Fuel cell anode pressure fluctuation suppression method and system

Abstract

The invention provides a fuel cell anode pressure fluctuation suppression method and system, and relates to the technical field of battery intelligent control, and the method comprises the steps: carrying out the real-time sensing of the stack operation condition of a target fuel cell, and obtaining the stack operation state parameters; performing critical working condition identification, pressure trend prediction and adaptive model prediction control based on the operation state parameters of the electric pile, and deciding an optimal control instruction for balancing pressure stability, action smoothness and system energy efficiency; the optimal control instruction is converted into the opening degree of a proportional valve at the front end of an ejector and the rotating speed of an anode circulating pump, and the flow of hydrogen entering an anode and the amount of recycled gas are finally adjusted by changing the opening degree and the rotating speed, so that the anode pressure fluctuation of the target fuel cell is inhibited; through multi-level information fusion and decision optimization, accurate and smooth control of the anode pressure is realized, and the dynamic performance and reliability of the fuel cell system are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery intelligent control, and in particular to a fuel cell anode pressure fluctuation suppression method and system. BACKGROUND

[0002] As a clean and efficient energy conversion device, the anode hydrogen recirculation system of a proton exchange membrane fuel cell (PEMFC) is crucial for improving hydrogen utilization and preventing flooding. The mixed recirculation scheme of an ejector-circulating pump has high operating efficiency and has become the mainstream technology.

[0003] However, this scheme has inherent technical problems in the critical load interval where the ejector begins to take effect: the system faces a working mode switch from a circulating pump-dominated to an ejector-dominated mode. In this critical region, the ejector flow and the pump-driven gas flow are coupled with each other, and the dynamic characteristics show strong nonlinearity and time-varying nature, resulting in fluctuations in the anode pressure. In the prior art, the control is mainly realized based on a PID control algorithm, and the controller parameters are fixed and unchanged. When the system operates in the critical condition, the control algorithm cannot adapt to the rapidly changing dynamic characteristics, and the anode pressure oscillates, resulting in a decline in control performance.

[0004] In addition, as the power core of a vehicle, the controller of a fuel cell system needs to be integrated into an embedded control unit (ECU) for vehicles. This environment imposes extremely stringent requirements on the real-time performance, reliability, and computational resource occupation of the control algorithm. Many optimization algorithms that work well in simulation or laboratory environments often cannot be directly deployed on vehicle controllers with single-chip microcomputers (MCUs) as the core and limited resources due to their complex calculations and long iteration times, which greatly reduces their engineering application value.

[0005] Therefore, there is a control difficulty in the pressure fluctuations of the anode hydrogen recirculation of a proton exchange membrane fuel cell during the critical load switching, and the stringent resource limitations of the embedded system for vehicles make it a key technical challenge to develop an intelligent collaborative management method that takes into account high performance and real-time performance. SUMMARY

[0006] To solve the above problems, the present application proposes a fuel cell anode pressure fluctuation suppression method and system, which realizes precise and smooth control of the anode pressure through multi-level information fusion and decision optimization, and significantly improves the dynamic performance and reliability of the fuel cell system.

[0007] According to some embodiments, the present application adopts the following technical solutions: A fuel cell anode pressure fluctuation suppression method, comprising: real-time sensing of the stack operating conditions of a target fuel cell to obtain stack operating state parameters; Based on the stack operating state parameters, critical condition identification, pressure trend prediction and adaptive model predictive control are performed to determine optimal control instructions for balance pressure stability, action smoothness and system energy efficiency; The optimal control instructions are converted into the opening degree of the proportional valve at the front end of the ejector and the rotating speed of the anode circulating pump, and by changing the opening degree and the rotating speed, the hydrogen flow rate entering the anode and the recirculation gas amount are finally adjusted to suppress the anode pressure fluctuation of the target fuel cell.

[0008] According to some embodiments, the present application adopts the technical solutions as follows: A fuel cell anode pressure fluctuation suppression system comprises: A state perception module is configured to perceive the stack operation of a target fuel cell in real time to obtain stack operating state parameters; An instruction decision module is configured to perform critical condition identification, pressure trend prediction and adaptive model predictive control based on the stack operating state parameters to determine optimal control instructions for balance pressure stability, action smoothness and system energy efficiency; An instruction execution module is configured to convert the optimal control instructions into the opening degree of the proportional valve at the front end of the ejector and the rotating speed of the anode circulating pump, and by changing the opening degree and the rotating speed, the hydrogen flow rate entering the anode and the recirculation gas amount are finally adjusted to suppress the anode pressure fluctuation of the target fuel cell.

[0009] According to some embodiments, the present application adopts the technical solutions as follows: A computer program product comprises a computer program, which, when executed by a processor, implements the fuel cell anode pressure fluctuation suppression method.

[0010] According to some embodiments, the present application adopts the technical solutions as follows: A non-transitory computer readable storage medium is used to store computer instructions, which, when executed by a processor, implement the fuel cell anode pressure fluctuation suppression method.

[0011] According to some embodiments, the present application adopts the technical solutions as follows: An electronic device comprises a processor, a memory and a computer program; the processor is connected with the memory, and the computer program is stored in the memory; when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the fuel cell anode pressure fluctuation suppression method.

[0012] Compared with the prior art, the present application has the beneficial effects that: 1. By introducing a critical condition recognizer based on multi-feature fuzzy fusion and a parameter-adjustable high-precision prediction model, the application realizes accurate perception of the pressure change trend of the anode; compared with the traditional fixed threshold switching method or single parameter model prediction control, the application can identify the trend of the system entering the critical condition in advance, reduce the pressure prediction error; this enables the model prediction controller to make optimization decisions based on more accurate prediction information, thereby reducing the amplitude of the anode pressure fluctuation during the ejector / circulating pump switching process.

[0013] 2. The double-layer parameter self-adaptive mechanism used in the application solves the core problem that the fixed parameter controller cannot adapt to the strong nonlinearity and time-varying characteristics of the fuel cell; the first layer of parameter preset based on condition recognition enables the controller to quickly switch control strategies according to the current operating mode; the second layer of feedback fine-tuning based on instantaneous performance can continuously compensate for model mismatch and external disturbances, improving the operating condition adaptability and robustness of the control algorithm.

[0014] 3. The application, through the model prediction control framework of multiple inputs and multiple outputs, simultaneously incorporates the control quantities of the two actuators into the optimization objective function and the constraint conditions for collaborative optimization; the control smoothing term in the optimization objective explicitly penalizes the drastic change of the control quantity, and in combination with the adaptive mechanism, the slow opening of the ejector valve and the smooth reduction of the circulating pump speed can be realized; this fundamentally avoids the mutation of the control command, realizes truly disturbance-free smooth switching, and eliminates the pressure impact and system oscillation caused by rigid switching.

[0015] 4. The application guarantees safe and reliable operation of the system through multiple mechanisms, first, the state constraints (pressure safety boundary) and actuator constraints (stroke limit and change rate limit) are explicitly added to the optimization problem of the model prediction control, ensuring that the control system command is within the safe range, effectively preventing overpressure, actuator saturation and other risks; second, the parameter adaptive mechanism automatically adopts more conservative control parameters when detecting an increase in model mismatch or a drastic disturbance, prioritizing system stability to prevent control divergence. This inherent safety design greatly reduces the probability of system failure due to control failure.

[0016] 5. An economy term is introduced into the optimization objective function to penalize the power consumption of the circulating pump, guiding the system to preferentially use the ejector with lower energy consumption under the premise of meeting performance requirements. By adaptively adjusting λ, the system can automatically enter a more economical operating mode when the dynamic performance requirement is not high. Compared with the traditional fixed-time or threshold control strategy, the application can save auxiliary system energy consumption, and at the same time, due to more accurate on-demand control, unnecessary hydrogen purge loss is reduced, and the overall hydrogen utilization rate is improved.

[0017] 6. The adaptive mechanism of the present application reduces the dependence on the accuracy of the initial parameters of the controller; although the underlying algorithm is complex, by using engineering implementation means such as look-up table, piecewise linear approximation, etc., and taking advantage of the computing power of modern embedded processors, the algorithm can be completely implemented on a cost-controllable hardware platform in real time (sampling period ≤ 10 ms); at the same time, the control algorithm has self-diagnosis and parameter self-learning capabilities, reducing the burden of on-site debugging and maintenance, and is conducive to the promotion and application of the control technology in large-scale commercial fuel cell vehicles. BRIEF DESCRIPTION OF DRAWINGS

[0018] The accompanying drawings, which form a part of this specification, are included to provide a further understanding of the application, and are incorporated in and constitute a part of this specification. The embodiments of these drawings are set to explain the application, and do not constitute an improper limitation on the application.

[0019] Figure 1 It is a schematic diagram of the overall structure of the fuel cell anode hydrogen supply and recycling system in Example 1.

[0020] Figure 2 It is a hierarchical technology architecture and information flow diagram of the control algorithm in Example 2.

[0021] Figure 3 It is a working principle diagram of the critical condition recognizer in Example 2.

[0022] Figure 4 It is a logic composition diagram of the double-layer parameter adaptive mechanism in Example 2.

[0023] Figure 5 It is a complete online rolling optimization flowchart of the integrated adaptive model predictive control in Example 2. DETAILED DESCRIPTION

[0024] The present application will be further described below in conjunction with the drawings and examples.

[0025] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0026] It should be noted that the terms used herein are only for the purpose of describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and it should also be understood that when the terms "comprise" and / or "comprising" are used in the specification, there is a presence of the features, steps, operations, devices, components and / or combinations thereof.

[0027] Example 1 An embodiment of the present application provides a fuel cell anode pressure fluctuation suppression method, comprising: Step S1: real-time sensing of stack operation of a target fuel cell is performed to obtain a stack operation state parameter; Step S2: based on the stack operation state parameter, critical condition identification, pressure trend prediction and adaptive model prediction control are performed to determine optimal control instructions of balance pressure stability, action smoothness and system energy efficiency; Step S3: the optimal control instructions are converted into a proportional valve opening degree of a front end of an ejector and a rotating speed of an anode circulating pump, and by changing the opening degree and the rotating speed, hydrogen flow entering the anode and recirculation gas amount are finally adjusted to suppress anode pressure fluctuation of the target fuel cell.

[0028] As an embodiment, the fuel cell anode pressure fluctuation suppression method of the present application aims to solve the problem of pressure fluctuation caused by the anode ejector and circulating pump of a proton exchange membrane fuel cell in a critical switching condition, and the core is to construct a closed-loop control framework with real-time state sensing, high-precision dynamic prediction and intelligent parameter adaptive capability. Through multi-level information fusion and decision optimization, the framework realizes accurate and smooth control of the anode pressure, significantly improves the dynamic performance and reliability of the fuel cell system, and the specific implementation method is described below.

[0029] I. Proton exchange membrane fuel cell (PEMFC) and anode hydrogen recirculation system Figure 1 The overall hardware structure and gas path connection relationship of the fuel cell anode hydrogen supply and recirculation system involved in the embodiment are shown. Figure 1 As shown in the figure, the system is composed of a hydrogen tank, a solenoid valve, a proportional valve, a water separator, an ejector, a circulating pump and a stack anode side, wherein the hydrogen tank, the solenoid valve and the proportional valve belong to the anode hydrogen supply system; the water separator, the ejector and the circulating pump belong to the recirculation system.

[0030] High-pressure hydrogen gas is decompressed, one way as a driving flow into the ejector, using its jet effect to produce negative pressure to suck the circulating hydrogen gas at the anode outlet; the other way as a supplementary flow, the ejector and the circulating pump are installed in parallel, and the circulating pump ensures hydrogen flow in low load conditions to prevent water flooding; the gas after the stack reaction is separated, most of it is recovered, and a small amount of waste gas is discharged through the purge valve; pressure and temperature sensors are arranged at key nodes to monitor the system state in real time; all sensor data are sent to the control unit, and the control unit accurately adjusts the proportional valve opening degree of the front end of the ejector and the rotating speed of the circulating pump according to the method of the embodiment to realize intelligent collaborative management of the anode pressure.

[0031] II. Overall architecture design and control process The framework of the method of the embodiment is composed of three functionally clear levels: a perception layer, a decision-making layer, and an execution layer, ensuring efficient coordination of data acquisition, information processing, and control execution.

[0032] 1. The perception layer is responsible for collecting key physical quantities reflecting the operating state of the stack. The sensor array deployed in this layer includes: a current sensor, which is connected in series to the output main circuit of the stack (usually located between the positive or negative output of the stack and the load / inverter), for real-time monitoring of the stack load current which is the most direct parameter for judging the system operating condition; pressure sensors, including an anode inlet pressure sensor and an anode outlet pressure sensor, the anode inlet pressure sensor being installed at the anode inlet of the stack for measuring the hydrogen pressure entering the stack whose dynamic change is the core target of control; the anode outlet pressure sensor is installed at the anode outlet of the stack for measuring the pressure of the hydrogen after reaction which can be used for system monitoring and diagnosis; temperature sensors, which are directly installed on the cooling liquid outlet end of the stack or the stack body, for monitoring the stack temperature because temperature changes directly affect the gas state and reaction rate.

[0033] After all sensor signals are filtered, amplified, and analog-to-digital converted by the signal conditioning circuit, they are sent to the decision-making layer.

[0034] 2. The decision-making layer is composed of a high-performance embedded microcontroller, which runs the control algorithm proposed in the embodiment.

[0035] The algorithm software adopts modular design, with three core functional modules integrated internally: a working condition identifier, an anode pressure dynamic prediction model, and a self-adaptive model predictive controller; these modules cooperate to complete the complete control process from state judgment, trend prediction to optimal decision-making, which are described as follows: 2.1 Critical operating condition identifier based on multi-feature fuzzy fusion Accurate identification of whether the system enters the critical operating condition of the ejector is the premise of effective control; the traditional method based on a single current threshold cannot adapt to dynamic changes. Therefore, a fuzzy reasoning method based on multi-feature information fusion is designed in the embodiment, which makes comprehensive judgments based on multiple state information to improve the accuracy and robustness of critical operating condition identification.

[0036] The inputs of the identifier are multiple characteristic quantities that can reflect the approach of the system to the critical state: (1) Normalized current distance (I) , whose calculation formula is:

[0037] where, is the real-time acquired stack load current; is the effective working current range of the ejector calibrated by previous experiments; is the center value of the range, which quantitatively describes the current load relative to the center of the critical interval; when the absolute value is small, it means that the system is near the center of the critical interval.

[0038] (2) the current rate of change (dI / dt), whose formula is:

[0039] It represents the intensity and direction of load change. A rapidly increasing load, such as a large positive value, will make the system enter or cross the critical interval more quickly, so the control strategy needs to be more proactive.

[0040] (3) the pressure fluctuation intensity (σp), which is obtained by calculating the standard deviation of the anode inlet pressure signal in a recent window, whose formula is:

[0041] where, is the average value of the anode inlet pressure in the window period.

[0042] Near the critical condition, due to the instability of airflow switching, the pressure often shows specific oscillation characteristics, which can effectively capture this feature.

[0043] Next is the fuzzification process, which converts each precise input into a fuzzy semantic description, for example, dividing into five fuzzy sets: "negative big (NB)", "negative small (NS)", "zero (ZE)", "positive small (PS)", and "positive big (PB)", and defining the corresponding membership functions (such as triangular or trapezoidal functions) for each set, so as to determine the "degree" of the current precise value belonging to each fuzzy set.

[0044] Here, the membership function has the independent variable as the precise value of the normalized current distance , and the dependent variable as the membership degree of the value belonging to a certain fuzzy set (such as NB, NS, ZE, etc.), whose value is between 0 and 1.

[0045] ​​The core of fuzzy inference relies on a pre-defined rule base, which is based on deep understanding of the system physics and extensive experimental data. The rules are usually in the form of "IF condition1 AND condition2... THEN conclusion"; for example, a typical rule could be: ZE (current is close to the center of the critical region) P (load is increasing) M (moderate fluctuations have occurred) THEN critical condition confidence High (high likelihood of being in critical state)".

[0046] The system activates all relevant rules and eventually combines the outputs of all rules into a precise critical condition confidence value between 0 and 1 through a defuzzification method (such as the center of gravity method) . The closer to 1, the more likely the system is in a critical switching state; this continuously changing confidence value provides more information than a simple Boolean judgment (yes / no) for subsequent smooth control.

[0047] The core idea of the center of gravity method is to calculate the "center of mass" or "center of gravity" of the combined output fuzzy set of all activated rules, and use the horizontal coordinate value of this center of gravity as the final precise output. The standard calculation formula is as follows:

[0048] where, is the calculated precise critical condition confidence, i.e. the final result; is the number of activated fuzzy rules; is the center value of the fuzzy set corresponding to the conclusion of the th rule. For example, if the conclusion is "high confidence", the "high" set on the universe of discourse may be defined as a triangular membership function, and the horizontal coordinate value of its vertex (such as 0.8) is the center value of the set .

[0049] is the activation strength of the th rule, which is obtained by "and" operation (usually taking the minimum value) of the membership degrees of all premise conditions of the rule, representing the credibility of the conclusion of the rule under the current input.

[0050] 2.2 Parameter-adjustable anode pressure dynamic prediction model The performance of MPC is highly dependent on the accuracy of its internal prediction model. In order to accurately describe the dynamic characteristics of the anode pressure under critical conditions of the ejector, a high-precision prediction model combining mechanism and data-driven methods with online adjustable parameters is constructed. The core of the model is based on the law of conservation of mass and the ideal gas equation of state, which is: 2.2.1 Theoretical basis and core equation derivation of the model The anode flow channel is regarded as a control body. According to the law of conservation of mass, the rate of change of hydrogen mass in the control body is equal to the inflow mass flow minus the outflow mass flow and the mass flow consumed by the reaction , which can be expressed as:

[0051] For the anode recirculation loop, the outflow mass flow is 0 without purging, and the ideal gas equation of state is introduced , where R is the ideal gas constant, (M is the molar mass of hydrogen), and the mass can be related to the pressure ; the anode volume is constant, and the derivative of both sides of the equation of state is taken and substituted into the mass conservation equation. After rearrangement, the core differential equation describing the dynamic change of the anode inlet pressure is obtained:

[0052] where M is the molar mass of hydrogen, and V is the anode volume.

[0053] In order to combine the flow more conveniently, the mass flow is usually converted to the molar flow , which is substituted into the above equation. The core equation of the anode pressure dynamic prediction model can be simplified as:

[0054] where is the absolute pressure of the anode inlet, with units of pascal ( ), which is the core state variable of the model and also the target that needs to be stabilized by the control system.

[0055] is the absolute temperature of the stack, with units of kelvin ( ), and temperature directly affects the density and volume of the gas, which is a key parameter for accurate calculation of the number of moles of gas.

[0056] Volume of anode flow channel, unit: cubic meter (m3) This parameter is a fixed parameter determined by the structure of fuel cell.

[0057] Molar flow rate of hydrogen consumed by electrochemical reaction, unit: mole per second (mol / s) This parameter is directly determined by the current operating state of the stack, and the calculation formula is:

[0058] Wherein, N is the number of single cells in the stack, I is the real-time collected load current of the stack (unit: ampere, A) , and F is the Faraday constant (96,485 C / mol) ; this formula reflects the stoichiometric relationship between current and hydrogen consumption in electrochemical reaction.

[0059] Total molar flow rate of hydrogen entering the anode flow channel from the outside (mol / s) This parameter is the flow rate contributed by the ejector and the flow rate contributed by the circulating pump :

[0060] 2.2.2 Ejector flow model and its adaptive correction The working characteristics of the ejector are complex, and its flow rate is affected by the pressure difference before and after the nozzle, the valve opening, and the flow state. In this embodiment, an improved equivalent nozzle model is used to describe its flow characteristics:

[0061] Wherein, is the working condition adaptive correction coefficient, which is a function of the critical working condition confidence calculated above, i.e. ; when the critical working condition with high confidence is identified (for example ), the internal flow of the ejector may become unstable, and there is a risk of flow separation or slight surge. At this time, through experimental data calibration or expert knowledge, the is set to a value slightly less than 1, such as between 0.92 and 0.98, to compensate for the prediction deviation of the model under this special working condition, so that the prediction result is closer to the actual situation.

[0062] is the flow coefficient, which is a function of the opening of the proportional valve of the ejector , and this function relationship is obtained through bench test data calibration, reflecting the flow efficiency of the valve at different openings.

[0063] Arefis the reference cross-sectional area of the ejector nozzle (m2), which is a fixed geometric parameter.

[0064] ρH2is the density of hydrogen (kg / m3), which can be calculated from the current pressure and temperature of hydrogen according to the ideal gas state equation.

[0065] Psupis the supply pressure of high-pressure hydrogen (Pa).

[0066] Choked flow function, which is used to describe the choking phenomenon of compressible fluid (hydrogen) in converging flow passage. Its characteristics are: when the back pressure ratio is lower than the critical pressure ratio (about 0.53 for hydrogen), the gas flow velocity at the outlet of the flow passage reaches the local sound speed, and the flow rate is no longer affected by the downstream back pressure, at this time , the flow rate reaches the maximum value; when the back pressure ratio is higher than the critical value, the flow is in a subcritical state, and the flow rate decreases with the increase of back pressure, at this time .

[0067] 2.2.3 Circulating pump flow rate model The circulating pump can be modeled as a linear or nearly linear function of the pump rotational speed:

[0068] where, ηvolis the volumetric efficiency of the circulating pump, taking into account factors such as internal leakage of the pump.

[0069] Ncurrentis the current rotational speed of the pump (unit: revolutions per minute), which is determined by the duty cycle of the PWM signal output by the controller.

[0070] Nmaxis the maximum safe rotational speed allowed for the pump.

[0071] Qmaxis the maximum volumetric flow rate that the pump can provide at the maximum rotational speed (m3 / s); in use, it needs to be converted to molar flow rate ( mol / s) according to the current temperature and pressure state of the gas.

[0072] 2.2.4 Discretization and online application of the model ​​​​​​To implement the predictive function in a digital controller, the aforementioned continuous differential equation model needs to be discretized. Using the first-order Euler method or the more precise fourth-order Runge-Kutta method, the core equations of the anode pressure dynamic prediction model are transformed into a discrete state-space form suitable for computer computation.

[0073] in, For the next discrete time point Predicted anode inlet pressure; At the current discrete time point The actual measured anode inlet pressure; For the current discrete time point The opening degree of the ejector proportional valve; At the current discrete time point The duty cycle of the PWM signal output by the controller; At the current discrete time point The measured load current output by the fuel cell stack; At the current discrete time point The measured temperature of the fuel cell stack; For the current discrete time point of calculation The critical operating condition confidence level has a value between 0 and 1. The closer it is to 1, the greater the likelihood that the system is in a critical switching state.

[0074] This discrete model will be embedded into the predictive module of the controller. In each control cycle, the controller will predict the current time. Measured values Control sequence of future hypotheses and the confidence level of the critical operating condition calculated in real time. This model is called to recursively predict the anode inlet pressure trajectory at multiple future time points. This provides a basis for subsequent rolling optimization.

[0075] 2.3 Adaptive Model Predictive Controller 2.3.1 Controller Parameter Set The core innovation of the embodiment is to make the traditional model predictive control have the ability of online self-tuning key parameters, so as to actively adapt to the change of system dynamic characteristics, especially in complex critical switching conditions, a double-layer parameter adaptive mechanism is designed, which can not only feed forward according to the working condition, but also feedback according to the actual stack operating state of the fuel cell.

[0076] The adjustable key parameter set of the controller is defined as , and the definition of each parameter is as follows: : prediction horizon, which determines how many time steps the controller algorithm predicts the future behavior of the system in advance, a longer usually means better stability and robustness, but larger calculation amount.

[0077] : control horizon, which defines the number of future control steps that the optimizer can change.

[0078] : weight matrix of state (output) tracking error, which determines the importance of the controller to the pressure error ;The larger the value, the more the controller is committed to quickly eliminate the pressure error.

[0079] and : weight matrix of control amount change, corresponding to the change of injector valve opening and circulating pump speed respectively; larger or value will punish the violent change of control action, make the control more smooth, but may sacrifice the response speed.

[0080] : economic weight coefficient, used to punish the power consumption of the circulating pump in the objective function , guiding the system to preferentially use the ejector with lower energy consumption under the premise of meeting the performance.

[0081] 2.3.2 Implementation process of double-layer parameter adaptive control mechanism The implementation process of the double-layer parameter adaptive control mechanism of the controller is as follows: First layer: feedforward parameter preset based on working condition identification This layer plays the role of decision-making of the control system, which generates a set of basic parameters for the MPC controller through the preset, simulation and experiment optimized mapping function according to the critical working condition confidence calculated by the working condition identifier in real time:

[0082] The strategy of this mapping function is as follows: when ， At this point, the circulating pump dominates the mode: the system is relatively stable and far from the critical region; the strategy here prioritizes economy and smoothness, therefore... The function will output a set of preset parameters: a longer prediction time domain. To enhance robustness, medium-sized tracking weights Larger control variable weights and To ensure smooth movements and a greater emphasis on economic efficiency. To reduce pump consumption.

[0083] when At this point, the system enters a critical switching mode: it enters a region of critical dynamic complexity; the strategy shifts towards dynamic performance and rapid response. At this time, the mapping function is adjusted to another set of preset parameters: a moderate prediction time domain. To balance predictive power and computational speed, the tracking weights are increased. To suppress pressure fluctuations, a moderate control weight is used. and To prevent overshoot and reduce economic weights For the time being, priority will be given to ensuring control performance rather than energy efficiency.

[0084] when At this point, the system enters ejector-dominated mode: the switchover is largely complete, but stability must be ensured; the strategy focuses on a stable transition and energy efficiency optimization, with parameters set to a medium-to-long duration. ,medium smaller (The ejector is now working stably), larger (The pump will shut down smoothly), and the largest To maximize the use of the ejector.

[0085] Second layer: Closed-loop parameter fine-tuning based on instantaneous performance feedback This layer is used to fine-tune the control parameters output by the first layer, using the "coarse-tuning" parameters provided in the first layer. Based on this, a closed-loop optimizer is introduced to make fine adjustments according to the instantaneous performance of the control system in order to compensate for the effects of model errors and random disturbances.

[0086] First, define an instantaneous performance metric. This indicator is used to quantify the control effect at the current moment, and its calculation method is as follows:

[0087] in, This is the current pressure tracking error. These are all changes in the current control variables, namely changes in the ejector valve opening and changes in the circulating pump speed. and It is a weight used to balance tracking performance and the intensity of control actions.

[0088] Then, online optimization is performed using the gradient descent method; for the parameter set... One of the parameters Its adjustment amount The calculation is based on the gradient estimate of the parameter by the performance index:

[0089] In practical applications, gradient Since direct analytical solutions are difficult to obtain, numerical approximation methods are used for calculation.

[0090] in, It is for parameters The learning step size, and at the same time, to prevent over-adjustment or oscillation of parameters, the adjustment amount needs to be adjusted. Limitations can be imposed, and a dead zone can be set so that parameters are only updated when performance changes exceed a certain threshold.

[0091] Finally, the parameter set after two layers of adaptive adjustment is:

[0092] Where ΔΘ(k) represents all the parameters that need to be adjusted at a specific time k. Adjustment amount A set of.

[0093] 2.3.3 Online rolling optimization process with integrated adaptive control algorithm: Integrating the above components, each control cycle (sampling time) The complete algorithm flow within is as follows: (1) Initialization: The system starts up and loads the default controller parameters. Initialize the state estimator.

[0094] (2) Main loop (at each sampling time) ): 1) Data Acquisition and Preprocessing: Reading data from sensors Wait for the latest fuel cell stack operating status parameter data.

[0095] 2) Working condition recognition: call working condition recognizer, calculate current critical working condition confidence .

[0096] 3) Parameter adaptive adjustment: First layer (feedforward preset): calculate reference parameter set according to , through mapping function .

[0097] Second layer (feedback fine-tuning): based on the change of recent performance index , calculate parameter adjustment amount , get the final adaptive parameter set .

[0098] 4) State estimation and prediction: based on the measurement value at the current time , the future assumed control sequence and the real-time calculated critical working condition confidence , use the parameter adjustable prediction model to calculate the anode pressure trajectory in the future step .

[0099] 5) Optimization problem construction and solution: use the adaptive parameter set and the anode pressure trajectory to construct the finite time domain optimization problem in the following form, the objective function is composed of the pressure tracking term, the control smoothing term and the economy term, the pressure tracking term is used to punish the deviation of the predicted pressure from the reference pressure, the control smoothing term is used to punish the sharp change of the ejector proportional valve opening and the circulating pump speed, and the economy term is used to punish the operating power consumption of the circulating pump, which is expressed by the formula as follows:

[0100] At the same time, a series of constraint conditions must be met, including: State constraint: , i.e. pressure safety boundary, to ensure that the pressure is within the safe range.

[0101] Input constraint: , i.e. stroke limit, physical limit of actuator.

[0102] Input rate constraint: , i.e. change rate limit, limit the action speed of actuator to ensure smoothness.

[0103] Then, call embedded optimization solver (such as efficient set method, interior point method) to solve this constrained quadratic programming (QP) problem, to get the optimal control sequence in the future step .

[0104] ​(3) Control implementation: adopt the rolling optimization strategy in predictive control, only the first control amount in the optimal control sequence is sent to the actuator Actual output to the ejector proportional valve and the circulating pump actuator.

[0105] (4) State update and cycle: update the time index, Wait for the next sampling period to come, and then return to step a to start a new round of perception, decision-making, and execution cycle.

[0106] 3. The execution layer is responsible for converting the control instructions of the decision-making layer into control amounts of the ejector proportional valve and the anode circulating pump, the operation control of the two key actuators.

[0107] These two components receive instructions from the controller, adjust the hydrogen flow into the anode and the recirculation gas amount by changing the valve opening and pump speed.

[0108] 4. Overall control process In each fixed sampling period (10 milliseconds): The latest data of the perception layer is first sent to the working condition identifier, which quickly determines whether the current working mode of the system is close to the critical point of the ejector effect.

[0109] The identification result is passed to the prediction model, which predicts the trend of anode pressure changes in the future period based on the current stack operating state and different assumed control actions.

[0110] Finally, the adaptive model predictive controller integrates the current stack operating state and the predicted trend of anode pressure changes in the future period, uses an optimization algorithm to calculate the optimal control instructions that can best balance pressure stability, action smoothness, and system energy efficiency, and sends them to the execution layer's ejector valve and circulating pump. After the actuator acts, the system state changes, and a new round of perception, decision-making, and execution cycle starts, forming a continuous optimization intelligent control closed loop.

[0111] 5. Real-time guarantee and hardware implementation To ensure that the above complex algorithm can meet the strict real-time requirements in the vehicle controller, the invention has carried out targeted optimization in software and hardware levels.

[0112] 1. Calculation optimization: The fuzzy reasoning process in the working condition identification is converted into a look-up table (Look-up Table) through offline calculation, which directly looks up the table to obtain , greatly reducing the calculation time The complex nonlinear function (such as critical flow function ) is implemented by piecewise linear approximation or pre-computed lookup table.

[0113] High-efficiency QP solving algorithm optimized for embedded systems is selected, and hardware acceleration function of the processor is fully utilized.

[0114] 2. Task scheduling: The running period of the adaptive model predictive controller, especially the second layer fine-tuning algorithm with large calculation amount, can be set as an integer multiple of the main loop period, for example, once every 5 or 10 MPC periods. In this way, it is ensured that the parameters can follow the system changes, and a large amount of calculation is avoided every period, balancing the calculation load and adaptive speed.

[0115] 3. Hardware platform: The core processor is selected from high-performance digital signal processors (DSP) or high-performance microcontrollers, such as TI's C2000 series DSP or MCU with ARM Cortex-M7 core. These processors have high main frequency, hardware floating point operation unit and sufficient on-chip memory (RAM and Flash), which provides hardware guarantee for the running of complex algorithms. Reliable sensor signal conditioning circuit and actuator driving circuit are designed to ensure the accuracy of data acquisition and the execution of control instructions.

[0116] Embodiment 2 In an embodiment of the present application, a fuel cell anode pressure fluctuation suppression method is provided, which suppresses anode pressure fluctuation of a fuel cell anode hydrogen supply and recirculation system, Figure 1 as shown in the fuel cell anode hydrogen supply and recirculation system, Figure 2 The hierarchical technical architecture and information flow of the core control algorithm of this embodiment are depicted, which is divided into three core levels from left to right, forming a complete intelligent control closed loop.

[0117] The leftmost side is the perception and working condition identification layer, which is responsible for real-time acquisition and preprocessing of sensor data, and calling a multi-feature fuzzy fusion critical working condition identifier, analyzing multiple feature quantities such as current and pressure change trend, judging whether the system is currently in a circulation pump dominant, critical switching or ejector dominant mode, and outputting a quantitative confidence.

[0118] The middle layer is the prediction and adaptive layer, mainly including an anode pressure dynamic prediction model, which can predict the change trend of anode pressure according to the current state and working condition confidence, and provide reference for the execution of the control algorithm; The rightmost side is the decision and execution layer, i.e. the adaptive model predictive controller, which adjusts the key parameter set of the model predictive controller, such as prediction time domain and control weight, online according to the working condition confidence and real-time control performance through a double-layer parameter adaptive mechanism. In addition, rolling optimization calculation can be performed to generate optimal control instructions and deliver them to the actuator.

[0119] In this embodiment, the core controller adopts a 32-bit multi-core microcontroller with built-in hardware floating point operation unit and lockstep core; the controller communicates with the vehicle upper layer VCU through CAN bus to obtain driving instructions; the sensor system includes current sensor , pressure sensor , PT100 platinum resistance temperature sensor . The actuator is a proportional type ejector control valve and a three-phase brushless DC circulating pump.

[0120] After the system is powered on, the controller is initialized and the preset parameters are loaded. The reference parameters of model predictive control are set as follows: sampling period , prediction horizon , corresponding time is 0.3 seconds, control horizon . The initial value of the weight matrix is: pressure tracking weight , ejector valve opening change weight , circulating pump speed change weight , economy weight .

[0121] I. Working condition identifier Figure 3 is the working principle of the critical working condition identifier, which includes four key steps: First, multi-feature extraction, the identifier calculates key feature quantities from sensor data, including normalized current distance to reflect the relative position of the load, current rate of change to represent the load change trend, and pressure fluctuation intensity to capture specific oscillation patterns.

[0122] Then, the fuzzification process converts each precise feature quantity value into the membership degree of different fuzzy language variables.

[0123] Then, fuzzy reasoning, which is the core step, the system performs logical reasoning according to the preset fuzzy rule base, and the rule form is usually if a certain feature quantity meets a certain condition, then the critical confidence belongs to a certain state.

[0124] Finally, the defuzzification step combines and converts the results of fuzzy reasoning into an accurate critical working condition confidence value.

[0125] Compared with single threshold judgment, this method can identify the critical state of the system earlier and more accurately.

[0126] In this embodiment, the critical working condition identifier is implemented as a separate task module in software, and the implementation steps are as follows: 1. Feature quantity calculation: read , calculate the normalized current distance At the same time, the rate of change of current and the standard deviation of anode pressure signal in the past 50 sampling points are calculated as the pressure fluctuation intensity .

[0127] 2. Fuzzification: Triangular membership functions are defined for respectively. For example, the domain of is [-1, 1], and the center points of its fuzzy sets are set to -1, -0.5, 0, 0.5, 1 respectively.

[0128] 3. Rule base and reasoning: A fuzzy rule base containing 15 rules is established based on expert experience. For example, the rule: is adopted, and the Mamdani min-max reasoning method and the center of gravity method are used to solve fuzzification, outputting the confidence .

[0129] II. Parameter adaptive model predictive control Figure 4 The logic structure of the double-layer parameter adaptive mechanism is shown, which aims to dynamically adjust the key parameter set of the model predictive controller to adapt to different working conditions.

[0130] The first layer is a feedforward parameter preset layer based on working condition recognition. According to the real-time calculated critical working condition confidence, a set of basic parameters is quickly queried or calculated through a preset mapping function. For example, when the confidence is high, the preset parameters will tend to a shorter prediction horizon and a higher tracking weight, so that the controller responds more quickly.

[0131] The second layer is a closed-loop parameter fine-tuning layer based on performance feedback. On the basis of the basic parameters provided by the first layer, according to the changes of the instantaneous control performance index, the parameters are finely adjusted in a small range by using optimization algorithms such as gradient descent, to compensate for model errors and disturbances.

[0132] Finally, the two layers of adjustment are superimposed to form a complete adaptive parameter set for the current control period, so that the controller always maintains the best performance.

[0133] Figure 5 The complete online rolling optimization process of the integrated adaptive model predictive control is shown in the form of a flowchart: The flow starts with system initialization and parameter loading, and then enters the main loop. In each sampling period, first, data acquisition and preprocessing are performed; then, based on the acquired data, the critical condition confidence is calculated; then, double-layer parameter adaptation is performed to obtain the adjusted controller parameters; the next step is to use the adjusted prediction model to perform state prediction with the current state as the initial condition; then, a constrained optimization problem is constructed and solved to obtain the optimal control sequence for a period of time in the future; then, the immediate control quantity in the optimal sequence is applied to the actuator; finally, the next sampling period is waited for, and the data acquisition step is returned to start a new round of loop. This flow is repeated to realize continuous closed-loop optimization control of the anode pressure.

[0134] The online execution flow of the adaptive model predictive control algorithm is as follows: 1. Data acquisition and condition identification: at each sampling time k, the data are acquired, and the identifier is called to obtain .

[0135] 2. Double-layer parameter adaptation: The first layer (feedforward preset): according to the value of , the basic parameters are determined by table lookup method.

[0136] The second layer (feedback fine-tuning): the integral absolute error of the pressure tracking error in the past 1 second is calculated as the performance index . If increases by more than 10% compared with the last period, it is fine-tuned according to , otherwise it is reduced. The fine-tuning amount is limited within ±20%.

[0137] 3. Rolling optimization and control implementation: Using the current state estimate value as the initial condition, the adjusted parameter set and the prediction model are used to construct an optimization problem. An embedded QP solver is called to solve the problem to obtain the optimal control sequence . The first control quantity in the sequence is converted into a PWM signal to control the electromagnetic valves of the circulating pump and the ejector valve.

[0138] Example 3 In an embodiment of the present application, a fuel cell anode pressure fluctuation suppression system is provided, comprising: A state perception module configured to perceive the stack operation of the target fuel cell in real time to obtain stack operation state parameters; An instruction decision module is configured to make critical condition identification, pressure trend prediction and adaptive model prediction control based on the stack operating state parameters, and to decide optimal control instructions for balance pressure stability, action smoothness and system energy efficiency. An instruction execution module is configured to convert the optimal control instructions into the opening degree of the proportional valve at the front end of the ejector and the rotating speed of the anode circulating pump, and to finally adjust the hydrogen flow entering the anode and the recirculation gas amount by changing the opening degree and the rotating speed, so as to suppress the anode pressure fluctuation of the target fuel cell.

[0139] Embodiment 4 In an embodiment of the present application, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the fuel cell anode pressure fluctuation suppression method.

[0140] Embodiment 5 In an embodiment of the present application, a non-transitory computer readable storage medium is provided, which is used to store computer instructions, and the computer instructions, when executed by a processor, implement the fuel cell anode pressure fluctuation suppression method.

[0141] Embodiment 6 In an embodiment of the present application, an electronic device is provided, comprising a processor, a memory and a computer program; wherein the processor is connected with the memory, and the computer program is stored in the memory; when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device implements the fuel cell anode pressure fluctuation suppression method.

[0142] The present application is described in reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the computer or other programmable data processing device produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks

[0143] These computer program instructions can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are performed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide a device for implementing the functions specified in the flowcharts and / or block diagrams.Figure 1 one or more processes and / or functions specified in one or more blocks Figure 1 one or more processes and / or functions specified in one or more blocks

[0144] The above description is only the specific implementation of the present application, and is not intended to limit the protection scope of the present application. It should be understood by those skilled in the art that various modifications or changes can be made on the basis of the technical solutions of the present application without creative labor, and still fall within the protection scope of the present application.

Claims

1. A method for suppressing anode pressure fluctuations in a fuel cell, characterized in that, include: The stack operation status of the target fuel cell is monitored in real time to obtain stack operation status parameters; Based on the stack operating state parameters, critical operating conditions are identified, pressure trends are predicted, and adaptive model predictive control is performed to determine the optimal control command that balances pressure stability, smoothness of action, and system energy efficiency. The optimal control command is translated into the opening degree of the proportional valve at the front end of the ejector and the rotational speed of the anode circulation pump. By changing the opening degree and rotational speed, the flow rate of hydrogen entering the anode and the amount of recirculated gas are ultimately adjusted to suppress the anode pressure fluctuation of the target fuel cell.

2. The method for suppressing anode pressure fluctuations in a fuel cell as described in claim 1, characterized in that, The stack operating parameters include at least the stack load current, anode inlet pressure, anode outlet pressure, and stack temperature.

3. The method for suppressing anode pressure fluctuations in a fuel cell as described in claim 1, characterized in that, The critical operating condition identification is based on the real-time sensed stack operating status parameters, and calculates the current critical operating condition confidence of the target fuel cell through a multi-feature fuzzy fusion algorithm. The input variables of the multi-feature fuzzy fusion algorithm include normalized current distance, current change rate, and pressure fluctuation intensity. The normalized current distance is calculated based on the current stack load current and the preset effective operating current range of the ejector. The pressure fluctuation intensity is calculated by performing standard deviation calculation on recent anode pressure signals.

4. The method for suppressing anode pressure fluctuations in a fuel cell as described in claim 1, characterized in that, The pressure trend prediction is achieved by using the stack operating status parameters and critical condition confidence levels, and through a parameter-adjustable dynamic prediction model of anode pressure, to predict the trajectory of anode pressure changes in the future time domain. The adjustable-parameter dynamic prediction model for anode pressure is based on the law of conservation of mass, and its differential equation is expressed as: in, It is the absolute pressure at the anode inlet. It is the ideal gas constant. It is the absolute temperature of the fuel cell stack. It is the volume of the anode flow channel. It is the molar flow rate of hydrogen gas consumed in the electrochemical reaction. This is the total molar flow rate of hydrogen entering the anode channel from the outside; this parameter represents the flow rate contributed by the ejector. and the flow rate contributed by the circulating pump sum.

5. The method for suppressing anode pressure fluctuations in a fuel cell as described in claim 3, characterized in that, The adaptive model predictive control includes: A two-layer parameter adaptive mechanism is adopted to adjust the key parameter set of the model predictive control algorithm online based on the critical operating condition confidence and instantaneous control performance index. The key parameter set includes at least the prediction time domain, control time domain, state tracking weight matrix, and control quantity change weight matrix. Using the adjusted set of key parameters and the trajectory of anode pressure changes in the future time domain, a constrained finite-time optimization problem is constructed and solved to generate optimal control commands for the coordinated control of the ejector proportional valve and the anode circulation pump.

6. The method for suppressing anode pressure fluctuations in a fuel cell as described in claim 5, characterized in that, The two-layer parameter adaptive mechanism includes: First-layer feedforward preset: Based on the critical operating condition confidence level, a set of basic parameters are obtained by querying or calculating through a preset mapping function; The second layer of feedback fine-tuning: using the basic parameters as initial values, and based on the changes in instantaneous performance indicators, the key parameter set is fine-tuned online using the gradient descent method.

7. The method for suppressing anode pressure fluctuations in a fuel cell as described in claim 5, characterized in that, The objective function of the finite-time optimization problem consists of three terms: The pressure tracking term is used to penalize deviations between predicted pressure and reference pressure. A smoothing control term is used to penalize drastic changes in ejector proportional valve opening and circulating pump speed; The economics item is used to penalize the operating power consumption of the circulating pump.

8. A fuel cell anode pressure fluctuation suppression system, characterized in that, include: The status sensing module is configured to: sense the stack operation status of the target fuel cell in real time and obtain the stack operation status parameters; The instruction decision module is configured to: identify critical operating conditions, predict pressure trends, and perform adaptive model predictive control based on the stack operating state parameters, and determine the optimal control instruction that balances pressure stability, smoothness of action, and system energy efficiency. The instruction execution module is configured to convert the optimal control instruction into the opening degree of the proportional valve at the front end of the ejector and the rotational speed of the anode circulation pump. By changing the opening degree and rotational speed, the flow rate of hydrogen entering the anode and the amount of recirculated gas are ultimately adjusted to suppress the anode pressure fluctuation of the target fuel cell.

9. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement a fuel cell anode pressure fluctuation suppression method as described in any one of claims 1-7.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform a method for suppressing anode pressure fluctuations in a fuel cell as described in any one of claims 1-7.

Images