A hydrogen control system and method for a fuel cell vehicle under network connection

Abstract

The present invention proposes a hydrogen control system and method for a fuel cell vehicle under networking, the system comprising: a proportional flow valve, a fuel cell stack, a purge valve, a pressure sensor, a circulating pump and other components. By designing a decoupled sliding mode controller, the decoupling of flow and pressure is achieved while compensating for the inaccuracy of the model, and the stability of the feedback control law is proved by the Lyapunov principle. In view of the contradiction between the dynamic response performance and the jitter of the sliding mode control, a method for adaptive optimization of sliding mode control parameters based on gradient descent is further designed, and the system stability under variable load is improved by a feedforward controller. At the same time, the sliding mode tuning parameter MAP self-learning method is adopted to ensure the stability of the closed-loop system while solving the gradient optimization delay problem under transient conditions. In addition, the road information obtained through the Internet of Vehicles is used to optimize the output power of the fuel cell stack, and the algorithm is assigned to different cores of the on-board chip to improve the algorithm operation speed and reduce the chip load.

Description

Technical Field

[0001] The present invention relates to the technical field of fuel cells, and in particular to a hydrogen control system and method for a fuel cell vehicle under networking. Background Art

[0002] The stable operation of fuel cells requires the coordinated work of the hydrogen system, air supply system and water-heat management system. The general air system supplies excess and appropriate pressure air into the fuel cell stack to prevent the efficiency of the fuel cell stack from decreasing due to oxygen starvation. The water-heat management system controls the working temperature, gas temperature and humidity of the fuel cell stack.

[0003] There are two types of hydrogen systems: direct discharge and hydrogen recycling. During the working process, the fuel cell cannot use all the hydrogen entering the stack at one time, and the direct discharge directly discharges the unreacted hydrogen into the air, which is less economical. The hydrogen recycling method recycles the unreacted hydrogen again, thereby improving the utilization rate of hydrogen. At the same time, hydrogen recycling can also improve the water balance in the stack to avoid flooding in the stack and optimize the working environment of the stack. The reasonable control of the circulation device of the hydrogen recycling system directly affects the efficiency of hydrogen utilization, so it is of great significance to optimize the control of the subcomponents of the hydrogen system.

[0004] The hydrogen system needs to reduce the pressure of 30-70Mpa high-pressure hydrogen stored in the tank to the appropriate pressure, and adjust it to the appropriate flow rate through the proportional flow valve. It is worth noting that the excessive pressure difference between the anode and cathode gases will damage the fragile proton exchange membrane. The pressure difference between the anode and cathode gases needs to be controlled within a reasonable value. The parasitic power generated by the anode decompression process is smaller than that of the cathode gas boosting process, so the anode pressure is generally controlled to follow the cathode pressure. Unlike the cathode exhaust gas that is directly discharged into the air, the unreacted hydrogen in the stack needs to be circulated to the anode inlet again to participate in the reaction, and it needs to pass through the steam-water separator for dehydration. At this time, due to the internal hydrogen reaction of the stack and the obstruction of the steam-water separator, the pressure of the recycled hydrogen will be lower than the required value, so the recycled hydrogen needs to be pressurized twice by the circulating pump. However, there is a coupling relationship between the outlet flow and pressure of the proportional flow valve and the circulating pump, which makes it difficult to adjust the opening of the proportional flow valve and the speed of the circulating pump.

[0005] The controllers of the current existing technical solutions ignore the impact of future traffic road information on the entire vehicle, and under special working conditions, the output power of the battery stack is higher than the actual demand. The control of the hydrogen system often adjusts the hydrogen system pressure separately. Considering the decoupling of flow and pressure, the control is also mostly concentrated on feedforward decoupling or state space decoupling, which depends on the accuracy of the model and is sensitive to changes in system parameters. It is difficult to implement. The corresponding decoupling control cannot simultaneously solve the steady-state control of the battery stack startup or transient fluctuations in working conditions. In addition, the corresponding optimization controller ignores the computing power limitations of the on-board chip, resulting in poor real-time and stability of the control. Summary of the invention

[0006] In view of the above shortcomings, the present invention proposes a hydrogen control system and method for fuel cell vehicles under networking. By designing a decoupled sliding mode controller, the decoupling of flow and pressure is achieved while compensating for the inaccuracy of the model, and the stability of the feedback control law is proved by the Lyapunov principle. In view of the contradiction between the dynamic response performance and the jitter of the sliding mode control, a method for adaptive optimization of sliding mode control parameters based on gradient descent is further designed, and the system stability under variable load is improved by a feedforward controller. At the same time, the sliding mode tuning parameter MAP self-learning method is adopted to ensure the stability of the closed-loop system while solving the gradient optimization delay problem under transient conditions. In addition, the road information obtained through the Internet of Vehicles is used to optimize the output power of the battery stack, and the algorithm is assigned to different cores of the on-board chip to improve the algorithm operation speed and reduce the chip load.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A networked fuel cell vehicle hydrogen control system, the system comprising:

[0009] Hydrogen storage bottle, pressure reducing valve, proportional flow valve, fuel cell, steam-water separator, purge valve, pressure sensor, circulation pump, check valve, drain valve;

[0010] Wherein, a pressure reducing valve and a proportional flow valve are connected between the hydrogen storage bottle and the battery stack, the battery stack is connected to a steam-water separator, the steam-water separator is connected to a purge valve, the purge valve is connected to a pressure sensor, a circulation pump and a one-way valve, and the one-way valve is connected to a drain valve;

[0011] The pressure reducing valve is used to reduce the high-pressure hydrogen pressure of the hydrogen storage bottle to the value required by the fuel cell, the proportional flow valve is used to adjust the flow of the hydrogen system into the stack, and the circulating pump is used to pressurize and recirculate the hydrogen; the steam-water separator is used to separate the hydrogen and the entrained water at the anode outlet of the fuel cell stack and discharge them through the drain valve, and the purge valve is used to remove the nitrogen that has penetrated and accumulated in the cathode cavity.

[0012] Another object of an embodiment of the present invention is to provide a method for controlling hydrogen in a networked fuel cell vehicle. The method is implemented by the above-mentioned networked fuel cell vehicle hydrogen control system, and the method comprises the following steps:

[0013] 1. Establishment of control model:

[0014] 1) Proportional flow valve modeling, the proportional flow valve formula is as follows:

[0015] Formula 1:

[0016] Where: P pv is the pressure at the front end of the proportional flow valve, T air is the atmospheric temperature, ρ H2 is the density of hydrogen, c 1～3 is a constant, u1 is the control input of the proportional flow valve;

[0017] 2) Intake manifold modeling, the intake manifold formula is as follows:

[0018] Formula 2:

[0019] Among them, V sm is the intake manifold volume, M pv,out is the outlet flow of the proportional flow valve, M pu,out is the outlet flow of the circulating pump, M an,in is the anode inlet flow rate of the stack, R is the molar gas constant, M sm,in Intake manifold inlet flow, M sm,out Intake manifold outlet flow;

[0020] 3) Stack modeling: Since the stack ignores the internal nitrogen and hydrogen permeation phenomena, and ignores the effect of water on the anode pressure and flow, the stack differential equation is as follows:

[0021] Formula 3:

[0022] Formula 4: P an,out =P an -ΔP st ;

[0023] Formula 5:

[0024] Where: P an Anode chamber pressure, M an,in Anode inlet flow, M react Hydrogen flow rate for anode reaction, M an,out Anode outlet hydrogen flow rate, Δp st The pressure drop at the anode outlet of the stack caused by the hydrogen reaction, C w Anode flow field width, Cd Anode flow field depth, L anode flow field length, Molar mass of hydrogen, N cell Number of cells in the stack, I req The stack current required, n is the number of anode flow field channels, c 4～10 constant;

[0025] 4) Modeling of the return air manifold. The formula of the return air manifold is as follows:

[0026] Formula 6:

[0027] Where: P rm Return manifold pressure, M rm,in Return manifold inlet flow, M rm,out Return manifold outlet flow, V rm Return manifold volume, T st Stack temperature;

[0028] 5) Circulation pump modeling, the circulation pump formula is as follows:

[0029] Formula 7:

[0030] Where: c 11 is the proportional parameter of the circulating pump speed and the control input voltage, d pu is the diameter of the circulating pump blade. The outlet pressure of the circulating pump can be calculated by its isentropic efficiency and mechanical efficiency. The calculation is relatively complicated. Generally, its inlet and outlet pressure-speed map is measured by experiment. The present invention adopts the data of Busch's MINKMA0018A circulating pump. In order to facilitate the design of the controller, the outlet pressure of the circulating pump is set to k pu u2, where k pu is a variable related to the inlet pressure of the circulating pump, and u2 is the control input of the circulating pump.

[0031] As a further technical solution of the present invention, it also includes multi-core controller task allocation, using Infineon TC275TD-Step (DAP) as a fuel cell controller, and allocating the power correction, feedforward control, disturbance observation, feedback decoupling control and parameter optimization of the fuel cell hydrogen system to the three cores of the processor CPU0, CPU1, and CPU2 in parallel. The steps include:

[0032] 1) Assign the battery stack output power correction task and feedforward control algorithm under the network connection to the first core CPU0 of the multi-core processor to run, update the battery stack required power in real time according to road information and SOC, and output the feedforward controller for transient control;

[0033] 2) A feedback decoupling controller was designed based on the required flow and pressure information of the battery stack predicted and updated by CPU0, and the algorithm was assigned to CPU1 for parallel operation;

[0034] 3) Assign the stochastic gradient optimization algorithm to CPU2 for parallel operation, and optimize the sliding mode control tuning parameters under five step sizes at the same time. Take the average of the five optimized tuning parameters as the final output tuning parameters, and make a stability judgment on the tuning parameters. The stable tuning parameters are updated to overwrite the preset offline tuning parameter Map.

[0035] As a further technical solution of the present invention, the steps of allocating the grid-connected stack output power correction task and the feedforward control algorithm to the first core CPU0 of the multi-core processor for execution, updating the stack required power in real time according to road information and SOC, and outputting the feedforward controller for transient control include:

[0036] 1.1) Determine whether the fuel cell stack is in the startup phase. If so, calculate the corresponding static feedforward control law to compensate for the disturbance caused by the hysteresis response of the hydrogen system subcomponents in the startup phase;

[0037] At the moment when the fuel cell starts to start, due to the hysteresis response of each subcomponent and the slow response speed of the flow and pressure input based on the feedback control law, the flow supplied to the stack at the moment of startup cannot meet the needs of the fuel cell, resulting in a large disturbance in the system during the startup phase. Therefore, a static feedforward control law is designed to avoid the problem of slow response of the feedback control law, increase the response speed of the system, and improve the tracking performance of the control input during the startup phase, thereby optimizing the fluctuation of the system during the startup phase;

[0038] The static feedforward controller is designed as follows:

[0039] 1.1.1) Determine the system state variables;

[0040] State 1: Anode inlet pressure:

[0041] Formula 8:

[0042] State 2: Circulation pump speed:

[0043] Formula 9: x2 = c 11 u2;

[0044] 1.1.2) Define the system output: y1 proportional valve outlet mass flow, y2 anode inlet pressure:

[0045] Formula 10:

[0046] 1.1.3) Control input: k pv Proportional flow valve control input voltage, Vpu Hydrogen circulation pump motor control input voltage;

[0047] Formula 11:

[0048] 1.1.4) Define φ1, φ2:

[0049] The hydrogen supply subsystem generally supplies excess hydrogen to the anode to prevent the fuel cell efficiency from being reduced and the proton exchange membrane from being damaged due to hydrogen starvation. Different systems have different hydrogen excess ratios. The hydrogen excess ratio reference value of the present invention uses experimental data measured by a 115KW fuel cell stack to calculate the hydrogen flow rate required at the anode inlet under different working conditions. The calculation formula is as follows:

[0050] Formula 12:

[0051] Where: F is Faraday's constant, is the molar mass of hydrogen, N cell is the number of cells in the stack, I req Current required for the battery stack:

[0052] Then φ1, φ2 are defined as follows:

[0053] Formula 13:

[0054] 1.1.5) Take a derivative of φ1 and φ2:

[0055] Formula 14:

[0056] At the beginning of the operation of the fuel cell, due to the hysteresis of the entire fuel cell system, the flow rate of the recycled hydrogen after being pressurized by the circulation pump is negligible compared to the flow rate output by the proportional flow valve, so φ1 here ignores the influence of the circulation pump fluid. Simplified as follows:

[0057] Formula 15:

[0058] 1.1.6) Order Thus, the static feedforward control law is obtained:

[0059] Formula 16:

[0060] 1.2) Obtain vehicle location information in real time;

[0061] 1.3) Obtain road information in real time;

[0062] 1.4) Obtain traffic light information in real time;

[0063] 1.5) Obtain power battery SOC in real time;

[0064] 1.6) Determine the future state of the vehicle based on road information, traffic light information and power battery SOC: rapid acceleration, rapid deceleration, slow acceleration, slow deceleration.

[0065] 1.7) Four groups of correction factors are set to correct the stack demand current information sent by the FCU at the next moment;

[0066] 1.8) Update the flow and pressure information of the hydrogen system demand supply and write it into the shared register;

[0067] 1.9) Determine whether there is a step change in the power required by the fuel cell stack. If so, calculate the corresponding dynamic feedforward control law to perform disturbance compensation;

[0068] When the power demand of the fuel cell changes in a step, it will cause large hydrogen flow demand and fluctuations, which in turn will cause large hydrogen pressure fluctuations, resulting in insufficient transient fuel supply to the stack and worsening the working conditions of the stack. Static feedforward alone cannot meet the demand for fuel cell flow and pressure in transient conditions, so a dynamic feedforward control law is designed to eliminate the system disturbance caused by the step change in the power demand, thereby improving the robustness and stability of the system and reducing the burden on the feedback controller. The coupled dynamic disturbance of the circulating pump at this time is solved in the feedback control law and ignored in the dynamic feedforward control law.

[0069] The dynamic feedforward controller is designed as follows:

[0070] 1.9.1) Correct the control input to the following form:

[0071] Formula 17:

[0072] 1.9.2) Correct the required flow rate M an,ref :

[0073] Formula 18:

[0074] Where: M an,pre is the corrected fuel cell demand flow, I pre The corrected battery stack requires current.

[0075] 1.9.3) Yes Derivative:

[0076] Formula 19:

[0077]

[0078] 1.9.4) Substitute equations 16 and 17 into equation 19 and let The dynamic feedforward control law is derived, and the specific formula is as follows:

[0079] Formula 20:

[0080] 1.10) An interrupt is triggered, and CPU1 reads the updated stack required flow and pressure information and feedforward control information;

[0081] 1.11) Reading completed, clear interrupt;

[0082] 1.12) Enter the next loop.

[0083] As a further technical solution of the present invention, the steps of designing a feedback decoupling controller according to the required flow and pressure information of the battery stack predicted and updated by CPU0, and distributing the algorithm to CPU1 for parallel operation include:

[0084] 2.1) Read the corrected stack required pressure and flow. If the reading time exceeds 1ms, interrupt the reading and use the last data;

[0085] 2.2) Tracking differentiator to smooth flow and pressure;

[0086] The differential tracker is a single-input dual-output module, which is designed to smooth the input signal to avoid the coexistence of overshoot and rise time when the input signal changes suddenly. The differential tracker can also achieve filtering at the same time; therefore, two sets of differential trackers are designed to optimize M an,pre and P pu,pre , the input is the M corresponding to the required power of the battery stack an,pre and P pu,pre , output r1 is the processed signal, output r2 is the differential of the output, and the specific formula of the tracking differentiator is designed as follows:

[0087] 2.2.1) Proportional flow valve:

[0088] Formula 21:

[0089] 2.2.2) Circulation pump:

[0090] Formula 22:

[0091] Among them: τ is the sampling period, ξ1,ξ2 are the parameters that determine the tracking speed;

[0092] 2.3) Read the current output of the hydrogen system;

[0093] 2.4) The disturbance observer collects the output for disturbance observation;

[0094] The disturbance observer can not only estimate the unknown disturbance in the system in real time and compensate it in the control loop, but also reduce the mutual influence between the system channels, thereby enhancing the system's ability to suppress disturbances and improving control performance and robustness. Therefore, the present invention designs two sets of disturbance observers, regards the flow control of the proportional flow valve and the pressure control of the circulating pump as disturbances, and decouples the disturbance effect on the adaptive sliding mode controller. The disturbance observer is designed as follows:

[0095] 2.4.1) Proportional flow valve: y1(k) is the proportional flow valve outlet flow:

[0096] Formula 23:

[0097] 2.4.2) Circulation pump: y2(k) is the outlet pressure of the circulation pump:

[0098] Formula 24:

[0099] Where: β is the observer tuning parameter, δ is the length of the linear segment of the fal function, and α is a constant between 0 and 1.

[0100] 2.5) Read the tracking differentiator output r1, r2 and the disturbance observer output z1, z2;

[0101] 2.6) Update the feedback controller input parameters;

[0102] 2.7) Calculation of control law of feedback controller;

[0103] The core of sliding mode control is to design a sliding mode surface so that the system state quickly tends to and keeps sliding on the surface, so as to achieve the purpose of stable control of the system. Fuel cells are complex nonlinear systems with strong coupling. It is difficult to accurately describe the system parameters with a model. Sliding mode control has strong robustness to the uncertainty of system parameters, external disturbances and nonlinear factors, and has fast response speed and strong convergence, which can be well applied to fuel cell systems. Therefore, this paper adopts two sets of first-order sliding mode control as the feedback control law of the system, and decouples and corrects them through tracking differentiators and disturbance observers. The specific design process of sliding mode control is as follows:

[0104] 2.7.1) The error design is as follows:

[0105] Formula 25:

[0106] The disturbance from the circulating pump flow is output by observer 1 as z 1,1 Compensation, the pressure disturbance from the proportional flow valve is output by observer 2 1,2 compensate;

[0107] 2.7.2) Design the sliding surface and take n s =1;

[0108] Formula 26:

[0109] 2.7.3) Approach rate: Constant approach rate, and replace the sign function with a saturation function to alleviate the inherent chattering problem of sliding mode control, where ε is the tuning parameter of the sliding mode control law, Parameters for adjusting the thickness of the boundary layer;

[0110] Formula 27:

[0111] 2.7.4) Derivative s and make corrections based on the disturbance observer output:

[0112] Formula 28:

[0113] Combining equations 27 and 28, we can obtain the decoupled sliding mode feedback control law:

[0114] Formula 29:

[0115] 2.8) Read the disturbance observer output z3;

[0116] 2.9) Modify the feedback control law output to complete decoupling;

[0117] The modified feedback control law is as follows:

[0118] Formula 30:

[0119] 2.10) Determination of stability of feedback control law;

[0120] The determination process is as follows:

[0121] 2.10.1) Design Lyapunov function:

[0122] Formula 31:

[0123] 2.10.2) Derivative the designed Lyapunov function:

[0124] Formula 32:

[0125] 2.10.3) Whether the control rate is stable when ε1 and ε2 are both greater than zero;

[0126] 2.11) Read the optimization data of shared register 2 and judge the optimization time.

[0127] 2.12) When the optimization time is less than 50ms, directly read the feedback control law tuning parameters; otherwise, read the offline tuning parameter Map;

[0128] 2.13) Output the final control law;

[0129] 2.14) Enter the next loop.

[0130] As a further technical solution of the present invention, the stochastic gradient optimization algorithm is assigned to CPU2 for parallel operation, and the sliding mode control tuning parameters under five step sizes are optimized at the same time, and the average of the five optimized tuning parameters is taken as the final output tuning parameter, and the stability of the tuning parameters is determined at the same time. The steps of updating the stable tuning parameters to cover the preset offline tuning parameter Map include:

[0131] 3.1) Read the current output of the hydrogen system;

[0132] The traditional sliding mode control tuning parameter ε is only applicable to a system with a relative working condition. For the complex and changeable fuel cell hydrogen supply subsystem, the sliding mode control law with constant tuning parameters is difficult to meet the system requirements. Due to the diversity of working conditions, it is difficult to enumerate various tuning parameters by calibration. The gradient descent optimization method is simple and intuitive, suitable for large-scale data, highly practical, and can well meet the parameter optimization requirements of the system. Therefore, the present invention obtains the sliding mode control law tuning parameter ε through the stochastic gradient descent optimization method;

[0133] 3.2) Establish cost function;

[0134] The cost function is as follows:

[0135] 3.2.1) Proportional flow valve:

[0136] Formula 33:

[0137] 3.2.2) Circulation pump

[0138] Formula 34:

[0139] 3.3) Tuning parameter ε iteration:

[0140] The iteration process is as follows:

[0141] 3.3.1) Proportional flow valve:

[0142] Formula 35:

[0143] 3.3.2) Circulation pump:

[0144] Formula 36:

[0145] Among them: μ is used to adjust the iteration step size of each step;

[0146] 3.4) Simultaneously optimize the sliding mode control tuning parameters under five step sizes, and take the average of the five optimized tuning parameters as the final output tuning parameter ε 1,grad , ε 2,grad ;

[0147] 3.5) For ε 1,grad , ε 2,grad Parameter determination, ε 1,grad , ε 2,grad If not all values ​​are less than zero, an interrupt is triggered and CPU1 reads the preset tuning parameter Map;

[0148] 3.6)ε 1,grad , ε 2,grad All are less than zero, and the optimization time is judged;

[0149] 3.7) If the optimization time is greater than 50ms, CPU1 directly reads the preset tuning parameter Map;

[0150] 3.8) The optimization time is less than 50ms, and CPU1 directly reads the data stored in shared register 2;

[0151] 3.9) Update offline tuning Map data;

[0152] 3.10) Output the final control law;

[0153] Formula 37:

[0154] 3.11) Enter the next loop.

[0155] Compared with the prior art, the present invention has the following beneficial effects:

[0156] 1. Taking into account the impact of road traffic information on the output power of the fuel cell vehicle, the output power of the fuel cell stack is predicted and corrected, and then the dynamic feedforward error and the sliding surface of the feedback control law are corrected to improve the economy of the vehicle.

[0157] 2. A feedforward control law based on static feedforward and dynamic feedforward was designed. Static feedforward was used to optimize the flow and pressure fluctuations caused by the hysteresis response of various sub-components of the hydrogen system at the moment of startup of the fuel cell stack. Dynamic feedforward was used to solve the interference caused by transient fluctuations in the power demand of the fuel cell, further improving the stability of control.

[0158] 3. Adaptive sliding mode control is used to compensate for the inaccuracy of the control model. At the same time, the sliding mode control parameters are adaptively adjusted through the gradient descent optimization method. A set of offline tuning parameters is preset to avoid the problem of slow optimization speed affecting the system stability, thereby improving the robustness of the system.

[0159] 4. The flow and pressure coupling interference problem in the hydrogen system is solved. By designing two sets of disturbance observers, the flow and pressure are regarded as interference, and the interference is compensated by the adaptive feedback control law, realizing the decoupling control of the flow and pressure of the hydrogen system. This decoupling does not depend on the accuracy of the model and does not need to accurately describe the impact of the coupling interference on the system. At the same time, two sets of tracking differentiators are designed to optimize the input of the adaptive feedback controller, further reducing the system disturbance.

[0160] 5. Improve the speed of optimization algorithm and reduce the load on vehicle chips by allocating parallel algorithms to multiple cores. BRIEF DESCRIPTION OF THE DRAWINGS

[0161] Figure 1 A schematic diagram of a fuel cell hydrogen control system provided in an embodiment of the present invention.

[0162] Figure 2 This is a schematic diagram of a circulating pump type hydrogen system model provided in an embodiment of the present invention.

[0163] Figure 3 A schematic diagram of task allocation for a multi-core controller provided in an embodiment of the present invention.

[0164] Figure 4 The embodiments of the present invention provide the required excess hydrogen ratio of the anode under different load conditions.

[0165] Figure 5 A schematic diagram of a control block diagram provided by an embodiment of the present invention.

[0166] Figure 6 A comparison diagram of the anode and cathode pressure difference provided in an embodiment of the present invention.

[0167] Figure 7 This is a diagram showing the effect of the stacking pressure control provided by an embodiment of the present invention.

[0168] Figure 8 This is a diagram showing the effect of the inflow flow control provided by an embodiment of the present invention.

[0169] Fig. 9 This is a pressure effect diagram after feedforward correction provided by an embodiment of the present invention.

[0170] Notes on the figure numbers: 1-hydrogen storage bottle, 2-pressure reducing valve, 3-proportional flow valve, 4-fuel cell, 5-steam-water separator, 6-purge valve, 7-pressure sensor, 8-circulation pump, 9-check valve, 10-drain valve. DETAILED DESCRIPTION

[0171] The technical solution of the present invention is further described in detail below in conjunction with specific implementation methods.

[0172] See also Figures 1 to 9 , a networked fuel cell vehicle hydrogen control system, the system comprising:

[0173] Hydrogen storage bottle 1, pressure reducing valve 2, proportional flow valve 3, battery stack 4, steam-water separator 5, purge valve 6, pressure sensor 7, circulation pump 8, check valve 9, drain valve 10;

[0174] Among them, a pressure reducing valve 2 and a proportional flow valve 3 are connected between the hydrogen storage bottle 1 and the battery stack 4, the battery stack 4 is connected to a steam-water separator 5, the steam-water separator 5 is connected to a purge valve 6, the purge valve 6 is connected to a pressure sensor 7, a circulation pump 8 and a one-way valve 9, and the one-way valve 9 is connected to a drain valve 10;

[0175] The pressure reducing valve 2 is used to reduce the high-pressure hydrogen pressure of the hydrogen storage bottle to the value required by the fuel cell, the proportional flow valve 3 is used to adjust the flow of the hydrogen system into the stack, and the circulating pump 8 is used to pressurize and recirculate the hydrogen; the steam-water separator 5 is used to separate the hydrogen and the entrained water at the anode outlet of the fuel cell 4 and discharge them through the drain valve 10, and the purge valve 6 is used to remove the nitrogen that has penetrated and accumulated in the cathode cavity.

[0176] Another object of an embodiment of the present invention is to provide a method for controlling hydrogen in a networked fuel cell vehicle. The method is implemented based on the above-mentioned networked fuel cell vehicle hydrogen control system, and the method includes the following steps:

[0177] 1. Establishment of control model:

[0178] 1) Proportional flow valve modeling, the proportional flow valve formula is as follows:

[0179] Formula 1:

[0180] Where: P pv is the pressure at the front end of the proportional flow valve, T air is the atmospheric temperature, ρ H2 is the density of hydrogen, c 1～3 is a constant, u1 is the control input of the proportional flow valve;

[0181] 2) Intake manifold modeling, the intake manifold formula is as follows:

[0182] Formula 2:

[0183] Among them, V sm is the intake manifold volume, M pv,out is the outlet flow of the proportional flow valve, M pu,out is the outlet flow of the circulating pump, M an,in is the anode inlet flow rate of the stack, R is the molar gas constant, M sm,in Intake manifold inlet flow, M sm,outIntake manifold outlet flow;

[0184] 3) Stack modeling: Since the stack ignores the internal nitrogen and hydrogen permeation phenomena, and ignores the effect of water on the anode pressure and flow, the stack differential equation is as follows:

[0185] Formula 3:

[0186] Formula 4: P an,out =P an -ΔP st ;

[0187] Formula 5:

[0188] Where: P an Anode chamber pressure, M an,in Anode inlet flow, M react Hydrogen flow rate for anode reaction, M an,out Anode outlet hydrogen flow rate, Δp st The pressure drop at the anode outlet of the stack caused by the hydrogen reaction, C w Anode flow field width, C d Anode flow field depth, L anode flow field length, Molar mass of hydrogen, N cell Number of cells in the stack, I req The stack current required, n is the number of anode flow field channels, c 4～10 constant;

[0189] 4) Modeling of the return air manifold. The formula of the return air manifold is as follows:

[0190] Formula 6:

[0191] Where: P rm Return manifold pressure, M rm,in Return manifold inlet flow, M rm,out Return manifold outlet flow, V rm Return manifold volume, T st Stack temperature;

[0192] 5) Circulation pump modeling, the circulation pump formula is as follows:

[0193] Formula 7:

[0194] Where: c 11 is the proportional parameter of the circulating pump speed and the control input voltage, d puis the diameter of the circulating pump blade. The outlet pressure of the circulating pump can be calculated by its isentropic efficiency and mechanical efficiency. The calculation is relatively complicated. Generally, its inlet and outlet pressure-speed map is measured by experiment. The present invention adopts the data of Busch's MINKMA0018A circulating pump. In order to facilitate the design of the controller, the outlet pressure of the circulating pump is set to k pu u2, where k pu is a variable related to the inlet pressure of the circulating pump, and u2 is the control input of the circulating pump.

[0195] This embodiment also includes multi-core controller task allocation, using Infineon TC275TD-Step (DAP) as a fuel cell controller, and allocating fuel cell hydrogen system power correction, feedforward control, disturbance observation, feedback decoupling control and parameter optimization to the three cores of the processor CPU0, CPU1, and CPU2 in parallel. The steps include:

[0196] 1) Assign the battery stack output power correction task and feedforward control algorithm under the network connection to the first core CPU0 of the multi-core processor to run, update the battery stack required power in real time according to road information and SOC, and output the feedforward controller for transient control;

[0197] 1.1) Determine whether the fuel cell stack is in the startup phase. If so, calculate the corresponding static feedforward control law to compensate for the disturbance caused by the hysteresis response of the hydrogen system subcomponents in the startup phase;

[0198] At the moment when the fuel cell starts to start, due to the hysteresis response of each subcomponent and the slow response speed of the flow and pressure input based on the feedback control law, the flow supplied to the stack at the moment of startup cannot meet the needs of the fuel cell, resulting in a large disturbance in the system during the startup phase. Therefore, a static feedforward control law is designed to avoid the problem of slow response of the feedback control law, increase the response speed of the system, and improve the tracking performance of the control input during the startup phase, thereby optimizing the fluctuation of the system during the startup phase;

[0199] The static feedforward controller is designed as follows:

[0200] 1.1.1) Determine the system state variables;

[0201] State 1: Anode inlet pressure:

[0202] Formula 8:

[0203] State 2: Circulation pump speed:

[0204] Formula 9: x2 = c 11 u2;

[0205] 1.1.2) Define the system output: y1 proportional valve outlet mass flow, y2 anode inlet pressure:

[0206] Formula 10:

[0207] 1.1.3) Control input: k pv Proportional flow valve control input voltage, V pv Hydrogen circulation pump motor control input voltage;

[0208] Formula 11:

[0209] 1.1.4) Define φ1, φ2:

[0210] The hydrogen supply subsystem generally supplies excess hydrogen to the anode to prevent the fuel cell efficiency from being reduced and the proton exchange membrane from being damaged due to hydrogen starvation. Different systems have different hydrogen excess ratios. The hydrogen excess ratio reference value of the present invention uses experimental data measured by a 115KW fuel cell stack to calculate the hydrogen flow rate required at the anode inlet under different working conditions. The calculation formula is as follows:

[0211] Formula 12:

[0212] Where: F is Faraday's constant, is the molar mass of hydrogen, N cell is the number of cells in the stack, I req is the current required by the stack; the excess hydrogen required by the anode under different load conditions is as follows Figure 4 As shown:

[0213] Then φ1, φ2 are defined as follows:

[0214] Formula 13:

[0215] 1.1.5) Take a derivative of φ1 and φ2:

[0216] Formula 14:

[0217] At the beginning of the operation of the fuel cell, due to the hysteresis of the entire fuel cell system, the flow rate of the recycled hydrogen after being pressurized by the circulation pump is negligible compared to the flow rate output by the proportional flow valve, so φ1 here ignores the influence of the circulation pump fluid. Simplified as follows:

[0218] Formula 15:

[0219] 1.1.6) Order Thus, the static feedforward control law is obtained:

[0220] Formula 16:

[0221] 1.2) Obtain vehicle location information in real time;

[0222] 1.3) Obtain road information in real time;

[0223] 1.4) Obtain traffic light information in real time;

[0224] 1.5) Obtain power battery SOC in real time;

[0225] 1.6) Determine the future state of the vehicle based on road information, traffic light information and power battery SOC: rapid acceleration, rapid deceleration, slow acceleration, slow deceleration.

[0226] 1.7) Four groups of correction factors are set to correct the stack demand current information sent by the FCU at the next moment;

[0227] 1.8) Update the flow and pressure information of the hydrogen system demand supply and write it into the shared register;

[0228] 1.9) Determine whether there is a step change in the power required by the fuel cell stack. If so, calculate the corresponding dynamic feedforward control law to perform disturbance compensation;

[0229] When the power demand of the fuel cell changes in a step, it will cause large hydrogen flow demand and fluctuations, which in turn will cause large hydrogen pressure fluctuations, resulting in insufficient transient fuel supply to the stack and worsening the working conditions of the stack. Static feedforward alone cannot meet the demand for fuel cell flow and pressure in transient conditions, so a dynamic feedforward control law is designed to eliminate the system disturbance caused by the step change in the power demand, thereby improving the robustness and stability of the system and reducing the burden on the feedback controller. The coupled dynamic disturbance of the circulating pump at this time is solved in the feedback control law and ignored in the dynamic feedforward control law.

[0230] The dynamic feedforward controller is designed as follows:

[0231] 1.9.1) Correct the control input to the following form:

[0232] Formula 17:

[0233] 1.9.2) Correct the required flow rate M an,ref :

[0234] Formula 18:

[0235] Where: M an,pre is the corrected fuel cell demand flow, I pre The corrected battery stack requires current.

[0236] 1.9.3) Yes Derivative:

[0237] Formula 19:

[0238] 1.9.4) Substitute equations 16 and 17 into equation 19 and let The dynamic feedforward control law is derived, and the specific formula is as follows:

[0239] Formula 20:

[0240] 1.10) An interrupt is triggered, and CPU1 reads the updated stack required flow and pressure information and feedforward control information;

[0241] 1.11) Reading completed, clear interrupt;

[0242] 1.12) Enter the next loop.

[0243] 2) A feedback decoupling controller was designed based on the required flow and pressure information of the battery stack predicted and updated by CPU0, and the algorithm was assigned to CPU1 for parallel operation;

[0244] 2.1) Read the corrected stack required pressure and flow. If the reading time exceeds 1ms, interrupt the reading and use the last data;

[0245] 2.2) Tracking differentiator to smooth flow and pressure;

[0246] The differential tracker is a single-input dual-output module, which is designed to smooth the input signal to avoid the coexistence of overshoot and rise time when the input signal changes suddenly. The differential tracker can also achieve filtering at the same time; therefore, two sets of differential trackers are designed to optimize M an,pre and P pu,pre , the input is the M corresponding to the required power of the battery stack an,pre and P pu,pre , output r1 is the processed signal, output r2 is the differential of the output, and the specific formula of the tracking differentiator is designed as follows:

[0247] 2.2.1) Proportional flow valve:

[0248] Formula 21:

[0249] 2.2.2) Circulation pump:

[0250] Formula 22:

[0251] Among them: τ is the sampling period, ξ1,ξ2 are the parameters that determine the tracking speed;

[0252] 2.3) Read the current output of the hydrogen system;

[0253] 2.4) The disturbance observer collects the output for disturbance observation;

[0254] The disturbance observer can not only estimate the unknown disturbance in the system in real time and compensate it in the control loop, but also reduce the mutual influence between the system channels, thereby enhancing the system's ability to suppress disturbances and improving control performance and robustness. Therefore, the present invention designs two sets of disturbance observers, regards the flow control of the proportional flow valve and the pressure control of the circulating pump as disturbances, and decouples the disturbance effect on the adaptive sliding mode controller. The disturbance observer is designed as follows:

[0255] 2.4.1) Proportional flow valve: y1(k) is the proportional flow valve outlet flow:

[0256] Formula 23:

[0257] 2.4.2) Circulation pump: y2(k) is the outlet pressure of the circulation pump:

[0258] Formula 24:

[0259] Where: β is the observer tuning parameter, δ is the length of the linear segment of the fal function, and α is a constant between 0 and 1.

[0260] 2.5) Read the tracking differentiator output r1, r2 and the disturbance observer output z1, z2;

[0261] 2.6) Update the feedback controller input parameters;

[0262] 2.7) Calculation of control law of feedback controller;

[0263] The core of sliding mode control is to design a sliding mode surface so that the system state quickly tends to and keeps sliding on the surface, so as to achieve the purpose of stable control of the system. Fuel cells are complex nonlinear systems with strong coupling. It is difficult to accurately describe the system parameters with a model. Sliding mode control has strong robustness to the uncertainty of system parameters, external disturbances and nonlinear factors, and has fast response speed and strong convergence, which can be well applied to fuel cell systems. Therefore, this paper adopts two sets of first-order sliding mode control as the feedback control law of the system, and decouples and corrects them through tracking differentiators and disturbance observers. The specific design process of sliding mode control is as follows:

[0264] 2.7.1) The error design is as follows:

[0265] Formula 25:

[0266] The disturbance from the circulating pump flow is output by observer 1 as z 1,1Compensation, the pressure disturbance from the proportional flow valve is output by observer 2 1,2 compensate;

[0267] 2.7.2) Design the sliding surface and take n s =1;

[0268] Formula 26:

[0269] 2.7.3) Approach rate: Constant approach rate, and replace the sign function with a saturation function to alleviate the inherent chattering problem of sliding mode control, where ε is the tuning parameter of the sliding mode control law, Parameters for adjusting the thickness of the boundary layer;

[0270] Formula 27:

[0271] 2.7.4) Derivative s and make corrections based on the disturbance observer output:

[0272] Formula 28:

[0273] Combining equations 27 and 28, we can obtain the decoupled sliding mode feedback control law:

[0274] Formula 29:

[0275] 2.8) Read the disturbance observer output z3;

[0276] 2.9) Modify the feedback control law output to complete decoupling;

[0277] The modified feedback control law is as follows:

[0278] Formula 30:

[0279] 2.10) Determination of stability of feedback control law;

[0280] The determination process is as follows:

[0281] 2.10.1) Design Lyapunov function:

[0282] Formula 31:

[0283] 2.10.2) Derivative the designed Lyapunov function:

[0284] Formula 32:

[0285] 2.10.3) Whether the control rate is stable when ε1 and ε2 are both greater than zero;

[0286] 2.11) Read the optimization data of shared register 2 and judge the optimization time.

[0287] 2.12) When the optimization time is less than 50ms, the feedback control law tuning parameters are directly read; otherwise, the offline tuning parameter Map is read;

[0288] 2.13) Output the final control law;

[0289] 2.14) Enter the next loop.

[0290] 3) Assign the stochastic gradient optimization algorithm to CPU2 for parallel operation, and optimize the sliding mode control tuning parameters under five step sizes at the same time. Take the average of the five optimized tuning parameters as the final output tuning parameters, and make a stability judgment on the tuning parameters. The stable tuning parameters are updated to overwrite the preset offline tuning parameter Map.

[0291] 3.1) Read the current output of the hydrogen system;

[0292] The traditional sliding mode control tuning parameter ε is only applicable to a system with a relative working condition. For the complex and changeable fuel cell hydrogen supply subsystem, the sliding mode control law with constant tuning parameters is difficult to meet the system requirements. Due to the diversity of working conditions, it is difficult to enumerate various tuning parameters by calibration. The gradient descent optimization method is simple and intuitive, suitable for large-scale data, highly practical, and can well meet the parameter optimization requirements of the system. Therefore, the present invention obtains the sliding mode control law tuning parameter ε through the stochastic gradient descent optimization method;

[0293] 3.2) Establish cost function;

[0294] The cost function is as follows:

[0295] 3.2.1) Proportional flow valve:

[0296] Formula 33:

[0297] 3.2.2) Circulation pump

[0298] Formula 34:

[0299] 3.3) Tuning parameter ε iteration:

[0300] The iteration process is as follows:

[0301] 3.3.1) Proportional flow valve:

[0302] Formula 35:

[0303] 3.3.2) Circulation pump:

[0304] Formula 36:

[0305] Among them: μ is used to adjust the iteration step size of each step;

[0306] 3.4) Simultaneously optimize the sliding mode control tuning parameters under five step sizes, and take the average of the five optimized tuning parameters as the final output tuning parameter ε 1,grad , ε 2,grad ;

[0307] 3.5) For ε 1,grad , ε 2,grad Parameter determination, ε 1,grad , ε 2,grad If not all values ​​are less than zero, an interrupt is triggered and CPU1 reads the preset tuning parameter Map;

[0308] 3.6)ε 1,grad , ε 2,grad All are less than zero, and the optimization time is judged;

[0309] 3.7) If the optimization time is greater than 50ms, CPU1 directly reads the preset tuning parameter Map;

[0310] 3.8) The optimization time is less than 50ms, and CPU1 directly reads the data stored in shared register 2;

[0311] 3.9) Update offline tuning Map data;

[0312] 3.10) Output the final control law;

[0313] Formula 37:

[0314] 3.11) Enter the next loop.

[0315] The present invention corrects the fuel cell output power according to traffic road information and power battery SOC, optimizes the stack output by correcting the dynamic feedforward error and the sliding surface of the feedback control law, and improves the economy and power. A control-oriented fuel cell hydrogen system model is established, and the control model is simplified through experimental data to facilitate the design of the control law.

[0316] The design of conventional controllers ignores the computing power of on-board chips, especially the optimization algorithm consumes a lot of computing power and is difficult to use on on-board controllers. Therefore, the present invention distributes the designed algorithm to the on-board multi-core chip for parallel operation, improves the algorithm operation speed, and reduces the on-board processor load. Power prediction correction and feedforward control are distributed to CPU0, disturbance observation and decoupling controller are distributed to CPU1, and stochastic gradient descent adaptive optimization is distributed to CPU2. The operation speed is improved and the processor load is reduced through multi-core parallelism.

[0317] A feedforward control law based on static feedforward and dynamic feedforward was designed. The static feedforward optimizes the flow and pressure fluctuation problems caused by the hysteresis response of various sub-components of the hydrogen system at the moment of stack startup. The dynamic feedforward compensates for the interference caused by transient fluctuations in the power demand of the fuel cell, further improving the stability of control.

[0318] The sliding mode control method is used to compensate for the inaccuracy of the control model. The sliding mode control tuning parameters under five step sizes are optimized simultaneously through the stochastic gradient descent method. The average of the five optimized tuning parameters is taken as the final output tuning parameters to reduce the noise interference caused by the gradient descent optimization process. At the same time, the stability of the output tuning parameters is judged, and the preset feedback control rate tuning parameters Map are continuously updated according to the stability judgment results, thereby improving the system's anti-interference ability and high robustness. In response to the coupling problem of flow and pressure in the hydrogen system, two sets of independent differential trackers and disturbance observers are designed. While optimizing the sliding mode control input, the flow, pressure and coupling are regarded as interference with each other, and this interference is compensated by the adaptive sliding mode feedback control law, realizing the decoupling control of the flow and pressure of the hydrogen system.

[0319] In order to verify the performance of the DASMC controller in this paper, a step load current is used to simulate the actual vehicle operating power change, and the performance of traditional sliding mode control SMC, decoupled sliding mode control DSMC and decoupled adaptive sliding mode control DASMC are compared and analyzed. The results show that the designed controller can realize the decoupling control of anode flow and pressure, and stabilize the hydrogen supply flow while reducing the anode-cathode pressure difference. The maximum anode-cathode pressure difference is about 0.01 bar, and the maximum flow supply error is 0.015 g / s; under the stack startup condition, the pressure fluctuation caused by the delayed response of sub-components is reduced by 0.122 bar; when the stack is running with variable load, it can quickly respond to the changes in pressure and flow within 0.02 s.

[0320] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0321] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for controlling hydrogen in a networked fuel cell vehicle, characterized in that the method is implemented based on a networked fuel cell vehicle hydrogen control system, and the method comprises the following steps:

1. Establishment of control model: 1) Proportional flow valve modeling, the proportional flow valve formula is as follows: Formula 1: Where: P pv is the pressure at the front end of the proportional flow valve, T air is the atmospheric temperature, ρ H2 is the density of hydrogen, c 1～3 is a constant, u1 is the control input of the proportional flow valve; 2) Intake manifold modeling, the intake manifold formula is as follows: Formula 2: Among them, V sm is the intake manifold volume, M pv,out is the outlet flow of the proportional flow valve, M pu,out is the outlet flow of the circulating pump, M an,in is the anode inlet flow rate of the stack, R is the molar gas constant, M sm,in Intake manifold inlet flow, M sm,out Intake manifold outlet flow; 3) Stack modeling: Since the stack ignores the internal nitrogen and hydrogen permeation phenomena, and ignores the effect of water on the anode pressure and flow, the stack differential equation is as follows: Formula 3: Formula 4: P an,out =P an -ΔP st ; Formula 5: Where: P an Anode chamber pressure, M an,in Anode inlet flow, M react Hydrogen flow rate for anode reaction, M an,out Anode outlet hydrogen flow rate, Δp st The pressure drop at the anode outlet of the stack caused by the hydrogen reaction, C w Anode flow field width, C d Anode flow field depth, L Anode flow field length, M H2 Molar mass of hydrogen, N cell Number of cells in the stack, I req The stack current required, n is the number of anode flow field channels, c 4～10 constant; 4) Modeling of the return air manifold. The formula of the return air manifold is as follows: Formula 6: Where: P rm Return manifold pressure, M rm,in Return manifold inlet flow, M rm,out Return manifold outlet flow, V rm Return manifold volume, T st Stack temperature; 5) Circulation pump modeling, the circulation pump formula is as follows: Formula 7: Where: c 11 is the proportional parameter of the circulating pump speed and the control input voltage, d pu is the diameter of the circulating pump blades. The outlet pressure of the circulating pump is calculated by its isentropic efficiency and mechanical efficiency. The outlet pressure of the circulating pump is set to k pu u2, where k pu is a variable related to the inlet pressure of the circulating pump, and u2 is the control input of the circulating pump; It also includes multi-core controller task allocation, using Infineon TC275T D-Step as the fuel cell controller, allocating the fuel cell hydrogen system power correction, feedforward control, disturbance observation, feedback decoupling control and parameter optimization to the three cores of the processor CPU0, CPU1, and CPU2. The parallel steps include: 1) Assign the battery stack output power correction task and feedforward control algorithm under the network connection to the first core CPU0 of the multi-core processor to run, update the battery stack required power in real time according to road information and SOC, and output the feedforward controller for transient control; 2) A feedback decoupling controller was designed based on the required flow and pressure information of the battery stack predicted and updated by CPU0, and the algorithm was assigned to CPU1 for parallel operation; 3) Assign the stochastic gradient optimization algorithm to CPU2 for parallel operation, and optimize the sliding mode control tuning parameters under five step sizes at the same time. Take the average of the five optimized tuning parameters as the final output tuning parameters, and determine the stability of the tuning parameters. The stable tuning parameters are updated to cover the preset offline tuning parameter Map; The steps of allocating the network-connected battery stack output power correction task and the feedforward control algorithm to the first core CPU0 of the multi-core processor for execution, updating the battery stack required power in real time according to road information and SOC, and outputting the feedforward controller for transient control include: 1.1) Determine whether the fuel cell stack is in the startup phase. If so, calculate the corresponding static feedforward control law to compensate for the disturbance caused by the hysteresis response of the hydrogen system subcomponents in the startup phase; The static feedforward controller is designed as follows: 1.1.1) Determine the system state variables; State 1: Anode inlet pressure: Formula 8: State 2: Circulation pump speed: Formula 9: x2 = c 11 u2; 1.1.2) Define the system output: y1 proportional valve outlet mass flow, y2 anode inlet pressure: Formula 10: 1.1.3) Control input: k pv Proportional flow valve opening, V pu Hydrogen circulation pump motor control input voltage; Formula 11: 1.1.4) Define φ1, φ2: The hydrogen supply subsystem oversupplies hydrogen into the anode. The required hydrogen flow at the anode inlet under different working conditions is calculated as follows: Formula 12: Where: F is Faraday's constant, is the molar mass of hydrogen, N cell is the number of cells in the stack, I req The current required for the battery stack; Then φ1, φ2 are defined as follows: Formula 13: 1.1.5) Take a derivative of φ1 and φ2: Formula 14: At the beginning of the operation of the fuel cell, due to the hysteresis of the entire fuel cell system, the flow rate of the recycled hydrogen after being pressurized by the circulation pump is negligible compared to the flow rate output by the proportional flow valve, so φ1 here ignores the influence of the circulation pump fluid. Simplified as follows: Formula 15: 1.1.6) Order Thus, the static feedforward control law is obtained: Formula 16: 1.2) Obtain vehicle location information in real time; 1.3) Obtain road information in real time; 1.4) Obtain traffic light information in real time; 1.5) Obtain power battery SOC in real time; 1.6) Determine the future state of the vehicle based on road information, traffic light information and power battery SOC: rapid acceleration, rapid deceleration, slow acceleration, slow deceleration; 1.7) Four groups of correction factors are set to correct the stack demand current information sent by the FCU at the next moment; 1.8) Update the flow and pressure information of the hydrogen system demand supply and write it into the shared register; 1.9) Determine whether there is a step change in the power required by the fuel cell stack. If so, calculate the corresponding dynamic feedforward control law to perform disturbance compensation; The dynamic feedforward controller is designed as follows: 1.9.1) Correct the control input to the following form: Formula 17: 1.9.2) Correct the required flow rate M an,ref : Formula 18: Where: M an,pre is the corrected fuel cell demand flow, I pre The current required for the battery stack after correction; 1.9.3) Yes Derivative: Formula 19: 1.9.4) Substitute equations 16 and 17 into equation 19 and let The dynamic feedforward control law is derived, and the specific formula is as follows: Formula 20: 1.10) An interrupt is triggered, and CPU1 reads the updated stack required flow and pressure information and feedforward control information; 1.11) Reading completed, clear interrupt; 1.12) Enter the next cycle; The steps of designing a feedback decoupling controller according to the required flow and pressure information of the battery stack updated by CPU0 and distributing the algorithm to CPU1 for parallel operation include: 2.1) Read the corrected stack required pressure and flow. If the reading time exceeds 1ms, interrupt the reading and use the last data; 2.2) Tracking differentiator to smooth flow and pressure; The differential tracker is a single-input dual-output module designed to smooth the input signal. Two sets of differential trackers are designed to optimize M an,pre and P pu,pre , the input is the M corresponding to the required power of the battery stack an,pre and P pu,pre , output r1 is the processed signal, output r2 is the differential of the output, and the specific formula of the tracking differentiator is designed as follows: 2.2.1) Proportional flow valve: Formula 21: 2.2.2) Circulation pump: Formula 22: Among them: τ is the sampling period, ξ1,ξ2 are the parameters that determine the tracking speed; 2.3) Read the current output of the hydrogen system; 2.4) The disturbance observer collects the output for disturbance observation; Two sets of disturbance observers are designed. The flow control of the proportional flow valve and the pressure control of the circulating pump are regarded as disturbances to each other. The disturbance effect is decoupled on the adaptive sliding mode controller. The disturbance observer is designed as follows: 2.4.1) Proportional flow valve: y1(k) is the proportional flow valve outlet flow: Formula 23: 2.4.2) Circulation pump: y2(k) is the outlet pressure of the circulation pump: Formula 24: Where: β is the observer tuning parameter, δ is the length of the linear segment of the fal function, and α is a constant between 0 and 1; 2.5) Read the tracking differentiator output r1, r2 and the disturbance observer output z1, z2; 2.6) Update the feedback controller input parameters; 2.7) Calculation of control law of feedback controller; The specific design process of the sliding mode decoupling controller is as follows: 2.7.1) Error design is as follows: Formula 25: The disturbance from the circulating pump flow is output by observer 1 as z 1,1 Compensation, the pressure disturbance from the proportional flow valve is output by observer 2 1,2 compensate; 2.7.2) Design the sliding surface and take n s =1; Formula 26: 2.7.3) Approach rate: Constant approach rate, and replace the sign function with a saturation function to alleviate the inherent chattering problem of sliding mode control, where ε is the tuning parameter of the sliding mode control law, Parameters for adjusting the thickness of the boundary layer; Formula 27: 2.7.4) Derivative s and make corrections based on the disturbance observer output: Formula 28: Combining equations 27 and 28, we can obtain the decoupled sliding mode feedback control law: Formula 29: 2.8) Read the disturbance observer output z3; 2.9) Modify the feedback control law output to complete decoupling; The modified feedback control law is as follows: Formula 30: 2.10) Determination of stability of feedback control law; The determination process is as follows: 2.10.1) Design Lyapunov function: Formula 31: 2.10.2) Derivative the designed Lyapunov function: Formula 32: 2.10.3) Whether the control rate is stable when ε1 and ε2 are both greater than zero; 2.11) Read the optimization data of shared register 2 and judge the optimization time. 2.12) When the optimization time is less than 50ms, directly read the feedback control law tuning parameters; otherwise, read the offline tuning parameter Map; 2.13) Output the final control law; 2.14) Enter the next cycle; The steps of assigning the stochastic gradient optimization algorithm to CPU2 for parallel operation, optimizing the sliding mode control tuning parameters under five step sizes, taking the average of the five optimized tuning parameters as the final output tuning parameters, and determining the stability of the tuning parameters, and updating the stable tuning parameters to cover the preset offline tuning parameter Map include: 3.1) Read the current output of the hydrogen system; The tuning parameter ε of the sliding mode control law is obtained by the stochastic gradient descent optimization method; 3.2) Establish cost function; The cost function is as follows: 3.2.1) Proportional flow valve: Formula 33: 3.2.2) Circulation pump: Formula 34: 3.3) Tuning parameter ε iteration: The iteration process is as follows: 3.3.1) Proportional flow valve: Formula 35: 3.3.2) Circulation pump: Formula 36: in: μ is used to adjust the iteration step size of each step; 3.4) Simultaneously optimize the sliding mode control tuning parameters under five step sizes, and take the average of the five optimized tuning parameters as the final output tuning parameter ε 1,grad , ε 2,grad ; 3.5) For ε 1,grad , ε 2,grad Parameter determination, ε 1,grad , ε 2,grad If not all values ​​are less than zero, an interrupt is triggered and CPU1 reads the preset tuning parameter Map; 3.6)ε 1,grad , ε 2,grad All are less than zero, and the optimization time is judged; 3.7) If the optimization time is greater than 50ms, CPU1 directly reads the preset tuning parameter Map; 3.8) The optimization time is less than 50ms, and CPU1 directly reads the data stored in shared register 2; 3.9) Update offline tuning Map data; 3.10) Output the final control law; Formula 37: 3.11) Enter the next loop.

2. A method for controlling hydrogen in a fuel cell vehicle connected to a network according to claim 1, characterized in that: A networked fuel cell vehicle hydrogen control system includes: Hydrogen storage bottle, pressure reducing valve, proportional flow valve, fuel cell, steam-water separator, purge valve, pressure sensor, circulation pump, check valve, drain valve; Wherein, a pressure reducing valve and a proportional flow valve are connected between the hydrogen storage bottle and the battery stack, the battery stack is connected to a steam-water separator, the steam-water separator is connected to a purge valve, the purge valve is connected to a pressure sensor, a circulation pump and a one-way valve, and the one-way valve is connected to a drain valve; The pressure reducing valve is used to reduce the high-pressure hydrogen pressure of the hydrogen storage bottle to the value required by the fuel cell, the proportional flow valve is used to adjust the flow of the hydrogen system into the stack, and the circulating pump is used to pressurize and recirculate the hydrogen; the steam-water separator is used to separate the hydrogen and the entrained water at the anode outlet of the fuel cell stack and discharge them through the drain valve, and the purge valve is used to remove the nitrogen that has penetrated and accumulated in the cathode cavity.

Images