layer and the theoretical data from the simulation layer, calculates the difference and assigns weights, and outputs the corrected target command. The execution control layer translates the modified target command into specific control voltages for the opening degrees of the first and second diversion valves, directly driving the physical valves to operate.

3. An execution method for a turbine-stage flame fuel cell hybrid control system based on hysteresis control and aging compensation as described in any one of claims 1-2, characterized in that, Includes the following steps: Step 1: Establish an interference-free temperature model through the simulation prediction layer of the intelligent control unit to set a theoretical operating benchmark for the system under ideal physical conditions; Step 2: Based on the theoretical operating benchmark, environmental parameters are collected through the data acquisition layer of the intelligent control unit, and the environmental parameters are compensated by the compensation and correction layer of the intelligent control unit to correct them to real-time target values ​​that conform to the current flight altitude and atmospheric conditions. Step 3: Based on the real-time target value, the compensation and correction layer of the intelligent control unit introduces an aging equation to compensate for the performance drift caused by long-term operation, ensuring that the corrected final instruction can reflect the true health status of the system and obtain accurate final instructions. Step 4: After obtaining the precise final instruction, the execution control layer of the intelligent control unit optimizes and tunes the process using an ant colony algorithm, driving the physical actuator to overcome the physical lag of fluid transmission and achieve millisecond-level tracking of the instruction.

4. The execution method of the turbine-stage flame fuel cell hybrid control system based on hysteresis control and aging compensation according to claim 3, characterized in that, Step one specifically involves: The intelligent control unit's simulation and prediction layer presets relevant parameters, including the simulated high-pressure turbine outlet temperature, the simulated gas turbine speed, and the simulated fuel distribution ratio; it uses fluid simulation software to simulate the theoretical temperatures at the outlet of the fuel-rich combustion chamber and the inlet of the fuel cell stack under ideal conditions. ; The calculation expression uses a time-dependent exponential decay function. To reflect thermal inertia: ； in, Key parameters include high-pressure turbine speed and fuel flow rate; The heat capacity constant of the equipment is determined experimentally. This refers to the duration during which the system operates under this condition. This equation characterizes the natural hysteresis of system temperature changes with time and key parameters in the absence of external disturbances.

5. The execution method of the turbine-stage flame fuel cell hybrid control system based on hysteresis control and aging compensation according to claim 3, characterized in that, Step two specifically involves: The data acquisition layer of the intelligent control unit collects actual flight data through environmental sensors distributed on the fuselage and internal operating condition sensors, including the actual ambient temperature and humidity outside the fuselage, the actual high-pressure turbine exhaust pressure, and the actual fuel cell output voltage. Define the error quantity: Let the actual collected temperature be... Calculate the deviation ； Multiple regression analysis: using temperature deviation With environmental parameters as independent variables and environmental parameters as dependent variables, a linear superposition matrix is ​​established; the system calculates the weighting coefficients of each disturbance factor. Thus we get: ； This process generates the first temperature model. This refers to the corrected target temperature that the system should achieve under the current flight environment, thereby accurately identifying anomalies caused by system malfunctions or non-environmental factors.

6. The execution method of the turbine-stage flame fuel cell hybrid control system based on hysteresis control and aging compensation according to claim 3, characterized in that, Step three specifically involves: The intelligent control unit's compensation and correction layer predicts the performance degradation of the fuel cell stack and combustion chamber through aging equations; the calculation logic is as follows: the system monitors the power conversion efficiency V of the fuel cell stack in real time; and records the second-order difference of the historical efficiency V, i.e., the decay acceleration. An aging factor is constructed using a non-linear function of time intervals: ； Dynamic adjustment mechanism: When the control unit detects If the value exceeds the safety threshold, the system determines that the fuel cell stack alone cannot meet the power or thrust requirements of the entire machine. At this time, the control unit automatically sends an increase signal to the first diversion valve to increase the amount of fuel originally allocated to the rich combustion chamber from 50% to 60%, or directly increases the amount of afterburner injection in the lean combustion chamber.

7. The execution method of the turbine-stage flame fuel cell hybrid control system based on hysteresis control and aging compensation according to claim 3, characterized in that, Step four specifically involves: The execution control layer of the intelligent control unit constructs a hysteresis control model for the opening control of the first and second flow divider valves; due to the significant hysteresis in the airflow regulation between turbine stages, an ant colony algorithm is used to adjust the PID parameters and proportional coefficients. Integral coefficient Differential coefficients Perform closed-loop optimization and pre-compensation calculations; (4.1) Closed-loop optimization calculation steps: Step (4.1.1): Temperature difference criterion and optimization trigger: The control unit collects the expected temperature of the fuel cell stack in real time according to the sampling period. Compared with actual temperature Calculate the current temperature difference: ； If |ΔT(t)| < 0.5℃, it indicates that the current system is in thermal equilibrium, and the current PID model parameters remain unchanged; if |ΔT(t)| ≥ 0.5℃, an alarm is triggered, the system immediately clears the tabu list of the ant colony algorithm, and starts adaptive optimization of the PID parameters; Step (4.1.2): Constructing the parameter optimization space and fitness function: The three parameters to be optimized ( The range of values ​​for ) is discretized into grid nodes; Define the target cost function J to measure the control effect: ； in, The temperature overshoot of the system response. To adjust the settling time; These are preset weighting coefficients; Define virtual ants fitness function The lower the cost, the higher the fitness: ； Step (4.1.3): Calculation of state transition probability: Virtual Ants From nodes in parameter space Transfer to node (i.e., choose a certain) , or The probability of discrete values The calculation formula is as follows: ； in: The path at time t The concentration of pheromones on the surface; This is a heuristic function, typically taken as the reciprocal of the local error; As a factor of pheromone importance, As a heuristic factor; gather For ants The current set of neighboring nodes that can be selected; Step (4.1.4): Global optimal pheromone update: After all ants have completed one traversal of the parameter combinations, the fitness is selected. The highest globally optimal parameter set, and path pheromone. Update: ； in, ρ is the pheromone evaporation coefficient (0 < ρ < 1). Where is the pheromone constant. For the ants on the path in this iteration The pheromone increment released; the algorithm iterates to a set number of times or After convergence, output the globally optimal ( , , Parameter set; (4.2) Feedforward-based tuning operation: Step (4.2.1): Prediction of feedforward interference: When an acceleration command signal is received from the flight control system At that time, the simulation prediction layer calculates the impending sudden change in airflow rate at the high-pressure turbine based on the derivative of the command. ; ； Considering the system's thermal inertia and pipeline transport hysteresis, calculate the future... Expected temperature disturbance at time : ； in, This is the flow gain coefficient. The thermal disturbance coefficient, The time constant for airflow transport between turbine stages; Step (4.2.2): Calculation of feedforward control variables: To counteract anticipated temperature disturbances in advance, the system calculates and tunes the control components. : ； in, To compensate for the gain, This is the damping coefficient for the mutation rate; this formula ensures that the system generates valve action commands to suppress overshoot before the temperature rises. Step (4.2.3): Calculation of the final valve opening control law: The optimal PID control input obtained by the ant colony optimization algorithm is combined with the feedforward control input to calculate the total control opening command of the second diverter valve. : The system introduces an error prediction mechanism to reduce the pre-prediction error. ； The formula for the main valve opening control law is: ； The control module will include the above. This is converted into the corresponding PWM drive voltage, which directly drives the shunt to operate. Through this tuning, the system successfully bypasses the inherent defect of conventional PID control that requires waiting for error accumulation, ensuring that the SOFC stack remains strictly constrained within the absolute safe range of 600-1000 ℃ even under sudden changes in operating conditions.

Images