An intelligent vehicle cold start control method

Abstract

The present invention relates to the field of vehicle control technology, and discloses an intelligent automobile cold start control method, comprising the following steps: step S101, when the temperature of a fuel cell stack is less than a first temperature threshold, entering step S102; step S102, generating a characteristic matrix; step S103, inputting the characteristic matrix into a first control model, and outputting an opening adjustment value of an electromagnetic hydrogen exhaust valve and a speed adjustment value of an air compressor; step S104, when the output voltage of the fuel cell stack is greater than or equal to a voltage threshold, entering step S105; step S105, generating a characteristic sequence; step S106, inputting the characteristic sequence into a second control model, and outputting an output power adjustment value of a power battery; step S107, when the temperature of the fuel cell stack is greater than or equal to a second temperature threshold, the vehicle cold start ends; the present invention can dynamically adjust the vehicle through the first control model and the second control model, thereby ensuring the output stability and efficiency of the entire fuel stack and improving the vehicle cold start speed.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle control, and more specifically, to an intelligent automobile cold start control method. Background Art

[0002] With the development of new energy technologies, fuel cell electric vehicles have gradually become one of the important options for solving traffic pollution and energy shortage problems. They convert chemical energy into electrical energy through the reaction of fuel and oxidant, thereby driving the operation of fuel cell electric vehicles, which have the advantages of high efficiency, zero emissions and long driving range. However, in low temperature environments, the chemical reaction rate of fuel cells decreases, the conductivity of electrolytes deteriorates, and the water vapor generated by fuel cells may freeze, resulting in slow startup of fuel cell electric vehicles, which seriously affects the user's driving experience.

[0003] The Chinese patent with publication number CN114347863B and patent name "Vehicle Cold Start Method, Device, Readable Storage Medium, Electronic Device and Vehicle" discloses: when the power battery temperature is lower than the first temperature threshold, the power battery heating is started; when the power battery temperature is greater than or equal to the second temperature threshold, the fuel cell heating is started, and the fuel cell and the power battery can be heated and started; when the fuel cell temperature is greater than or equal to the third temperature threshold or the power battery temperature is less than the fourth temperature threshold, the vehicle's heating load is turned on to consume the electric energy generated by the fuel cell to avoid charging the power battery too early and causing overcurrent of the power battery.

[0004] The Chinese patent with publication number CN116512993B and patent name "A control method, device, vehicle and medium for a fuel cell cold start system" discloses: applying ultra-low voltage loading to the fuel cell so that the output voltage of the fuel cell is stabilized within the minimum voltage value for normal operation in a low-temperature environment, and calculating the output power of the fuel cell and the required driving power of the power motor. When the output power of the fuel cell is greater than the required driving power, and the required driving power is greater than the rated power of the power battery heater, the power motor drive is given priority, and the remaining power is supplied to the fuel cell for heating; when the required driving power is less than the rated power of the power battery heater, the power motor drive is given priority, followed by the output power of the fuel cell being the rated power, and finally the remaining power is supplied to the power battery for heating until the fuel cell reaches the temperature threshold, indicating that the cold start of the fuel cell is completed.

[0005] Both of the above solutions are based on preset thresholds for vehicle cold start control. During the actual vehicle cold start process, environmental conditions and vehicle status may change continuously, and a single threshold judgment may not be able to meet the dynamic needs of the vehicle. In addition, the fuel stack of a fuel cell electric vehicle is usually composed of multiple battery cells. The differences between battery cells may cause voltage imbalance and affect the output stability and efficiency of the entire fuel stack. Summary of the invention

[0006] The present invention provides an intelligent automobile cold start control method to solve the technical problems in the above-mentioned background technology.

[0007] The present invention provides an intelligent automobile cold start control method, comprising the following steps:

[0008] Step S101, after receiving the start command, obtain the temperature of the fuel cell stack, and if it is determined that the temperature is less than the first temperature threshold, then enter step S102, otherwise the vehicle starts normally;

[0009] Step S102, simultaneously turning on the fuel cell stack heater and the electromagnetic hydrogen exhaust valve heater, and collecting the operating parameters of each battery cell in the fuel cell stack at a preset time interval t within a preset time period T to generate a feature matrix; the element in the jth column of the i-th row of the feature matrix represents the voltage value, current value, temperature value and humidity value of the i-th battery cell at the j-th time point, where 1≤i≤M, 1≤j≤N, M represents the number of battery cells, N represents the number of time points, and N=T / t;

[0010] Step S103, inputting the characteristic matrix into the first control model, and outputting values ​​respectively representing the opening adjustment value of the electromagnetic hydrogen exhaust valve and the speed adjustment value of the air compressor in the first control time period;

[0011] Step S104, adjusting the opening of the electromagnetic hydrogen exhaust valve to the opening adjustment value within the first control time period, adjusting the speed of the air compressor to the speed adjustment value, and obtaining the output voltage of the fuel cell stack. If the output voltage is less than the voltage threshold, then repeating steps S102 to S104, otherwise proceeding to step S105;

[0012] Step S105, within a preset time period T, the operating parameters of the vehicle are collected at preset time intervals t to generate a feature sequence; the j-th sequence unit of the feature sequence represents the temperature, output voltage, output current, hydrogen supply pressure, air supply pressure, coolant temperature and ambient temperature of the fuel cell stack at the j-th time point;

[0013] Step S106, inputting the characteristic sequence into the second control model, and the output value represents the output power adjustment value of the power battery in the second control time period;

[0014] Step S107, adjust the output power of the power battery to the output power adjustment value within the second control time period, and obtain the temperature of the fuel cell stack and the vehicle driving demand power, and determine that its temperature is greater than or equal to the second temperature threshold and the output power of the power battery is greater than or equal to the vehicle driving demand power, then it indicates that the vehicle cold start is completed, otherwise repeat steps S105 to S107.

[0015] Furthermore, the first temperature threshold, the second temperature threshold, the voltage threshold, the preset time period T, the preset time interval t, the first control time period, the second control time period and the vehicle driving demand power are all custom parameters.

[0016] Furthermore, the vehicle driving demand power is equal to the sum of the vehicle idle power, the on-board electrical system load power, the power required by the fuel cell stack heater and the power required by the electromagnetic hydrogen exhaust valve heater; the idle torque and idle speed are obtained respectively by installing a torque sensor and a speed sensor on the output shaft of the power motor, and the idle power of the vehicle is obtained by multiplying the idle torque and the idle speed, and the on-board electrical system load power, the fuel cell stack heater and the electromagnetic hydrogen exhaust valve heater are obtained by an electronic power meter, and the on-board electrical system includes lighting, air conditioning and audio.

[0017] Further, the first control model includes N hidden layers, each hidden layer includes a first unit, a second unit and a third unit, the first unit includes H channels, where H is a custom parameter;

[0018] Each channel of the first unit of the u-th hidden layer inputs the elements of the 1st row to the Mth row of the u-th column of the feature matrix, that is, the voltage value, current value, temperature value and humidity value of the 1st battery cell to the Mth battery cell at the u-th time point, and each channel of the first unit of the u-th hidden layer outputs the first updated feature;

[0019] The second unit of the u-th hidden layer inputs the first updated feature output by each channel of the first unit of the u-th hidden layer, and outputs the second updated feature;

[0020] The third unit of the u-th hidden layer inputs the second updated feature output by the second unit of the u-th hidden layer, and outputs the third updated feature;

[0021] The third updated feature output by the third unit of the Nth hidden layer is input into the first classifier and the second classifier. The classification space of the first classifier represents the opening adjustment value of the electromagnetic hydrogen exhaust valve within the first control time period, and the classification space of the second classifier represents the speed adjustment value of the air compressor within the first control time period.

[0022] Furthermore, the calculation formula of the first control model includes:

[0023] ;

[0024] ;

[0025] ;

[0026] ;

[0027] ;

[0028] Definition: 1≤u≤N, 1≤v≤H, H represents the number of channels in the first unit, represents the second updated feature of the output of the second unit of the u-th hidden layer, represents the first updated feature of the v-th channel output of the first unit of the u-th hidden layer, It means concatenating the first updated features of the H channels output by the first unit of the u-th hidden layer. represents the weight parameter of the u-th hidden layer, , and They represent the first intermediate feature, the second intermediate feature, and the third intermediate feature of the vth channel of the first unit of the uth hidden layer, respectively. , and They respectively represent the first weight parameter, the second weight parameter and the third weight parameter of the vth channel of the first unit of the uth hidden layer, Represents the elements from the 1st row to the Mth row of the uth column of the feature matrix of the vth channel input of the first unit of the uth hidden layer, T represents the transpose operation of the matrix, and softmax represents the softmax activation function.

[0029] Furthermore, the third unit of each hidden layer of the first control model is constructed based on a long short-term memory network unit.

[0030] Furthermore, the second control model includes N time steps, the kth time step inputs the kth sequence unit of the feature sequence, where 1≤k≤N, and the output of the Nth time step is input to the third classifier, and the classification space of the third classifier represents the output power adjustment value of the power battery within the second control time period.

[0031] Furthermore, each time step of the second control model is constructed based on a gated recurrent network unit.

[0032] Furthermore, the training samples for training the first control model are obtained through a simulation platform, including the following steps:

[0033] Step S201, under the premise that the temperature of the fuel cell stack is less than a first temperature threshold, the simulation platform randomly generates operating parameters for each battery cell in the fuel cell stack;

[0034] Step S202, the simulation platform randomly generates A opening adjustment values ​​of the electromagnetic hydrogen exhaust valves and the speed adjustment value of the air compressor, and adjusts the opening of the electromagnetic hydrogen exhaust valve to the opening adjustment value and the speed of the air compressor to the speed adjustment value within the first control time period; wherein A is a custom parameter;

[0035] Step S203, recording the temperature and output voltage of A fuel cell stacks in the first control time period respectively, and obtaining the opening adjustment value of the electromagnetic hydrogen exhaust valve and the speed adjustment value of the air compressor corresponding to the maximum temperature or the maximum output voltage of the fuel cell stack as a sample label of a training sample;

[0036] Step S204, repeating steps S201 to S203 until B training samples are obtained; wherein B is a custom parameter.

[0037] Furthermore, the training samples for training the second control model are obtained through a simulation platform, including the following steps:

[0038] Step S301, under the premise that the output voltage of the fuel cell stack is greater than or equal to the voltage threshold, the simulation platform randomly generates the operating parameters of the vehicle;

[0039] Step S302, the simulation platform randomly generates output power adjustment values ​​of C power batteries, and adjusts the output power of the power batteries to the output power adjustment values ​​within a second control time period; wherein C is a custom parameter;

[0040] Step S303, recording the temperatures of the fuel cell stacks in C second control time periods respectively, and obtaining the output power adjustment value of the power battery corresponding to the maximum temperature of the fuel cell stack as a sample label of a training sample;

[0041] Step S304, repeating steps S301 to S303 until D training samples are obtained; wherein D is a custom parameter.

[0042] The beneficial effects of the present invention are as follows: the present invention establishes a nonlinear mapping relationship between the operating parameters of each battery cell in the fuel cell stack and the opening of the electromagnetic hydrogen exhaust valve and the speed of the air compressor through the first control model, establishes a nonlinear mapping relationship between the operating parameters of the vehicle and the output power of the power battery through the second control model, and dynamically adjusts the vehicle through the first control model and the second control model, thereby ensuring the output stability and efficiency of the entire fuel stack and improving the cold start speed of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of an intelligent automobile cold start control method of the present invention;

[0044] Figure 2It is a flow chart of obtaining a training sample for training a first control model through a simulation platform of the present invention;

[0045] Figure 3 It is a flow chart of obtaining training samples for training the second control model through a simulation platform of the present invention. DETAILED DESCRIPTION

[0046] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0047] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0048] like Figure 1-Figure 3 As shown, an intelligent automobile cold start control method comprises the following steps:

[0049] Step S101, after receiving the start command, obtain the temperature of the fuel cell stack, and if it is determined that the temperature is less than the first temperature threshold, then enter step S102, otherwise the vehicle starts normally;

[0050] Step S102, simultaneously turning on the fuel cell stack heater and the electromagnetic hydrogen exhaust valve heater, and collecting the operating parameters of each battery cell in the fuel cell stack at a preset time interval t within a preset time period T to generate a characteristic matrix;

[0051] The element in the jth column of the i-th row of the feature matrix represents the voltage value, current value, temperature value and humidity value of the i-th battery cell at the jth time point, where 1≤i≤M, 1≤j≤N, M represents the number of battery cells, N represents the number of time points, and N=T / t;

[0052] Step S103, inputting the characteristic matrix into the first control model, and outputting values ​​respectively representing the opening adjustment value of the electromagnetic hydrogen exhaust valve and the speed adjustment value of the air compressor in the first control time period;

[0053] Step S104, adjusting the opening of the electromagnetic hydrogen exhaust valve to the opening adjustment value within the first control time period, adjusting the speed of the air compressor to the speed adjustment value, and obtaining the output voltage of the fuel cell stack. If the output voltage is less than the voltage threshold, then repeating steps S102 to S104, otherwise proceeding to step S105;

[0054] Step S105, within a preset time period T, collecting the operating parameters of the vehicle at preset time intervals t to generate a feature sequence;

[0055] The jth sequence unit of the feature sequence represents the temperature, output voltage, output current, hydrogen supply pressure, air supply pressure, coolant temperature and ambient temperature of the fuel cell stack at the jth time point;

[0056] Step S106, inputting the characteristic sequence into the second control model, and the output value represents the output power adjustment value of the power battery in the second control time period;

[0057] Step S107, adjust the output power of the power battery to the output power adjustment value within the second control time period, and obtain the temperature of the fuel cell stack and the vehicle driving demand power, and determine that its temperature is greater than or equal to the second temperature threshold and the output power of the power battery is greater than or equal to the vehicle driving demand power, then it indicates that the vehicle cold start is completed, otherwise repeat steps S105 to S107.

[0058] It should be noted that, during the vehicle cold start control process, steps S102 to S104 represent low-voltage loading of the fuel cell stack to the minimum operating voltage to ensure the normal operation of the fuel cell stack, and steps S105 to S107 represent dynamic adjustment of the output power of the power battery to meet the vehicle's driving demand power, and gradually increase the temperature of the fuel cell stack, thereby achieving rapid starting of the vehicle under low temperature conditions; and the cold start control method provided by the present invention is targeted at PEMFC (proton exchange membrane fuel cell) type of fuel cell, but is not limited to PEMFC, and is also applicable to DMFC (methanol fuel cell) or PAFC (phosphoric acid fuel cell), etc.

[0059] In one embodiment of the present invention, the temperature of the fuel cell stack is obtained by an electronic temperature sensor. In order to improve the accuracy of the temperature measurement, it should be avoided to install it near the exhaust pipe or the cooling pipe. In addition, multiple electronic temperature sensors can also be installed to simultaneously obtain the temperature of the fuel cell stack, and the average temperature is calculated as the actual temperature of the fuel cell stack.

[0060] In one embodiment of the present invention, both the fuel cell stack heater and the electromagnetic hydrogen exhaust valve heater are PTC (positive temperature coefficient thermistor) heaters.

[0061] In one embodiment of the present invention, the first temperature threshold, the second temperature threshold, the voltage threshold, the preset time period T, the preset time interval t, the first control time period, the second control time period and the vehicle drive demand power are all custom parameters; preferably, the first temperature threshold is set to 0°, the second temperature threshold is set to the normal operating temperature of the fuel cell stack, such as 50°, the voltage threshold is set to the minimum operating voltage of the fuel cell stack, such as 0.5V, the preset time period T is set to 10 seconds, the preset time interval t is set to 1 second, the first control time period and the second control time period are both set to 10 seconds, and the vehicle drive demand power can also be set to the lower limit of the normal operating power of the vehicle, for example, if the normal operating power of the vehicle is between 10 and 20kW, the vehicle drive demand power is set to 10kW.

[0062] In one embodiment of the present invention, the vehicle driving demand power is equal to the sum of the vehicle idle power, the on-board electrical system load power, the power required by the fuel cell stack heater and the power required by the electromagnetic hydrogen exhaust valve heater; the idle torque and idle speed are obtained respectively by installing a torque sensor and a speed sensor on the output shaft of the power motor, and the idle power of the vehicle is obtained by multiplying the idle torque and the idle speed, and the on-board electrical system load power, the fuel cell stack heater and the electromagnetic hydrogen exhaust valve heater are obtained by an electronic power meter, and the on-board electrical system includes lighting, air conditioning and audio.

[0063] In one embodiment of the present invention, the first control model includes N hidden layers, each hidden layer includes a first unit, a second unit and a third unit, the first unit includes H channels, where H is a custom parameter, preferably, H is set to 8;

[0064] Each channel of the first unit of the u-th hidden layer inputs the elements of the 1st row to the Mth row of the u-th column of the feature matrix, that is, the voltage value, current value, temperature value and humidity value of the 1st battery cell to the Mth battery cell at the u-th time point, and each channel of the first unit of the u-th hidden layer outputs the first updated feature;

[0065] The second unit of the u-th hidden layer inputs the first updated feature output by each channel of the first unit of the u-th hidden layer, and outputs the second updated feature;

[0066] The third unit of the u-th hidden layer inputs the second updated feature output by the second unit of the u-th hidden layer, and outputs the third updated feature;

[0067] The third updated feature output by the third unit of the Nth hidden layer is input into the first classifier and the second classifier. The classification space of the first classifier represents the opening adjustment value of the electromagnetic hydrogen exhaust valve within the first control time period, and the classification space of the second classifier represents the speed adjustment value of the air compressor within the first control time period.

[0068] In one embodiment of the present invention, the calculation formula of the first control model includes:

[0069] ;

[0070] ;

[0071] ;

[0072] ;

[0073] ;

[0074] Definition: 1≤u≤N, 1≤v≤H, H represents the number of channels in the first unit, represents the second updated feature of the output of the second unit of the u-th hidden layer, represents the first updated feature of the v-th channel output of the first unit of the u-th hidden layer, It means concatenating the first updated features of the H channels output by the first unit of the u-th hidden layer. represents the weight parameter of the u-th hidden layer, , and They represent the first intermediate feature, the second intermediate feature, and the third intermediate feature of the vth channel of the first unit of the uth hidden layer, respectively. , and They respectively represent the first weight parameter, the second weight parameter and the third weight parameter of the vth channel of the first unit of the uth hidden layer, Represents the elements from the 1st row to the Mth row of the uth column of the feature matrix of the vth channel input of the first unit of the uth hidden layer, T represents the transpose operation of the matrix, and softmax represents the softmax activation function.

[0075] In one embodiment of the present invention, the third unit of each hidden layer of the first control model is constructed based on an LSTM (Long Short-Term Memory Network) unit. LSTM is a conventional technical means and will not be described in detail here.

[0076] In one embodiment of the present invention, the second control model includes N time steps, the kth time step inputs the kth sequence unit of the feature sequence, where 1≤k≤N, and the output of the Nth time step is input to the third classifier, and the classification space of the third classifier represents the output power adjustment value of the power battery within the second control time period.

[0077] In one embodiment of the present invention, each time step of the second control model is constructed based on a GRU (Gated Recurrent Network) unit. GRU is a conventional technical means and will not be described in detail here.

[0078] In one embodiment of the present invention, Figure 2 As shown, the training samples for training the first control model are obtained through a simulation platform, including the following steps:

[0079] Step S201, under the premise that the temperature of the fuel cell stack is less than a first temperature threshold, the simulation platform randomly generates operating parameters for each battery cell in the fuel cell stack;

[0080] Step S202, the simulation platform randomly generates A opening adjustment values ​​of the electromagnetic hydrogen exhaust valves and the speed adjustment value of the air compressor, and adjusts the opening of the electromagnetic hydrogen exhaust valve to the opening adjustment value and the speed of the air compressor to the speed adjustment value within the first control time period; wherein A is a custom parameter, preferably, A is set to 5;

[0081] Step S203, recording the temperature and output voltage of A fuel cell stacks in the first control time period respectively, and obtaining the opening adjustment value of the electromagnetic hydrogen exhaust valve and the speed adjustment value of the air compressor corresponding to the maximum temperature or the maximum output voltage of the fuel cell stack as a sample label of a training sample;

[0082] Step S204, repeat steps S201 to S203 until B training samples are obtained; wherein B is a custom parameter, preferably, B is set to 2000.

[0083] In one embodiment of the present invention, Figure 3 As shown, the training samples for training the second control model are obtained through a simulation platform, including the following steps:

[0084] Step S301, under the premise that the output voltage of the fuel cell stack is greater than or equal to the voltage threshold, the simulation platform randomly generates the operating parameters of the vehicle;

[0085] Step S302, the simulation platform randomly generates output power adjustment values ​​of C power batteries, and adjusts the output power of the power batteries to the output power adjustment values ​​within a second control time period; wherein C is a custom parameter, and preferably, C is set to 5;

[0086] Step S303, recording the temperatures of the fuel cell stacks in C second control time periods respectively, and obtaining the output power adjustment value of the power battery corresponding to the maximum temperature of the fuel cell stack as a sample label of a training sample;

[0087] Step S304, repeat steps S301 to S303 until D training samples are obtained; wherein D is a custom parameter, and preferably, D is set to 2000.

[0088] It should be noted that the parameters and adjustment values ​​randomly generated by the simulation platform are consistent with the actual cold start conditions of fuel cell electric vehicles. For example, the voltage value of the battery cell of the fuel cell stack cannot exceed its maximum output voltage, and the opening of the electromagnetic hydrogen exhaust valve cannot exceed its maximum opening.

[0089] It should be noted that the difference between the value outputted by the first control model and the second control model in each iteration and the sample label of its training sample is used as the loss function, and the gradient of the parameters in the first control model and the second control model is calculated by the chain rule. The gradient descent algorithm updates the parameters of the first control model and the second control model according to the gradient to minimize the loss function. The above steps are repeated until the first control model and the second control model converge or the number of iterations reaches the maximum number of iterations, wherein the gradient descent algorithm can be SGD (stochastic gradient descent) or Adam (adaptive moment estimation optimization algorithm), etc.

[0090] Furthermore, the above idea can be used to control the temperature and opening angle of the preheating grille during the cold start of the fuel cell electric vehicle to prevent heat loss and ensure that the fuel cell stack heats up quickly.

[0091] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. An intelligent automobile cold start control method, characterized in that: The following steps are involved: Step S101, after receiving the start command, obtain the temperature of the fuel cell stack, and if it is determined that the temperature is less than the first temperature threshold, then enter step S102, otherwise the vehicle starts normally; Step S102, simultaneously turning on the fuel cell stack heater and the electromagnetic hydrogen exhaust valve heater, and collecting the operating parameters of each battery cell in the fuel cell stack at a preset time interval t within a preset time period T to generate a characteristic matrix; feature The element in the jth column of the i-th row of the matrix represents the voltage value, current value, temperature value and humidity value of the i-th battery cell at the jth time point, where 1≤i≤M, 1≤j≤N, M represents the number of battery cells, N represents the number of time points, and N=T / t; Step S103, inputting the characteristic matrix into the first control model, and outputting values ​​respectively representing the opening adjustment value of the electromagnetic hydrogen exhaust valve and the speed adjustment value of the air compressor in the first control time period; Step S104, adjusting the opening of the electromagnetic hydrogen exhaust valve to the opening adjustment value within the first control time period, adjusting the speed of the air compressor to the speed adjustment value, and obtaining the output voltage of the fuel cell stack. If the output voltage is less than the voltage threshold, then repeating steps S102 to S104, otherwise proceeding to step S105; Step S105, within a preset time period T, collecting the operating parameters of the vehicle at preset time intervals t to generate a feature sequence; The jth sequence unit of the feature sequence represents the temperature, output voltage, output current, hydrogen supply pressure, air supply pressure, coolant temperature and ambient temperature of the fuel cell stack at the jth time point; Step S106, inputting the characteristic sequence into the second control model, and the output value represents the output power adjustment value of the power battery in the second control time period; Step S107, adjusting the output power of the power battery to the output power adjustment value within the second control time period, and obtaining the temperature of the fuel cell stack and the required driving power of the vehicle, and judging that the temperature is greater than or equal to the second temperature threshold and the output power of the power battery is greater than or equal to the required driving power of the vehicle, it indicates that the vehicle cold start is completed, otherwise, repeating steps S105 to S107; The first control model includes N hidden layers, each hidden layer includes a first unit, a second unit and a third unit, the first unit includes H channels, where H is a custom parameter; Each channel of the first unit of the u-th hidden layer inputs the elements of the 1st row to the Mth row of the u-th column of the feature matrix, that is, the voltage value, current value, temperature value and humidity value of the 1st battery cell to the Mth battery cell at the u-th time point, and each channel of the first unit of the u-th hidden layer outputs the first updated feature; The second unit of the u-th hidden layer inputs the first updated feature output by each channel of the first unit of the u-th hidden layer, and outputs the second updated feature; The third unit of the u-th hidden layer inputs the second updated feature output by the second unit of the u-th hidden layer, and outputs the third updated feature; The third updated feature output by the third unit of the Nth hidden layer is input into the first classifier and the second classifier, the classification space of the first classifier represents the opening adjustment value of the electromagnetic hydrogen exhaust valve in the first control time period, and the classification space of the second classifier represents the speed adjustment value of the air compressor in the first control time period; The second control model includes N time steps, the kth sequence unit of the kth time step input feature sequence, where 1≤k≤N, the output of the Nth time step is input to the third classifier, and the classification space of the third classifier represents the output power adjustment value of the power battery within the second control time period.

2. The intelligent automobile cold start control method according to claim 1, characterized in that: The first temperature threshold, the second temperature threshold, the voltage threshold, the preset time period T, the preset time interval t, the first control time period, the second control time period and the vehicle driving demand power are all custom parameters.

3. The intelligent automobile cold start control method according to claim 1, characterized in that: The vehicle driving demand power is equal to the sum of the vehicle idle power, the on-board electrical system load power, the power required by the fuel cell stack heater and the power required by the electromagnetic hydrogen exhaust valve heater; the idle torque and idle speed are obtained by installing a torque sensor and a speed sensor on the output shaft of the power motor, respectively, and the idle torque and the idle speed are multiplied to obtain the vehicle idle power, and the on-board electrical system load power, the fuel cell stack heater and the electromagnetic hydrogen exhaust valve heater are obtained by an electronic power meter, and the on-board electrical system includes lighting, air conditioning and audio.

4. The intelligent automobile cold start control method according to claim 1, characterized in that: The calculation formula of the first control model includes: ； ； ； ； ； Definition: 1≤u≤N, 1≤v≤H, H represents the number of channels in the first unit, represents the second updated feature of the output of the second unit of the u-th hidden layer, represents the first updated feature of the v-th channel output of the first unit of the u-th hidden layer, It means concatenating the first updated features of the H channels output by the first unit of the u-th hidden layer. represents the weight parameter of the u-th hidden layer, , and They represent the first intermediate feature, the second intermediate feature, and the third intermediate feature of the vth channel of the first unit of the uth hidden layer, respectively. , and They respectively represent the first weight parameter, the second weight parameter and the third weight parameter of the vth channel of the first unit of the uth hidden layer, Represents the elements from the 1st row to the Mth row of the uth column of the feature matrix of the vth channel input of the first unit of the uth hidden layer, T represents the transpose operation of the matrix, and softmax represents the softmax activation function.

5. The intelligent automobile cold start control method according to claim 1, characterized in that: The third unit of each hidden layer of the first control model is constructed based on a long short-term memory network unit.

6. The intelligent automobile cold start control method according to claim 1, characterized in that: Each time step of the second control model is built based on gated recurrent network units.

7. The intelligent automobile cold start control method according to claim 1, characterized in that: The training samples for training the first control model are obtained through a simulation platform, including the following steps: Step S201, under the premise that the temperature of the fuel cell stack is less than a first temperature threshold, the simulation platform randomly generates operating parameters for each battery cell in the fuel cell stack; Step S202, the simulation platform randomly generates A opening adjustment values ​​of the electromagnetic hydrogen exhaust valves and the speed adjustment value of the air compressor, and adjusts the opening of the electromagnetic hydrogen exhaust valve to the opening adjustment value and the speed of the air compressor to the speed adjustment value within the first control time period; wherein A is a custom parameter; Step S203, recording the temperature and output voltage of A fuel cell stacks in the first control time period respectively, and obtaining the opening adjustment value of the electromagnetic hydrogen exhaust valve and the speed adjustment value of the air compressor corresponding to the maximum temperature or the maximum output voltage of the fuel cell stack as a sample label of a training sample; Step S204, repeating steps S201 to S203 until B training samples are obtained; wherein B is a custom parameter.

8. The intelligent automobile cold start control method according to claim 1, characterized in that: The training samples for training the second control model are obtained through a simulation platform, including the following steps: Step S301, under the premise that the output voltage of the fuel cell stack is greater than or equal to the voltage threshold, the simulation platform randomly generates the operating parameters of the vehicle; Step S302, the simulation platform randomly generates output power adjustment values ​​of C power batteries, and adjusts the output power of the power batteries to the output power adjustment values ​​within a second control time period; wherein C is a custom parameter; Step S303, recording the temperatures of the fuel cell stacks in C second control time periods respectively, and obtaining the output power adjustment value of the power battery corresponding to the maximum temperature of the fuel cell stack as a sample label of a training sample; Step S304, repeating steps S301 to S303 until D training samples are obtained; wherein D is a custom parameter.

Images