Extended-range fuel cell electric vehicle power device and control method therefor

Abstract

An extended-range fuel cell electric vehicle power device includes a driving motor, a bidirectional converter, a chopper, a power cell, a fuel cell, a high-pressure hydrogen storage tank, an electric control valve, a controller, an accelerator pedal and a brake pedal. An output of the driving motor is connected to a transmission shaft of an electric vehicle through a speed change gearbox, and an input of the driving motor is connected to an alternating current output end of the bidirectional converter; a direct current input end of the bidirectional converter is connected in parallel to an output of the power cell and an output of the chopper, and an input of the chopper is connected to a power source output of the fuel cell.

Claims

1. An extended-range fuel cell electric vehicle power device, comprising a driving motor (1), a bidirectional converter (2), a chopper (3), a power cell (4), a fuel cell (5), a high-pressure hydrogen storage tank (6), an electric control valve (7), a controller (8), an accelerator pedal (9) and a brake pedal (10), wherein: an output of the driving motor is connected to a transmission shaft of an electric vehicle through a speed change gearbox, an input of the driving motor is connected to an alternating current output end of the bidirectional converter, a direct current input end of the bidirectional converter is connected in parallel to an output of the power cell and an output of the chopper, an input of the chopper is connected to a power source output of the fuel cell, a fuel inlet of the fuel cell is connected to an output of the high-pressure hydrogen storage tank through the electric control valve, an output of the controller is connected to a control port of the electric control valve and a control port of the bidirectional converter respectively, and an input end of the controller is connected to output signals of the accelerator pedal and the brake pedal respectively.

2. The extended-range fuel cell electric vehicle power device according to claim 1, wherein the controller is a BP neural network controller, and a structure of a BP neural network is a 2-5-2 structure, which means that, 2 neurons are provided in an input layer, 5 neurons are provided in a middle layer, and 2 neurons are provided in an output layer; and the input layer is used for signal transmission and outputs two controlled quantities y.sub.1 and y.sub.2.

3. The extended-range fuel cell electric vehicle power device according to claim 1, wherein the power cell is a lithium ion cell.

4. A method for controlling an extended-range fuel cell electric vehicle power device, wherein when a given power Pin is positive, an electric vehicle works in a driving state, when the given power Pin is negative, the electric vehicle works in a braking state, an actual output power Pout of a driving motor drives the electric vehicle, a signal of the actual output power Pout of the driving motor is fed back to an input end and compared with the given power Pin to form a feedback control system, the signal of the Pout and a signal of the Pin are inputted into a comparator to obtain a power error signal ΔP, and differentiation processing is performed on the power error signal ΔP to obtain a power change rate signal dP/dt, and two controlled quantities y.sub.1 and y.sub.2 are obtained from the ΔP and the dP/dt by a control algorithm, wherein the controlled quantity y.sub.1 controls an opening degree of an electric control valve, thus controlling an output electric power of a fuel cell; and the controlled quantity y.sub.2 controls a PWM signal of a control port of a bidirectional converter, thus controlling a working mode and a power of the bidirectional converter.

5. The method for controlling the extended-range fuel cell electric vehicle power device according to claim 4, wherein the controlling the working mode of the bidirectional converter comprises controlling an inversion state when the electric vehicle works in the driving state and a rectification state when the electric vehicle works in the braking state; in the inversion state, the driving motor works in an electric state, and a power cell converts a direct current into an alternating current through the bidirectional converter to provide electric power for the driving motor; in the rectification state, the driving motor works in a power generation state and converts energy generated during braking of the electric vehicle into electric energy, and an alternating current generated by the driving motor is rectified into a direct current through the bidirectional converter to charge the power cell; and when a signal is generated by an accelerator pedal, the driving motor works in the power generation state, and when a signal is generated by a brake pedal, the driving motor works in the electric state.

6. The method for controlling the extended-range fuel cell electric vehicle power device according to claim 4, wherein the control algorithm which is a BP neural network control algorithm comprises the following steps of: (1) establishing a structure of a BP neural network wherein a structure of the BP neural network is a 2-5-2 structure, which means that, a number of neurons in an input layer is that i=1,2, x.sub.i corresponds to two input variables, and the input layer is used for signal transmission; the first input variable is: x.sub.1:ΔP; the second input variable is: x.sub.2:dP/dt; a number of neurons in a middle layer is that j=1,2,3,4,5; and an input of the neurons in the middle layer is x.sub.j and an output of the neurons in the middle layer is x′.sub.j; a number of neurons in an output layer is that l=1,2 an input is y.sub.l, and y.sub.l corresponds to two controlled quantities y.sub.1 and y.sub.2; and (2) training the BP network.

7. The method for controlling the extended-range fuel cell electric vehicle power device according to claim 6, wherein the step (2) comprises: (21) forward propagation: calculating an output of the BP neural network, wherein the input of the neurons in the middle layer is a weighted sum of all the inputs, namely: $\begin{array}{cc}{x}_j=\underset{i=1}{\overset{2}{.Math.}}{\omega }_{ij}{x}_i;& \left(1\right)\end{array}$ the output x′.sub.j of the neurons in the middle layer excites x.sub.j by using an S function to obtain: $\begin{array}{cc}{x}_j^{\prime }=f\left(x_j\right)=\frac{1}{1+e^{-x}};\mathrm{then}\text{:}& \left(2\right)\\ \frac{\partial x_j^{\prime }}{\partial x_j}=x_j^{\prime }\left(1-x_j^{\prime }\right);& \left(3\right)\end{array}$ the output of the neurons in the output layer is that: $\begin{array}{cc}{y}_l=\underset{j}{\overset{5}{.Math.}}{\omega }_{jl}{x}_j^{\prime };& \left(4\right)\end{array}$ an error between a l.sup.th output of the BP neural network and a corresponding ideal output y.sub.l.sup.0 is that:
e.sub.l=y.sub.l.sup.0−y.sub.l  (5); taking a p.sup.th sample as an example, an error performance index function of the p.sup.th sample is that: $\begin{array}{cc}{E}_p=\frac{1}{2}\underset{l=1}{\overset{5}{.Math.}}{e}_l^2;& \left(6\right)\end{array}$ and (22) backward propagation: adjusting weights among the layers by gradient descent, wherein a learning algorithm of a connection weight ω.sub.jl between the output layer and the middle layer is that: $\begin{array}{cc}{\mathrm{\Delta \omega }}_{jl}=-\eta \frac{\partial E_p}{\partial {\omega }_{jl}}=\eta e_l\frac{\partial x_l}{\partial {\omega }_{jl}}=\eta e_l{x}_j^{\prime };& \left(7\right)\end{array}$ wherein η is a learning rate, η∈[0,1]; and Δω.sub.jl is a variation of the connection weight ω.sub.jl between the output layer and the middle layer; network weight at a moment k+1 is that:
ω.sub.jl(k+1)=ω.sub.jl(k)+Δω.sub.jl  (8); a learning algorithm of a connection weight ω.sub.ij between the middle layer and the input layer is that: $\begin{array}{cc}{\mathrm{\Delta \omega }}_{ij}=-\eta \frac{\partial E_p}{\partial {\omega }_{ij}}=\eta \underset{l=1}{\overset{5}{.Math.}}{e}_l\frac{\partial x_l}{\partial {\omega }_{ij}};\mathrm{wherein}\frac{\partial x_l}{\partial {\omega }_{ij}}=\frac{\partial x_l}{\partial x_j^{\prime }}.Math.\frac{\partial x_j^{\prime }}{\partial x_j}.Math.\frac{\partial x_j}{\partial {\omega }_{ij}}={\omega }_{il}.Math.\frac{\partial x_j^{\prime }}{\partial x_j}.Math.x_i={\omega }_{il}.Math.x_j^{\prime }\left(1-x_j^{\prime }\right).Math.x_i;& \left(9\right)\end{array}$ and Δω.sub.ij is a variation of the connection weight ω.sub.ij between the middle layer and the input layer; and a network weight at a moment t+1 is that:
ω.sub.ij(k+1)=ω.sub.ij(k)+Δω.sub.ij  (10).

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 is a structure diagram of the present invention;

(2) FIG. 2 is a diagram illustrating a control principle of the present invention;

(3) FIG. 3 is a flow chart illustrating energy control of an extended-range high-temperature fuel cell electric vehicle of the present invention; and

(4) FIG. 4 is a schematic diagram illustrating a control structure a PB neural network of the present invention.

DETAILED DESCRIPTION

(5) The technical solutions of the present invention are further illustrated with reference to the drawings and the specific embodiments, it should be understood that these embodiments are only used for illustrating the present invention and are not intended to limit the scope of the present invention, and after reading the present invention, modifications of various equivalent forms of the present invention made by those skilled in the art all fall within the scope defined by the appended claims of the present application.

(6) As shown in FIG. 1, an extended-range fuel cell electric vehicle power device according to the present invention includes a driving motor 1, a bidirectional converter 2, a chopper 3, a power cell 4, a fuel cell 5, a high-pressure hydrogen storage tank 6, an electric control valve 7, a controller 8, an accelerator pedal 9 and a brake pedal 10.

(7) An output of the driving motor is mechanical energy and is connected to a transmission shaft of an electric vehicle through a speed change gearbox, and an input of the driving motor is connected to an alternating current output end of the bidirectional converter. A direct current input end of the bidirectional converter is connected in parallel to an output of the power cell and an output of the chopper, an input of the chopper is connected to a power source output of the fuel cell, and a fuel inlet of the fuel cell is connected to an output of the high-pressure hydrogen storage tank, with the electric control valve connected therebetween. A control port of the electric control valve is connected to an output of the controller, and the output of the controller is also connected to a control port of the bidirectional converter to control a working state of the bidirectional converter. An input end of the controller is connected to output signals of the accelerator pedal and the brake pedal. The high-pressure hydrogen storage tank is provided with a fuel injection interface, and the power cell is provided with a charging interface.

(8) The driving motor converts electric energy into mechanical energy to drive a vehicle.

(9) Forward operation of the bidirectional converter inverts direct current energy of the power cell into direct current energy to provide the alternating current energy to the driving motor, and backward operation of the bidirectional converter charges the power cell with electric energy generated during braking of the driving motor to realize regenerative braking.

(10) The chopper changes direct current energy of the fuel cell into a voltage range acceptable at a direct current input end of the bidirectional converter.

(11) The power cell is used for storing electric power and may be a lithium ion cell and the like.

(12) The fuel cell converts chemical energy into electric energy, and is equivalent to a small power generation device.

(13) The high-pressure hydrogen storage tank is used for storing hydrogen.

(14) The electric control valve is used for controlling a supply amount of the hydrogen.

(15) The controller is composed of a high-performance chip and a peripheral circuit, collects the signals of the accelerator pedal and the brake pedal, and controls the working states of the electric control valve and the bidirectional converter. The controller may be TMS320F28335PGFA.

(16) The accelerator pedal is used for controlling a vehicle speed, and the signal is transmitted to the input end of the controller.

(17) The accelerator pedal is used for controlling a braking process of a vehicle, and the signal is transmitted to the input end of the controller.

(18) A method for controlling an extended-range fuel cell electric vehicle power device according to the present invention is shown in FIG. 2, wherein a control scheme thereof is that: when a given power Pin is positive, an electric vehicle works in a driving state, when the given power Pin is negative, the electric vehicle works in a braking state, an actual output power Pout of a driving motor drives the electric vehicle, and a signal of the actual output power Pout of the driving motor is fed back to an input end and compared with the given power Pin to form a feedback control system. The signal of the Pout and a signal of the Pin are inputted into a comparator to obtain a power error signal ΔP, and differentiation processing is performed on the power error signal ΔP to obtain a power change rate signal dP/dt, and two controlled quantities y.sub.1 and y.sub.2 are obtained from the ΔP and the dP/dt by a control algorithm, wherein the controlled quantity y.sub.1 controls an opening degree of an electric control valve, thus controlling an output electric power of a fuel cell, and the controlled quantity y.sub.2 controls a PWM (pulse width modulation) signal of a control port of a bidirectional converter, thus controlling a working mode and a power of the bidirectional converter. An energy control flow chart of an extended-range high-temperature fuel cell electric vehicle is shown in FIG. 3.

(19) The controlling the working mode of the bidirectional converter includes controlling an inversion state when the electric vehicle works in the driving state and a rectification state when the electric vehicle works in the braking state. In the inversion state, the driving motor works in an electric state, and a power cell converts a direct current into an alternating current through the bidirectional converter to provide electric power for the driving motor. In the rectification state, the driving motor works in a power generation state and converts energy generated during braking of the electric vehicle into electric energy, and an alternating current generated by the driving motor is rectified into a direct current through the bidirectional converter to charge the power cell. When a signal is generated by an accelerator pedal, the driving motor works in the power generation state, and when a signal is generated by a brake pedal, the driving motor works in the electric state.

(20) A BP neural network is used for control in the control algorithm above, and a learning ability of the BP neural network is used to learn a relationship between the driving motor and actions of the power cell, the fuel cell and a driver, so that corresponding control signals can be made quickly and accurately to control the driving motor.

(21) As shown in FIG. 4, a structure of the BP neural network is a 2-5-2 structure, which means that, a number of neurons in an input layer is that i=1,2, x.sub.i corresponds to two input variables, and the input layer is used for signal transmission; the first input variable is: x.sub.1:ΔP; the second input variable is: x.sub.2:dP/dt; a number of neurons in a middle layer is that j=1,2,3,4,5; and an input of the neurons in the middle layer is x.sub.j and an output of the neurons in the middle layer is x′.sub.j; and a number of neurons in an output layer is that l=1,2, an input is y.sub.l, and y.sub.l corresponds to two controlled quantities y.sub.1 and y.sub.2.

(22) After the structure of the BP neural network is established, the BP network is trained next, and a specific training process is as follows.

(23) Forward propagation refers to transmitting an input signal from the input layer to the middle layer and then to the output layer, and if the output layer obtains the expected output, the learning algorithm is finished; otherwise, the signal turns to backward propagation.

(24) A specific network learning algorithm includes: (1) forward propagation: calculating an output of a network, wherein the input of the neurons in the middle layer is a weighted sum of all the inputs, namely:

(25) $\begin{array}{cc}{x}_j=\underset{i=1}{\overset{2}{.Math.}}{\omega }_{ij}{x}_i;& \left(1\right)\end{array}$ the output x′.sub.j of the neurons in the middle layer excites x.sub.j by using an S function to obtain:

(26) $\begin{array}{cc}{x}_j^{\prime }=f\left(x_j\right)=\frac{1}{1+e^{-x}};\mathrm{then}\text{:}& \left(2\right)\\ \frac{\partial x_j^{\prime }}{\partial x_j}=x_j^{\prime }\left(1-x_j^{\prime }\right);& \left(3\right)\end{array}$ the output of the neurons in the output layer is that:

(27) $\begin{array}{cc}{y}_l=\underset{j}{\overset{5}{.Math.}}{\omega }_{jl}{x}_j^{\prime };& \left(4\right)\end{array}$ an error between a l.sup.th output of the BP neural network and a corresponding ideal output y.sub.l.sup.0 is that:
e.sub.l=y.sub.l.sup.0−y.sub.l  (5); taking a p.sup.th sample as an example, an error performance index function of the p.sup.th sample is that:

(28) 0$\begin{array}{cc}{E}_p=\frac{1}{2}\underset{l=1}{\overset{5}{.Math.}}{e}_l^2;& \left(6\right)\end{array}$ (2) backward propagation: adjusting weights among the layers by gradient descent, wherein a learning algorithm of a connection weight ω.sub.jl between the output layer and the middle layer is that:

(29) $\begin{array}{cc}{\mathrm{\Delta \omega }}_{jl}=-\eta \frac{\partial E_p}{\partial {\omega }_{jl}}=\eta e_l\frac{\partial x_l}{\partial {\omega }_{jl}}=\eta e_l{x}_j^{\prime };& \left(7\right)\end{array}$ wherein η is a learning rate, η∈[0,1]; a network weight a ta moment k+1 is that:
ω.sub.jl(k+1)=ω.sub.jl(k)+Δω.sub.jl  (8); a learning algorithm of a connection weight ω.sub.ij between the middle layer and the input layer is that:

(30) $\begin{array}{cc}{\mathrm{\Delta \omega }}_{ij}=-\eta \frac{\partial E_p}{\partial {\omega }_{ij}}=\eta \underset{l=1}{\overset{5}{.Math.}}{e}_l\frac{\partial x_l}{\partial {\omega }_{ij}};\mathrm{wherein}\frac{\partial x_l}{\partial {\omega }_{ij}}=\frac{\partial x_l}{\partial x_j^{\prime }}.Math.\frac{\partial x_j^{\prime }}{\partial x_j}.Math.\frac{\partial x_j}{\partial {\omega }_{ij}}={\omega }_{il}.Math.\frac{\partial x_j^{\prime }}{\partial x_j}.Math.x_i={\omega }_{il}.Math.x_j^{\prime }\left(1-x_j^{\prime }\right).Math.x_i;& \left(9\right)\end{array}$
and a network weight at a moment t+1 is that:
ω.sub.ij(k+1)=ω.sub.ij(k)+Δω.sub.ij  (10).

(31) Finishing controller structure design and learning algorithm design based on the steps (1) and (2) above can realize control by the BP neural network, and required samples can be obtained from experiments. A large amount of sample data recorded in test of a vehicle can be used as training samples. The neural network controller cannot be directly used after designing, and the sample data is needed to train and learn use of fuel materials. The designed control method needs samples for training, and the sample data comes from samples obtained during the actual test of the vehicle.