Method for controlling choking state of ejector in hydrogen supply system of hydrogen fuel cell

Abstract

The invention discloses a method for controlling the choking state of an ejector in a hydrogen supply system of a hydrogen fuel cell, which comprises the following steps of: constructing an SIMULINK model containing a hydrogen tank, a pressure reducing valve, a proportional valve, the ejector, a hydrogen supply pipeline, an anode stack and a circulating pump, exciting the proportional valve by a PRBS signal, acquiring the output of a back pressure ratio (pion = Pb / P), acquiring a dynamic model of the ejector through system identification, and calculating the choking state of the ejector in the hydrogen supply system of the hydrogen fuel cell. A fuzzy PID algorithm is adopted to control the opening degree of the proportional valve; the method specifically comprises the steps that in a hydrogen supply system, a model is updated by replacing proportional valve parameters with PRBS signals; the pi deviation and the change rate of the pi deviation serve as input, the opening degree of a proportional valve is adjusted through fuzzification, rule reasoning and defuzzification, and the ejector is kept in a choked state; the method is suitable for choking state control of the fuel cell hydrogen supply system, through cooperation of dynamic modeling and a control algorithm, in the typical power range of the fuel cell, the system stability and the hydrogen recovery efficiency are remarkably improved through the control method, and an efficient solution is provided for variable working condition operation of the fuel cell.

Description

Technical Field

[0001] This invention relates to the field of hydrogen fuel cell technology, and in particular to a method for controlling the clogging state of the ejector in a hydrogen fuel cell hydrogen supply system. Background Technology

[0002] Hydrogen fuel cells generate electricity through the electrochemical reaction of hydrogen and oxygen. The stability of the hydrogen supply system directly affects the performance and efficiency of the fuel cell. As the core component of the hydrogen supply system, the ejector draws low-pressure circulating hydrogen from the ejector end through a high-pressure hydrogen coil at the drive end, enabling hydrogen recovery and reuse. Its operating status is crucial to the system efficiency.

[0003] This invention constructs a dynamic model including a hydrogen tank, a proportional valve, an ejector, a hydrogen supply pipeline, an anode stack, and a circulating pump. Combining the ejector characteristics obtained through dynamic identification, a fuzzy PID control algorithm is used to dynamically adjust the opening of the proportional valve, keeping Π stable in the range of 0.52-0.58, while simultaneously satisfying the standard deviation of the pressure ratio change rate σdΠ / dt < 0.005, thereby ensuring efficient operation of the system under choke conditions.

[0004] The choking state is the efficient operating range of the ejector. Under these conditions, the hydrogen flow velocity through the nozzle reaches the speed of sound, and the flow rate is unaffected by downstream back pressure, ensuring a stable hydrogen recirculation ratio and maximizing recovery efficiency. Existing technologies largely focus on the theoretical derivation and structural parameter design of the ejector. While this can broaden the critical mode operating range, it lacks supporting control methods. Even with structurally optimized ejectors, the choking state can still fluctuate significantly due to a lack of real-time adjustments. This invention achieves stable maintenance of the ejector in the choking state by performing system identification and design on the established hydrogen supply system and then employing PID fuzzy control. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the technical solution adopted by this invention to achieve the above objectives is as follows:

[0006] (1) System modeling:

[0007] Construct a SIMULINK dynamic model including a hydrogen tank, proportional valve, ejector, hydrogen supply pipeline, anode stack, and circulation pump; one ejector model design suitable for a fuel cell hydrogen circulation supply system specifically includes:

[0008] Driver-side flow equation:

[0009]

[0010] The Mach number correction factor is given by the following equation:

[0011]

[0012] Where γ is the adiabatic coefficient and R is the molar gas constant. is the molar mass of hydrogen gas;

[0013] The hydrogen tank is connected to a proportional valve, which receives the opening adjustment Δu from the fuzzy PID controller; the proportional valve is connected to an ejector, and the ejector inlet pressure sensor collects the inlet pressure P1 at the drive end, and the temperature sensor collects the inlet temperature at the drive end. The ejector mixing chamber outlet is connected to the anode stack hydrogen inlet via a hydrogen supply pipeline, and a back pressure sensor on the pipeline collects the ejector back pressure P. b The anode stack outlet is divided into an exhaust branch and a recovery branch. The recovery branch is connected to the circulation pump inlet via a hydrogen supply pipeline, and the stack outlet pressure is used as the circulation pump inlet parameter. The circulation pump outlet is connected to the ejector end inlet via a hydrogen supply pipeline.

[0014] (2) A dynamic identification method includes:

[0015] The excitation signal PRBS is generated as the proportional valve opening input, with a sequence length of 127 bits, an amplitude of ±20% of the rated opening of the proportional valve, and a signal bandwidth covering 0.1-10Hz.

[0016] When performing model identification, the prediction error method is used to estimate parameters when identifying the Hammerstein Wiener dynamic model using MATLAB System Identification.

[0017] The static nonlinearity of the model is fitted with a third-order polynomial, with the input being the proportional valve opening and the output being the equivalent opening after nonlinear correction.

[0018] u nl = a0 + a1u + a2u 2 + a3u 3

[0019] Where u is the actual opening, and a0 to a3 are the fitting coefficients, which are identified through PRBS excitation data;

[0020] The dynamic linear component of the rule base model is composed of a second-order time-delay transfer function.

[0021]

[0022] Where K is the gain coefficient, T1 and T2 are time constants, and τ is the time delay;

[0023] The noise processing of the model adopts the ARX model. The ARX model suppresses random noise in the collected back pressure ratio and other data by constructing an autoregressive relationship between the input and output.

[0024] (3) Fuzzy PID control methods include:

[0025] Input variable processing is performed, with pressure deviation e=Π ref The inputs are −Π and the rate of change ec=de / dt. Seven Gaussian membership functions are used for fuzzification. The universe of discourse of e is [-0.1, 0.1], the universe of discourse of ec is [−0.05, 0.05], and the fuzzy subset is {NB, NM, NS, ZO, PS, PM, PB}.

[0026] Rule-based reasoning and defuzzification are performed based on a rule base of 25 congestion-specific rules. The Mamdani reasoning method of "small activation, large synthesis" is adopted, and the discrete centroid method is used to substitute the discrete points in the universe of discourse ΔKp and their corresponding total membership degrees into the formula.

[0027]

[0028] in It is the precise output ΔKp after deblurring. To output discrete points in the universe of discourse, For discrete points The corresponding membership value, where n is the number of discrete points;

[0029] For parameter updates and critical maintenance, the PID controller parameters are updated using formulas.

[0030] Kp=Kp0+ΔKp, Ki=Ki0+ΔKi, Kd=Kd0+ΔKd

[0031] Where Kp0, Ki0, and Kd0 are initial values;

[0032] Output opening adjustment amount Δu:

[0033]

[0034] The integral term employs an anti-integral saturation algorithm, pausing integration when |e|>0.05 to avoid overshoot; the proportional valve opening is dynamically adjusted and deviation correction is performed, increasing the output weight of Δu when Π < 0.52 and decreasing the output weight of Δu when Π > 0.58.

[0035] Compared with the prior art, the present invention, by adopting the above technical solution, has the following technical effects.

[0036] This invention constructs a SIMULINK dynamic model, and the constructed ejector model fully considers nonlinear characteristics, enabling accurate reproduction of the flow transfer process in a real hydrogen supply system. By generating a PRBS signal as the proportional valve opening input, and combining it with the acquired back pressure ratio signal, the dynamic model is identified using the Hammerstein Wiener toolbox in MATLAB System Identification, effectively capturing the system's dynamic characteristics. Fuzzy PID control is employed, using the pressure ratio deviation and its rate of change as input. After fuzzification processing with a variable universe of discourse and inference from a choke zone rule base, the proportional valve opening adjustment is output through centroid defuzzification, achieving precise control of the ejector choke state. The overall method can stably maintain the ejector choke state under different operating conditions, improving hydrogen recovery efficiency and system stability. Attached Figure Description

[0037] Figure 1 A flowchart illustrating the design of a method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system;

[0038] Figure 2 A schematic diagram of the SIMULINK model connection for a hydrogen fuel cell hydrogen supply system;

[0039] Figure 3 This is a block diagram of a fuzzy PID controller.

[0040] Figure 4 This is a schematic diagram of the control model. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be described in more detail below with reference to the accompanying drawings.

[0042] like Figure 1 As shown, this embodiment provides a control method for stably maintaining the ejector blockage state, which mainly includes:

[0043] A SIMULINK dynamic model was constructed, including a hydrogen tank, proportional valve, ejector, hydrogen supply pipeline, anode stack, and circulating pump. The simulation step size was set to 0.001s, and the simulation time was set to 1800s to ensure that the model could accurately simulate the dynamic operation of the system and capture subtle changes in the system at different times. Figure 2 As shown, the specific connection method is as follows:

[0044] It is equipped with a 35MPa hydrogen tank connected to a proportional valve, and the proportional valve receives the opening adjustment Δu output by the fuzzy PID controller.

[0045] The proportional valve is connected to the ejector, and the ejector inlet is equipped with a pressure sensor to collect the inlet pressure P1 of the drive end and a temperature sensor to collect the inlet temperature of the drive end. ;

[0046] The outlet of the ejector's mixing chamber is connected to the hydrogen inlet of the anode stack via a hydrogen supply pipeline, and a back pressure sensor on the pipeline collects the ejector back pressure P. b ;

[0047] The anode stack outlet is divided into an exhaust branch and a recovery branch. The recovery branch is connected to the inlet of the circulation pump via a hydrogen supply pipeline, and the outlet pressure of the stack is used as the inlet parameter of the circulation pump.

[0048] The outlet of the circulating pump is connected to the inlet of the ejector via a hydrogen supply pipeline.

[0049] The flow equation for the key component, the ejector drive end, is described as follows:

[0050]

[0051] in, The nozzle throat area is set to 0.0005 m². The flow coefficient is set to 0.8 based on the actual situation. The inlet pressure of the ejector end, For the inlet pressure of the drive end, The inlet temperature of the drive end;

[0052] The Mach number correction factor is given by the following equation:

[0053]

[0054] Where γ is the adiabatic coefficient and R is the molar gas constant. is the molar mass of hydrogen gas;

[0055] Excitation signal application during dynamic identification:

[0056] The PRBS signal generated by the m-sequence generator is used as the input signal for proportional valve opening control;

[0057] The sequence length is set to 127 bits, and its amplitude is limited to ±20% of the rated opening of the proportional valve. The signal frequency scanning range covers 0.1 - 10Hz. During the operation, the signal generator sends a PRBS signal to the proportional valve according to the set parameters for a duration of 300s, during which relevant response data of the ejector are collected.

[0058] Model identification operation in dynamic identification:

[0059] During signal application, the ejector back pressure P is simultaneously acquired.b The ratio Π to the primary flow nozzle inlet pressure P1 is Π=P b / P1 is used as an output variable;

[0060] Simultaneously monitor the proportional valve opening u and the primary flow (hydrogen from the hydrogen tank through the proportional valve to the ejector drive end) mass flow rate q. m1 Secondary flow (anode reflux hydrogen delivered by the circulating pump) mass flow rate q m2 Key process parameters;

[0061] Using the MATLAB System Identification toolbox, a Hammerstein Wiener dynamic model was constructed using system identification methods;

[0062] The static nonlinearity element employs a third-order polynomial fitting algorithm. Parameter optimization is performed on actual operating data using the least squares method to determine the coefficients a0 = -0.5, a1 = 1.2, a2 ​​= -0.03, and a3 = 0.001. This establishes the relationship between the proportional valve opening u and the equivalent opening u after nonlinear correction. nl The mapping relationship between them;

[0063] The dynamic linear element employs a second-order time-delay transfer function:

[0064]

[0065] The noise reduction mechanism uses the ARX model to effectively suppress sensor noise.

[0066] The principle of fuzzy PID controller is as follows: Figure 3 As shown, the parameter settings and initialization are as follows:

[0067] Set target back pressure ratio Π ref =0.55; Initialize the fuzzy membership function, for the input variable pressure ratio deviation e=Π ref −Π and the rate of change ec=de / dt are respectively fuzzed using 7 Gaussian membership functions;

[0068] The universe of discourse for e is set to [-0.1, 0.1], and the universe of discourse for ec is set to [-0.05, 0.05]. The fuzzy subsets are both {NB, NM, NS, ZO, PS, PM, PB}, and the membership function width parameters σe=0.03 and σec=0.015.

[0069] Initialize a rule base containing 25 congestion zone-specific rules. Some core rules are shown in the table below:

[0070] Table 1

[0071] Serial Number condition ΔKp ΔKi ΔKd 1 e is NB and ec is NB. PB NB PS 2 e is NM and ec is NS PM NM ZO 3 e is NS and ec is ZO PS ZO ZO 4 e is ZO and ec is ZO ZO ZO ZO 5 e is PS and ec is ZO NS ZO ZO 6 e is PM and ec is PS NM PM ZO

[0072] The above rules are designed for the congestion zone (0.52≤Π≤0.58) to correct PID parameters.

[0073] Figure 4 This is a fuzzy PID control model diagram of the ejector critical state control method in the hydrogen fuel cell hydrogen supply system described in this invention. The model is built on the SIMULINK platform and integrates the dynamically identified Hammerstein Wiener model with the fuzzy PID control strategy, achieving precise closed-loop control of the ejector clogging state.

[0074] The left-hand module is the step input for the target back pressure ratio, representing the desired ejector choke pressure ratio of the system. Π is calculated using an adder. ref The deviation from the actual back pressure ratio Π of the ejector is e=Π ref −Π, the “Zero-Order Hold” module discretizes the deviation signal;

[0075] The fuzzy PID control module has a built-in rule library of 25 critical zone rules. It takes the deviation e and the rate of change of deviation de / dt as inputs, and defuzzifies the PID parameters through the Mamdani inference method and the centroid method to output the PID parameter correction.

[0076] The dynamic identification model module is a nonlinear model of the ejector obtained through dynamic identification. It accurately simulates the pressure and flow characteristics of the ejector under different operating conditions, providing a high-fidelity simulation environment for the verification of control strategies.

Claims

1. A method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system, characterized in that, include: System modeling was performed, and a SIMULINK model was constructed that includes a hydrogen tank, proportional valve, ejector, hydrogen supply pipeline, anode stack, and circulation pump. The ejector model takes into account its nonlinear characteristics. Dynamic identification generates a pseudo-random binary sequence (PRBS) signal as the proportional valve opening input; the ejector back pressure P is acquired. b The ratio Π to the primary flow nozzle inlet pressure P1 is Π=P b / P1 is used as the output; the dynamic model of the ejector is obtained through system identification; Fuzzy PID control, with pressure ratio deviation e=Π ref −Π and its rate of change ec=de / dt are inputs; after fuzzification by changing the universe of discourse, inference is performed through the established critical region-specific rule base; the centroid method is used to defuzzify and output the proportional valve opening adjustment amount Δu; the standard deviation σ of the pressure ratio change rate is calculated. dΠ / dt When σ dΠ / dt When the value is less than 0.005 and 0.52≤Π≤0.58, it is determined to be a congested state; the opening of the proportional valve is dynamically adjusted to maintain the congested state of the system and keep the flow delivery stable.

2. The method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system as described in claim 1, characterized in that, The connection relationships of the modules in the SIMULINK model are as follows: The hydrogen tank is connected to a proportional valve, which receives the opening adjustment Δu from the fuzzy PID controller; the proportional valve is connected to an ejector, and the ejector inlet pressure sensor collects the inlet pressure P1 at the drive end, and the temperature sensor collects the inlet temperature at the drive end. The ejector mixing chamber outlet is connected to the anode stack hydrogen inlet via a hydrogen supply pipeline, and a back pressure sensor on the pipeline collects the ejector back pressure P. b The anode stack outlet is divided into an exhaust branch and a recovery branch. The recovery branch is connected to the circulation pump inlet via a hydrogen supply pipeline, and the stack outlet pressure is used as the circulation pump inlet parameter. The circulation pump outlet is connected to the ejector end inlet via a hydrogen supply pipeline.

3. The method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system as described in claim 1, characterized in that, The ejector model design method is as follows: Assuming that heat exchange is neglected and the gas flow is a single-phase flow, establish the flow rate and energy conservation relationships at the driving end, ejector end, and compression end. Driver-side flow equation: in, This represents the area of ​​the nozzle throat. For flow coefficient, The inlet pressure of the ejector end, For the inlet pressure of the drive end, For the inlet temperature of the drive end, Calculations are performed under both congested and non-congested conditions; Ejector energy conservation equation: in The enthalpy of hydrogen at the outlet of the mixing chamber is determined by the driving end ( ) and ejector end ( The superposition of hydrogen enthalpy ensures thermodynamic energy balance.

4. The method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system according to claim 1, characterized in that, In the dynamic identification step: The pseudo-random binary sequence (PRBS) has a sequence length of 127 bits, an amplitude of ±20% of the rated opening, a signal bandwidth covering 0.1-10Hz, and an autocorrelation function satisfying:

5. The method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system according to claim 1, characterized in that, The dynamic identification is performed using the Hammerstein-Wiener model structure through the MATLAB System Identification toolbox, which includes a static nonlinear component, a dynamic linear component, and an ARX noise processing module.

6. A method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system according to claim 1 or 4, characterized in that, The Hammerstein Wiener model obtained through dynamic identification includes: Its model structure consists of a static nonlinear element, a dynamic linear element, and noise processing. The static nonlinear element uses a third-order polynomial fitting, with the input being the proportional valve opening and the output being the equivalent opening after nonlinear correction. The formula is as follows: the nl =a0+a1u+a2u 2 +a3u 3 Where u is the actual opening, and a0 to a3 are the fitting coefficients, which are identified through PRBS excitation data; The dynamic linear element is a second-order time-delay transfer function: Where K is the gain coefficient, T1 and T2 are time constants, and τ is the time delay; The noise processing uses the ARX model to suppress noise in the acquired data.

7. The method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system according to claim 1, characterized in that, The fuzzy PID control steps include: Pressure deviation e=Π ref −Π and the rate of change ec = de / dt are used as input variables, where Π ref The target back pressure ratio is set to 0.

55. Seven Gaussian membership functions are used to fuzzify e and ec. The universe of discourse range of e is defined as [-0.1, 0.1], and the universe of discourse range of ec is defined as [−0.05, 0.05]. Their fuzzy subsets are consistent and set as {NB (negative large), NM (negative medium), NS (negative small), ZO (zero), PS (positive small), PM (positive medium), PB (positive large)}.

8. A method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system according to claim 1 or 7, characterized in that, The fuzzy PID control includes fuzzy rule reasoning and defuzzification steps: The regulation library contains 25 core control regulations for congested states, in the form of "IF (fuzzy subset of e) AND (fuzzy subset of ec) THEN (fuzzy subset of ΔKp, ΔKi, ΔKd)". Typical rules include: if e is NB and ec is NB, then ΔKp is PB, ΔKi is NB, and ΔKd is PS; if e is NS and ec is ZO, then ΔKp is PS, ΔKi is ZO, and ΔKd is ZO. Here, ΔKp, ΔKi, and ΔKd are the correction values ​​for proportional, integral, and derivative parameters, respectively, and the fuzzy subsets are consistent with e and ec. The Mamdani inference method is used to determine the activation intensity of each rule through "smaller" operations and synthesize the output fuzzy set through "larger" operations. The centroid method is used for defuzzification to calculate the precise values ​​of ΔKp, ΔKi, and ΔKd.

9. A method for controlling ejector blockage in a hydrogen fuel cell hydrogen supply system according to claim 1 or 7, characterized in that, The parameter updates and congestion state maintenance of the fuzzy PID control specifically include: The ΔKp, ΔKi, and ΔKd calculated using the center-of-gravity method are used to update the PID controller parameters using the following formula: Kp=Kp0+ΔKp, Ki=Ki0+ΔKi, Kd=Kd0+ΔKd The proportional valve opening adjustment Δu is output by the PID controller, and the formula is: Among them, the integral term adopts an anti-integral saturation algorithm; for the congestion state deviation correction strategy, when Π<0.52, the output weight of Δu is increased to accelerate the increase of the proportional valve opening and increase the primary flow rate to improve the back pressure ratio; when Π>0.58, the output weight of Δu is decreased to slow down the adjustment of the proportional valve opening and avoid the back pressure ratio from dropping too quickly.

Images