CN117622095A - Hybrid system energy management method based on optimization of air supply condition of fuel cell - Google Patents

Abstract

The invention provides a hybrid system energy management method based on optimization of air supply conditions of a fuel cell, which designs a vehicle speed predictor based on a bidirectional LSTM, and can obtain more accurate required power of the whole vehicle in a short-term time domain in the future; calculating the optimal air supply condition offline by using a particle swarm optimization algorithm, so as to realize higher system economy; the optimal air supply condition is considered in the updating calculation of the cost function of the energy management strategy, so that the energy management process is more in line with the actual working condition of the fuel cell hybrid system, various defects caused by regarding the polarization characteristic and the hydrogen consumption characteristic of the electric pile as a fixed form in the prior art are overcome, and the practicability is higher. The invention can accurately predict the future vehicle speed under various different working conditions and reasonably distribute the energy of the hybrid power source, thereby effectively optimizing the hydrogen fuel economy of the hybrid power system on the premise of ensuring the normal output of the fuel cell system.

Description

Hybrid system energy management method based on optimization of air supply condition of fuel cell

Technical Field

The invention belongs to the technical field of energy management of fuel cell hybrid electric vehicles, and particularly relates to an energy management method of a hybrid system based on optimization of air supply conditions of a fuel cell.

Background

Currently, fuel cell systems used in hybrid vehicles generally have problems of slow power response, unidirectional power generation, and energy recovery, so that there is still a need to combine energy storage devices such as batteries to improve dynamic performance and efficiency of the power system. Energy management strategies can play a critical role in terms of the rational distribution of power between the fuel cell system and the battery, as well as the efficiency of operation of the hybrid power system. In the prior art of energy management of some fuel cell hybrid systems, the polarization characteristics of the fuel cell stack and the hydrogen consumption characteristics of the fuel cell system are mostly regarded as fixed forms, whereas in actual fuel cell systems, different air supply conditions will lead to variations in stack and system output performance. Therefore, how to optimize the hybrid system energy management method for the actual fuel cell air supply conditions, so that the economy of such a power system is improved, is a technical problem that needs to be solved urgently in the art.

Disclosure of Invention

In view of the above, the present invention provides a hybrid system energy management method based on optimization of air supply conditions of a fuel cell, specifically comprising the following steps:

s1, respectively establishing a fuel cell stack model, a cathode air supply system model, a storage battery equivalent circuit model and a vehicle dynamics model aiming at a fuel cell hybrid system; the fuel cell stack model specifically comprises: a galvanic pile polarization characteristic and hydrogen consumption model; the cathode air supply system model specifically includes: an air compressor model, a cathode supply pipeline model, a cathode flow channel model, a backflow pipeline model and a back pressure valve model;

s2, designing a particle swarm offline optimization algorithm aiming at the relation between different cathode peroxy ratios and cathode pressures and net output power of the fuel cell system, and calculating the optimal peroxy ratio and cathode pressure; fitting the optimal air supply conditions corresponding to different load current ranges by combining the load current;

s3, training a two-way long-short-term memory network (LSTM) by utilizing vehicle road circulation working condition data to obtain a predictor, and performing online prediction on vehicle speed information in a future short-term time domain; inputting current vehicle speed and historical vehicle speed information measured in real time by a vehicle sensor into the predictor to obtain a predicted vehicle speed information prediction result, and calculating to obtain the vehicle demand power in a future short-term time domain on the basis of the predicted vehicle speed information prediction result;

s4, establishing a Model Predictive Control (MPC) energy management strategy based on each model established in the step S1, wherein the power of the fuel cell system and the SOC of the storage battery are used as state quantities, the change rate of the output power of the fuel cell system is used as a manipulated variable, and the required power of the whole vehicle is used as disturbance quantity; minimizing the hydrogen consumption of a fuel cell hybrid system as an objective function of an energy management strategy for predicting the power of the fuel cell system and the SOC of a storage battery; converting the objective function into a constraint quadratic programming problem, and solving the constraint quadratic programming problem through an active set algorithm to obtain an optimal control sequence of the system;

s5, extracting a first variable of an optimal control sequence, and acquiring current disturbance information through the historical polarization characteristics of the fuel cell stack at the previous moment; updating the stack polarization characteristics and the fuel cell hydrogen consumption model in combination with the optimal air supply conditions in the step S2, and executing the step S4 again, and finally applying the first variable of the obtained optimal control sequence for minimizing the hydrogen consumption to the vehicle control system; and (3) returning to the step (S3) to repeatedly execute the corresponding steps along with the continuous operation of the vehicle.

Further, step S2 specifically includes:

s21, designing and executing a particle swarm offline optimization algorithm:

(1) Initializing the population scale of a particle swarm and the position and speed of each particle, setting the upper limit and the lower limit of the position and the update speed, an acceleration constant and the maximum iteration times; setting the net power of the fuel cell system as an adaptive function;

(2) Calculating a current peroxide ratio for each particle, a cathode pressure update rate, and a corresponding value for each particle:

v _ij ＝wv _ij +c ₁ r ₁ (pBest _ij -x _ij )+c ₂ r ₂ (gBest _i -x _ij )

x _ij (t+1)＝x _ij (t)+v _ij (t+1)

wherein i represents the ith particle, j represents the jth dimension of the particle, w is the inertial weight, c ₁ And c ₂ For acceleration constant r ₁ And r ₂ Is a random number, v _ij Is the current peroxide ratio, cathode pressure update rate, x _ij Is the value of each particle, pBest is the optimal individual extremum, gBest is the optimal global extremum, and t is the time variable;

s22, fitting the optimal air supply condition:

the following table of fuel cell cathode air supply parameters was obtained by fitting the calculated optimum peroxide ratio and cathode pressure in combination with load current:

wherein lambda is _O2 To the peroxide ratio, y ₁ ，y ₂ ,…，y _n For optimum peroxide ratio, R ₁ ，R ₂ ,…，R _n For optimum cathode pressure value, P _ca For cathode pressure, P _atm Is at ambient pressure, I _st Is the load current.

Further, the step S3 specifically includes:

s31, establishing a training data set for training a bidirectional LSTM network:

the method comprises the steps of splicing seven road circulation working condition data of CHTC_ B, CHTC _ D, CHTC _ C, CHTC _LT, CHTC_TT, CHTC_HT and UDDSHDV to form a running database, extracting data from the running database, executing normalization processing, and forming a training set and a testing set by utilizing the data after the normalization processing;

s32, building and training a bidirectional LSTM network structure:

aiming at an input layer, a forward layer, a reverse layer and an output layer which form a bidirectional LSTM network, respectively defining the number of neurons, initial weights and thresholds of each layer; inputting training set data into a network for training until the prediction precision requirement is met, then testing and evaluating the trained network prediction effect by using the data of a test set, and if the prediction precision requirement is met, storing the trained network as a predictor for real-time prediction of the vehicle speed;

s33, online predicting the vehicle speed:

and inputting the current vehicle speed and the historical vehicle speed information measured by the vehicle sensor in real time into the vehicle speed predicted by the predictor, and calculating to obtain the vehicle demand power in the future short-term time domain.

Further, the step S4 specifically includes:

s41, building an energy management strategy model and determining an objective function:

the predictive model of the energy management strategy is described as a linearized discrete-time state space model of the form:

wherein x (k) = [ P ] _fc (k-1)，SOC(k)] ^T ，u(k)＝ΔP _fc (k)，d(k)＝P _req (k)，y(k)＝[P _fc (k),SOC(k)] ^T ；A、B _u 、B _d C, D are matrix coefficients, P _fc For fuel cell system power, ΔP _fc For the rate of change of the power of the fuel cell system, i.e. the manipulated variable, P _req The power required by the whole vehicle, namely the disturbance quantity, and k is a discrete time variable;

the corresponding objective function takes the following form:

wherein J is _MPC As a goal of the energy management strategy,equivalent hydrogen consumption rate for hybrid power system, SOC ₀ For initial SOC, t _f To predict the time domain, t _m To control the time domain, alpha ₁ 、α ₂ 、α ₃ All are constant coefficients;

s42, solving a quadratic programming:

the objective function is converted into the following quadratic programming problem and solved by an active set algorithm:

wherein,

the hessian matrix of the quadratic programming function is a positive definite matrix, the optimal problem is strict convex quadratic programming, and the quadratic programming function has a unique global minimum value in a prediction range.

Further, the step S5 specifically includes:

s51, forward calculation:

extracting a first variable of the optimal control sequence calculated in the step S4, and acquiring current disturbance information through historical polarization characteristics of the electric pile at the previous moment, namely the relation between the output voltage and the current density of the fuel cell; calculating and updating the polarization characteristic of the electric pile and the hydrogen consumption model of the fuel cell by combining the optimal air supply parameters obtained in the step S2;

s52, solving the objective function again by using the updated fuel cell hydrogen consumption model, re-acquiring the minimum optimal control sequence, and applying the first variable in the minimum optimal control sequence to the vehicle control system; as the vehicle continues to run, a new vehicle speed prediction is started in step S3, and the following steps are repeatedly executed.

According to the hybrid system energy management method based on the optimization of the air supply condition of the fuel cell, provided by the invention, the vehicle speed predictor based on the bidirectional LSTM is designed, so that the accurate required power of the whole vehicle in the future short-term time domain can be obtained; calculating the optimal air supply condition offline by using a particle swarm optimization algorithm, so as to realize higher system economy; the optimal air supply condition is considered in the updating calculation of the energy management strategy objective function, so that the energy management process is more in line with the actual working condition of the fuel cell hybrid system, various defects caused by regarding the polarization characteristic and the hydrogen consumption characteristic of the electric pile as a fixed form in the prior art are overcome, and the practicability is higher. The invention can accurately predict the future vehicle speed under various different working conditions and reasonably distribute the energy of the hybrid power source, thereby effectively optimizing the hydrogen fuel economy of the hybrid power system on the premise of ensuring the normal output of the fuel cell system.

Drawings

FIG. 1 is a block flow diagram of a method provided by the present invention;

FIG. 2 is a schematic diagram of a prediction of a vehicle speed in a future short-term time domain in an example according to the present invention;

FIG. 3 is a schematic diagram of an optimal air supply condition calculated in an example according to the present invention;

fig. 4 is a graph showing hydrogen consumption characteristics corresponding to different air supply conditions obtained based on examples of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made apparent and fully in view of the accompanying drawings, in which some, but not all embodiments of the invention are shown. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The hybrid system energy management method based on the optimization of the air supply condition of the fuel cell provided by the invention, as shown in fig. 1, specifically comprises the following steps:

s1, respectively establishing a fuel cell stack model, a cathode air supply system model, a storage battery equivalent circuit model and a vehicle dynamics model aiming at a fuel cell hybrid system; the fuel cell stack model specifically comprises: a galvanic pile polarization characteristic and hydrogen consumption model; the cathode air supply system model specifically includes: an air compressor model, a cathode supply pipeline model, a cathode flow channel model, a backflow pipeline model and a back pressure valve model;

s2, designing a particle swarm offline optimization algorithm aiming at the relation between different cathode peroxy ratios and cathode pressures and net output power of the fuel cell system, and calculating the optimal peroxy ratio and cathode pressure; fitting the optimal air supply conditions corresponding to different load current ranges by combining the load current;

s3, training a two-way long-short-term memory network (LSTM) by utilizing vehicle road circulation working condition data to obtain a predictor, and performing online prediction on vehicle speed information in a future short-term time domain; inputting current vehicle speed and historical vehicle speed information measured in real time by a vehicle sensor into the predictor to obtain a predicted vehicle speed information prediction result, and calculating to obtain the vehicle demand power in a future short-term time domain on the basis of the predicted vehicle speed information prediction result;

s4, establishing a Model Predictive Control (MPC) energy management strategy based on each model established in the step S1, wherein the power of the fuel cell system and the SOC of the storage battery are used as state quantities, the change rate of the output power of the fuel cell system is used as a manipulated variable, and the required power of the whole vehicle is used as disturbance quantity; minimizing the hydrogen consumption of a fuel cell hybrid system as an objective function of an energy management strategy for predicting the power of the fuel cell system and the SOC of a storage battery; converting the objective function into a constraint quadratic programming problem, and solving the constraint quadratic programming problem through an active set algorithm to obtain an optimal control sequence of the system;

s5, extracting a first variable of an optimal control sequence, and acquiring current disturbance information through the historical polarization characteristics of the fuel cell stack at the previous moment; updating the stack polarization characteristics and the fuel cell hydrogen consumption model in combination with the optimal air supply conditions in the step S2, and executing the step S4 again, and finally applying the first variable of the obtained optimal control sequence for minimizing the hydrogen consumption to the vehicle control system; and (3) returning to the step (S3) to repeatedly execute the corresponding steps along with the continuous operation of the vehicle.

In a preferred embodiment of the present invention, step S2 specifically includes:

s21, designing and executing a particle swarm offline optimization algorithm:

(1) Initializing the population scale of a particle swarm and the position and speed of each particle, setting the upper limit and the lower limit of the position and the update speed, an acceleration constant and the maximum iteration times; setting the net power of the fuel cell system as an adaptive function;

(2) Calculating a current peroxide ratio for each particle, a cathode pressure update rate, and a corresponding value for each particle:

v _ij ＝wv _ij +c ₁ r ₁ (pBest _ij -x _ij )+c ₂ r ₂ (gBest _i -x _ij )

x _ij (t+1)＝x _ij (t)+v _ij (t+1)

wherein i represents the ith particle, j represents the jth dimension of the particle, w is the inertial weight, c ₁ And c ₂ For acceleration constant r ₁ And r ₂ Is a random number, v _ij Is the current peroxide ratio, cathode pressure update rate, x _ij Is the value of each particle, pBest is the optimal individual extremum, gBest is the optimal global extremum, and t is the time variable;

s22, fitting the optimal air supply condition:

the following table of fuel cell cathode air supply parameters was obtained by fitting the calculated optimum peroxide ratio and cathode pressure in combination with load current:

wherein lambda is _O2 To the peroxide ratio, y ₁ ，y ₂ ,…，y _n For optimum peroxide ratio, R ₁ ，R ₂ ,…，R _n For optimum cathode pressure value, P _ca For cathode pressure, P _atm Is at ambient pressure, I _st Is the load current.

In a preferred embodiment of the present invention, step S3 specifically includes:

s31, establishing a training data set for training a bidirectional LSTM network:

the method comprises the steps of splicing seven road circulation working condition data of CHTC_ B, CHTC _ D, CHTC _ C, CHTC _LT, CHTC_TT, CHTC_HT and UDDSHDV to form a running database, extracting data from the running database, executing normalization processing, and forming a training set and a testing set by utilizing the data after the normalization processing;

s32, building and training a bidirectional LSTM network structure:

aiming at an input layer, a forward layer, a reverse layer and an output layer which form a bidirectional LSTM network, respectively defining the number of neurons, initial weights and thresholds of each layer; inputting training set data into a network for training until the prediction precision requirement is met, then testing and evaluating the trained network prediction effect by using the data of a test set, and if the prediction precision requirement is met, storing the trained network as a predictor for real-time prediction of the vehicle speed;

s33, online predicting the vehicle speed:

and inputting the current vehicle speed and the historical vehicle speed information measured by the vehicle sensor in real time into the vehicle speed predicted by the predictor, and calculating to obtain the vehicle demand power in the future short-term time domain.

In a preferred embodiment of the present invention, step S4 specifically includes:

s41, building an energy management strategy model and determining an objective function:

the predictive model of the energy management strategy is described as a linearized discrete-time state space model of the form:

wherein x (k) = [ P ] _fc (k-1)，SOC(k)] ^T ，u(k)＝ΔP _fc (k)，d(k)＝P _req (k)，y(k)＝[P _fc (k),SOC(k)] ^T ；A、B _u 、B _d C, D are matrix coefficients, P _fc For fuel cell system power, ΔP _fc For the rate of change of the power of the fuel cell system, i.e. the manipulated variable, P _req The power required by the whole vehicle, namely the disturbance quantity, and k is a discrete time variable;

the corresponding objective function takes the following form:

wherein J is _MPC As a goal of the energy management strategy,equivalent hydrogen consumption rate for hybrid power system, SOC ₀ For initial SOC, t _f For prediction of time domain, it can be chosen to be 3, t _m For controlling the time domain, it can be chosen to be 2, α ₁ 、α ₂ 、α ₃ All are constant coefficients;

s42, solving a quadratic programming:

the objective function is converted into the following quadratic programming problem and solved by an active set algorithm:

wherein,

the hessian matrix of the quadratic programming function is a positive definite matrix, the optimal problem is strict convex quadratic programming, and the quadratic programming function has a unique global minimum value in a prediction range.

In a preferred embodiment of the present invention, step S5 specifically includes:

s51, forward calculation:

extracting a first variable of the optimal control sequence calculated in the step S4, and acquiring current disturbance information through historical polarization characteristics of the electric pile at the previous moment, namely the relation between the output voltage and the current density of the fuel cell; calculating and updating the polarization characteristic of the electric pile and the hydrogen consumption model of the fuel cell by combining the optimal air supply parameters obtained in the step S2;

s52, solving the objective function again by using the updated fuel cell hydrogen consumption model, re-acquiring the minimum optimal control sequence, and applying the first variable in the minimum optimal control sequence to the vehicle control system; as the vehicle continues to run, a new vehicle speed prediction is started in step S3, and the following steps are repeatedly executed.

Specifically, fig. 2 shows a result of predicting a vehicle speed in a future short-term time domain in an example according to the present invention.

Specifically, fig. 3 (a) shows the optimum operating point of the air compressor, and fig. 3 (b) shows the optimum fuel cell cathode air supply conditions at different load currents.

Specifically, fig. 4 shows a comparison of hydrogen consumption characteristics of the fuel cell system corresponding to the two air supply conditions of the peroxy ratio of 1.9, the pressure ratio of 1.4, and the peroxy ratio of 1.8, and the pressure ratio of 2.0, respectively.

It should be understood that, the sequence number of each step in the embodiment of the present invention does not mean that the execution sequence of each process should be determined by the function and the internal logic of each process, and should not limit the implementation process of the embodiment of the present invention.

Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (5)

1. A hybrid system energy management method based on optimization of fuel cell air supply conditions, characterized by: the method specifically comprises the following steps:

s1, respectively establishing a fuel cell stack model, a cathode air supply system model, a storage battery equivalent circuit model and a vehicle dynamics model aiming at a fuel cell hybrid system; the fuel cell stack model specifically comprises: a galvanic pile polarization characteristic and hydrogen consumption model; the cathode air supply system model specifically includes: an air compressor model, a cathode supply pipeline model, a cathode flow channel model, a backflow pipeline model and a back pressure valve model;

s2, designing a particle swarm offline optimization algorithm aiming at the relation between different cathode peroxy ratios and cathode pressures and net output power of the fuel cell system, and calculating the optimal peroxy ratio and cathode pressure; fitting the optimal air supply conditions corresponding to different load current ranges by combining the load current;

s3, training the bidirectional LSTM by utilizing vehicle road circulation working condition data to obtain a predictor for carrying out online prediction on the vehicle speed information in a future short-term time domain; inputting current vehicle speed and historical vehicle speed information measured in real time by a vehicle sensor into the predictor to obtain a predicted vehicle speed information prediction result, and calculating to obtain the vehicle demand power in a future short-term time domain on the basis of the predicted vehicle speed information prediction result;

s4, establishing an MPC energy management strategy based on each model established in the step S1, wherein the power of the fuel cell system and the SOC of the storage battery are used as state quantities, the change rate of the output power of the fuel cell system is used as a manipulated variable, and the required power of the whole vehicle is used as disturbance quantity; minimizing the hydrogen consumption of a fuel cell hybrid system as an objective function of an energy management strategy for predicting the power of the fuel cell system and the SOC of a storage battery; converting the objective function into a constraint quadratic programming problem, and solving the constraint quadratic programming problem through an active set algorithm to obtain an optimal control sequence of the system;

s5, extracting a first variable of an optimal control sequence, and acquiring current disturbance information through the historical polarization characteristics of the fuel cell stack at the previous moment; updating the stack polarization characteristics and the fuel cell hydrogen consumption model in combination with the optimal air supply conditions in the step S2, and executing the step S4 again, and finally applying the first variable of the obtained optimal control sequence for minimizing the hydrogen consumption to the vehicle control system; and (3) returning to the step (S3) to repeatedly execute the corresponding steps along with the continuous operation of the vehicle.

2. The method of claim 1, wherein: the step S2 specifically comprises the following steps:

s21, designing and executing a particle swarm offline optimization algorithm:

(1) Initializing the population scale of a particle swarm and the position and speed of each particle, and setting the upper limit, the lower limit and the acceleration constant of the position and the update speed and the maximum iteration times; setting the net power of the fuel cell system as an adaptive function;

(2) Calculating a current peroxide ratio for each particle, a cathode pressure update rate, and a corresponding value for each particle:

v _ij ＝wv _ij +c ₁ r ₁ (pBest _ij -x _ij )+c ₂ r ₂ (gBest _i -x _ij )

x _ij (t+1)＝x _ij (t)+v _ij (t+1)

wherein i represents the ith particle, j represents the jth dimension of the particle, w is the inertial weight, c ₁ And c ₂ For acceleration constant r ₁ And r ₂ Is a random number, v _ij Is the current peroxide ratio, cathode pressure update rate, x _ij Is the value of each particle, pBest is the optimal individual extremum, gBest is the optimal global extremum, and t is the time variable;

s22, fitting the optimal air supply condition:

the following table of fuel cell cathode air supply parameters was obtained by fitting the calculated optimum peroxide ratio and cathode pressure in combination with load current:

wherein lambda is _O2 To the peroxide ratio, y ₁ ，y ₂ ,…，y _n For optimum peroxide ratio, R ₁ ，R ₂ ,…，R _n For optimum cathode pressure value, P _ca For cathode pressure, P _atm Is at ambient pressure, I _st Is the load current.

3. The method of claim 1, wherein: the step S3 specifically comprises the following steps:

s31, establishing a training data set for training a bidirectional LSTM network:

the method comprises the steps of splicing seven road circulation working condition data of CHTC_ B, CHTC _ D, CHTC _ C, CHTC _LT, CHTC_TT, CHTC_HT and UDDSHDV to form a running database, extracting data from the running database, executing normalization processing, and forming a training set and a testing set by utilizing the data after the normalization processing;

s32, building and training a bidirectional LSTM network structure:

aiming at an input layer, a forward layer, a reverse layer and an output layer which form a bidirectional LSTM network, respectively defining the number of neurons, initial weights and thresholds of each layer; inputting training set data into a network for training until the prediction precision requirement is met, then testing and evaluating the trained network prediction effect by using the data of a test set, and if the prediction precision requirement is met, storing the trained network as a predictor for real-time prediction of the vehicle speed;

s33, online predicting the vehicle speed:

and inputting the current vehicle speed and the historical vehicle speed information measured by the vehicle sensor in real time into the vehicle speed predicted by the predictor, and calculating to obtain the vehicle demand power in the future short-term time domain.

4. The method of claim 1, wherein: the step S4 specifically comprises the following steps:

s41, building an energy management strategy model and determining an objective function:

the predictive model of the energy management strategy is described as a linearized discrete-time state space model of the form:

wherein x (k) = [ P ] _fc (k-1)，SOC(k)] ^T ，u(k)＝ΔP _fc (k)，d(k)＝P _req (k)，y(k)＝[P _fc (k),SOC(k)] ^T ；A、B _u 、B _d C, D are matrix coefficients, P _fc For fuel cell system power, ΔP _fc For the rate of change of the power of the fuel cell system, i.e. the manipulated variable, P _req The power required by the whole vehicle, namely the disturbance quantity, and k is a discrete time variable;

the corresponding objective function takes the following form:

wherein J is _MPC As a goal of the energy management strategy,equivalent hydrogen consumption rate for hybrid power system, SOC ₀ For initial SOC, t _f To predict the time domain, t _m To control the time domain, alpha ₁ 、α ₂ 、α ₃ All are constant coefficients;

s42, solving a quadratic programming:

the objective function is converted into the following quadratic programming problem and solved by an active set algorithm:

s.t.AU≤b

wherein,

the hessian matrix of the quadratic programming function is a positive definite matrix, the optimal problem is strict convex quadratic programming, and the quadratic programming function has a unique global minimum value in a prediction range.

5. The method of claim 1, wherein: the step S5 specifically comprises the following steps:

s51, forward calculation:

extracting a first variable of the optimal control sequence calculated in the step S4, and acquiring current disturbance information through historical polarization characteristics of the electric pile at the previous moment, namely the relation between the output voltage and the current density of the fuel cell; calculating and updating the polarization characteristic of the electric pile and the hydrogen consumption model of the fuel cell by combining the optimal air supply parameters obtained in the step S2;

s52, solving the objective function again by using the updated fuel cell hydrogen consumption model, re-acquiring the minimum optimal control sequence, and applying the first variable in the minimum optimal control sequence to the vehicle control system; as the vehicle continues to run, a new vehicle speed prediction is started in step S3, and the following steps are repeatedly executed.