JP2007026683A - Fuel cell system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To supply hydrogen from a plurality of hydrogen storage alloy tanks to a fuel cell under high-level control. <P>SOLUTION: In this fuel cell system, a temperature and a hydrogen pressure inside hydrogen storage alloy tanks MH1, ..., and MHn are monitored, and based on data acquired from monitoring, supply valves V1, ..., and Vn and heating valves H1, ..., and Hn are controlled. Control of the supply valves and the heating valves is carried out by using control data stored in a controller 4. An optimum state transition path is found by reinforcement learning on simulation, which models the fuel cell system 100, and the control data are generated, based on a state and action stored in the state transition path. In this way, hydrogen can be fed from a plurality of hydrogen storage alloy tanks to the fuel cell along the optimum state transition path. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a fuel cell system, and more particularly to a fuel cell system capable of supplying hydrogen from a plurality of hydrogen storage alloy tanks to a fuel cell by advanced control.

Conventionally, a fuel cell system that includes a plurality of hydrogen storage alloy tanks, supplies hydrogen from some tanks to the fuel cell, and switches the tank so that hydrogen is supplied from the remaining tanks to the fuel cell when the supply pressure drops. Is known (see, for example, Patent Document 1).
On the other hand, a plant simulation method using a reinforcement learning method that performs a simulation of a plant that exhibits complex behavior by applying the reinforcement learning method is known (see, for example, Patent Document 2).
JP 2001-295996 A JP 2004-178492 A

In the above-described conventional fuel cell system, the tank is switched only according to the supply pressure, but this may not allow optimal operation. For example, even if a tank to which hydrogen is to be supplied next fails, there is a problem in that the failure is not known until the supply from the tank is started, and heating of the tank is continued wastefully.
Accordingly, an object of the present invention is to provide a fuel cell system capable of supplying hydrogen from a plurality of hydrogen storage alloy tanks to a fuel cell by advanced control.

In a first aspect, the present invention performs a plurality of hydrogen storage alloy tanks that store hydrogen in a hydrogen storage alloy, a fuel cell, and supply and stop of hydrogen from the hydrogen storage alloy tank to the fuel cell. Supply / stop means for heating, heating means for heating each hydrogen storage alloy tank, temperature sensor for detecting the temperature in each hydrogen storage alloy tank, and hydrogen pressure in each hydrogen storage alloy tank A pressure sensor to detect, a control means for monitoring the supply / stop means, the heating means, the temperature sensor, and the pressure sensor, and instructing the supply / stop means and the heating means according to their states. A fuel cell system is provided.
In the fuel cell system according to the first aspect, the temperature and hydrogen pressure in each hydrogen storage alloy tank are monitored, and the supply / stop means and heating means are controlled based on these data. For this reason, hydrogen can be supplied from a plurality of hydrogen storage alloy tanks to the fuel cell by advanced control. For example, if a tank to which hydrogen is to be supplied next fails, the hydrogen pressure does not rise rapidly even if heated, so that the failure is known and heating of the tank can be stopped before starting supply from the tank.

In a second aspect, the present invention provides the fuel cell system according to the first aspect, wherein the supply / stop unit, the heating unit, the temperature sensor, and the pressure sensor according to the states of the supply / stop unit and the pressure sensor. And storage means for storing control data in which actions to be instructed next are respectively set to the heating means, and the control data is stored in the supply / stop means and the heating means on a simulation modeling the fuel cell system. The supply / stop means recorded in the state transition path is found by reinforcement learning that repeatedly evaluates the state of the fuel cell, the temperature sensor, and the pressure sensor. Based on the state of the heating means, the temperature sensor, and the pressure sensor, and the instructions given to the supply / stop means and the heating means. The control means monitors the supply / stop means, the heating means, the temperature sensor, and the pressure sensor, and follows the supply / stop means and the heating means according to their states. The fuel cell system is characterized in that an action to be instructed is read from the control data and instructed to the supply / stop unit and the heating unit.
The behavior of the fuel cell system is complex, and it is difficult for humans to assume the optimum behavior according to the state. Furthermore, if the number of hydrogen storage alloy tanks is 100, for example, the number of temperature sensors is 100 and the number of pressure sensors is 100. If the supply / stop means is a supply valve provided in each tank, the number of supply valves is also 100. Furthermore, if the heating means is a heating valve provided in each tank, the number of heating valves is also 100. That is, the number of states and the number of actions become enormous, and it is quite difficult for humans to assume the optimum action according to the state.
Therefore, in the fuel cell system according to the second aspect, the state transition path is found by reinforcement learning on a simulation modeling the fuel cell system, and control data is created based on the state and action recorded in the state transition path. Then, control is performed using the control data. Thereby, hydrogen can be supplied from a plurality of hydrogen storage alloy tanks to the fuel cell along an optimal state transition path according to the state.

  According to the fuel cell system of the present invention, hydrogen can be supplied from a plurality of hydrogen storage alloy tanks to the fuel cell by advanced control. Further, hydrogen can be supplied from the plurality of hydrogen storage alloy tanks to the fuel cell along an optimum state transition path according to the state.

  Hereinafter, the present invention will be described in more detail with reference to the embodiments shown in the drawings. Note that the present invention is not limited thereby.

FIG. 1 is an explanatory diagram of a fuel cell system 100 according to the first embodiment.
This fuel cell system 100 includes n (≧ 2) hydrogen storage alloy tanks MH1,..., MHn, a fuel cell 1, and hydrogen storage alloy tanks MH1 to MHn to the fuel cell 1 for storing hydrogen in the hydrogen storage alloy. , Vn interposed between the hydrogen storage alloy tanks MH1,..., MHn and the hydrogen pipe 2, and the exhaust of the fuel cell 1. Heat exchanger 3 for heating the hydrogen storage alloy tanks MH1,..., MHn using heat, and heating valves H1, interposed between the hydrogen storage alloy tanks MH1,. , Hn and temperature sensors T1,..., Tn for detecting temperatures in the hydrogen storage alloy tanks MH1,..., MHn, and pressure sensors P1,. , Pn and supply valve V1 ..., Vn and heating valves H1, ..., Hn, temperature sensors T1, ..., Tn and pressure sensors P1, ..., Pn are monitored, and supply valves V1, ..., Vn and heating valve H1 according to their current state. ,..., Hn and a controller 4 for instructing the next action.

  The heat exchanger 3 receives the exhaust heat of the fuel cell 1 by a heat exchange medium (such as water). The heating valves H1,..., Hn circulate the heat exchange medium between the hydrogen storage alloy tanks MH1,.

FIG. 2 is a conceptual diagram showing the control data 5 stored in the controller 4.
The control data 5 includes supply valves V1,..., Vn, heating valves H1,..., Hn and pressure sensors P1,. List of “current state” consisting of Pn ′ and temperature change rates T1 ′,..., Tn ′ and “next action” of supply valves V1,..., Vn and heating valves H1,. It consists of multiple sets. Each set is prioritized.
It should be noted that (pressure change rate Px ′) = (current pressure Px) − (pressure Px before one control time), (temperature change rate Tx ′) = (current temperature Tx) − (one control time before) Temperature Tx).

FIG. 3 is a flowchart showing the procedure of the hydrogen supply control process executed by the controller 4.
In step S1, the supply valves V1,..., Vn and the heating valves H1,. That is, the supply of hydrogen to the fuel cell 1 is stopped, and the heating of the hydrogen storage alloy tanks MH1, ..., MHn is stopped.

  In step S2, the abnormality detection timer built in the controller 4 is started from “0”.

In step S3, the current values of the pressure sensors P1,..., Pn and the temperature sensors T1,..., Tn are read and the pressure change speeds P1 ′, ..., Pn ′ and the temperature change speeds T1 ′,. To do.
In step S4, supply valves V1, ..., Vn and heating valves H1, ..., Hn and pressure sensors P1, ..., Pn and temperature sensors T1, ..., Tn current values and pressure change rates P1 ', ..., Pn' And whether the “current state” consisting of the temperature change rates T1 ′,..., Tn ′ exists in the control data 5.
In step S5, if the “current state” exists in the control data 5, the process proceeds to step S6, and if not, the process proceeds to step S7.
In step S6, the “next action” of the supply valves V1,..., Vn and the heating valves H1,..., Hn corresponding to the “current state” is read from the control data 5, and the supply valves V1,. A control command is sent to H1,..., Hn. Then, the process returns to step S2.

In step S7, it is determined whether the abnormality detection timer has timed out. If not, the process proceeds to step S8, and if timed out, the process proceeds to step S9.
In step S8, only the heating valve Hi of the hydrogen storage alloy tank MHi having the highest hydrogen pressure and the smallest number i is opened. Then, the process returns to step S3.

  In step S9, the supply valves V1,..., Vn and the heating valves H1,. That is, since there is an abnormal situation in which the state not described in the control data 5 continues, the supply of hydrogen to the fuel cell 1 is stopped, and the heating of the hydrogen storage alloy tanks MH1, ..., MHn is stopped. Then, the process ends.

FIG. 4 shows a control data creation device that finds an optimal state transition path through reinforcement learning on a simulation modeling the fuel cell system 100 and creates control data 5 based on the state and behavior recorded in the state transition path. FIG.
The control data creation device 200 includes a fuel cell simulator 21 that models the fuel cell 1, a hydrogen storage alloy tank MH1,..., MHn, temperature sensors T1,..., Tn, pressure sensors P1,. .., Vn and heating valves H1,..., Hn, a hydrogen tank simulator 22, a heat exchanger / simulator 23 that models the heat exchanger 3, and a load / simulator 24 that models a load. And a state transition / evaluation unit 25 that gives behavior data to each simulator and evaluates and records the state transition caused thereby.
The control data creation device 200 is a computer in terms of hardware.

FIG. 5 is a flowchart showing the procedure of reinforcement learning processing by the control data creation device 200.
In step P1, the state of each simulator and valve is initialized.
In step P2, the operation of each simulator is started.

In step P3, pressure data and the like are collected and the rate of change is calculated to obtain the “current state”.
In step P4, if the “current state” is the same as before, the process returns to step P3, and if the “current state” is different from the previous, the process proceeds to step P5.

In step P5, action data is randomly generated and given to each simulator.
In step P6, as a result of giving behavior data to each simulator, if the fuel cell does not stop, the process proceeds to step P7, and if the fuel cell stops, the process proceeds to step P8.

  In Step P7, “0” is given to the reward, and the behavior evaluation function is updated. Then, the process returns to Step P3.

  In Step P8, a negative value is given to the reward, and the behavior evaluation function is updated. Then, the process returns to Step P1.

  The state transition / evaluation unit 25 creates a state transition diagram by performing reinforcement learning processing over a sufficient time. Next, the control data 5 is created by associating each state in the state transition diagram with the next action in which the time until the stop of the fuel cell is the longest. When there are a plurality of next actions that take the longest time to stop the fuel cell in a certain state, one of those actions is selected. For example, the action of opening the heating valve Hi of the hydrogen storage alloy tank MHi having the smallest number i among the plurality of next actions is selected. The control data 5 may be stored in the controller 4.

6 to 8 illustrate state transition diagrams.
This state transition diagram is created under the following conditions for the sake of simplicity.
(1) There are only three hydrogen storage alloy tanks MH1, MH2, and MH3.
(2) The state transition is triggered by whether the hydrogen storage alloy tanks MH1, MH2, and MH3 are emptied or become abnormal during use. The time to empty is longer than the time to become abnormal during use.
(3) Action opens the valve corresponding to any one hydrogen storage alloy tank, and closes the valve corresponding to the remaining hydrogen storage alloy tanks.
(4) Square boxes represent states Sa to SE.
(5) The number written in the square box represents the number of the hydrogen storage alloy tank to be opened in the next action.
(6) An arrow coming out of the square box represents a state transition.
(7) The length of the arrow represents the length of usable time.

FIG. 9 illustrates control data 5 created from the state transition diagrams of FIGS.
According to the control data 5, the following control is performed.
(1) In the initial state Sa, only the valve corresponding to the hydrogen storage alloy tank MH1 is opened.
(1-1) If normal, the hydrogen storage alloy tank MH1 is emptied and enters the state Sb, so that only the valve corresponding to the hydrogen storage alloy tank MH2 is opened.
(1-2) If the hydrogen storage alloy tank MH1 is abnormal, the state Sg is reached, so that only the valve corresponding to the hydrogen storage alloy tank MH2 is opened.
(2) In the state Sb, only the valve corresponding to the hydrogen storage alloy tank MH2 is opened.
(2-1) If normal, the hydrogen storage alloy tank MH2 becomes empty and enters the state Sc, so that only the valve corresponding to the hydrogen storage alloy tank MH3 is opened.
(2-2) If the hydrogen storage alloy tank MH2 is abnormal, the state Sd is entered. Therefore, only the valve corresponding to the hydrogen storage alloy tank MH3 is opened.
The same applies hereinafter.

  The fuel cell system of the present invention can be used as a fuel cell system mounted on an automobile or a ship.

1 is a configuration explanatory view showing a fuel cell system according to Example 1. FIG. FIG. 3 is a conceptual diagram illustrating control data according to the first embodiment. FIG. 3 is a flowchart showing a procedure of hydrogen supply control processing according to the first embodiment. It is structure explanatory drawing which shows a control data production apparatus. It is a flowchart which shows the procedure of a reinforcement learning process. It is an illustration figure of a state transition diagram. FIG. 7 is a state transition diagram continued from FIG. 6. FIG. 8 is a state transition diagram continued from FIG. 7. It is an illustration figure of the control data created from the state transition diagram of FIGS.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Fuel cell 2 Hydrogen piping 3 Heat exchanger 4 Controller 100 Fuel cell system H1, ..., Hn Heating valve MH1, ..., MHn Hydrogen storage alloy tank P1, ..., Pn Pressure sensor T1, ..., Tn Temperature sensor V1, ..., Vn supply valve

Claims (2)

A plurality of hydrogen storage alloy tanks for storing hydrogen in the hydrogen storage alloy, a fuel cell, and supply / stop means for supplying and stopping hydrogen from the hydrogen storage alloy tank to the fuel cell; Heating means for heating the hydrogen storage alloy tank, a temperature sensor for detecting the temperature in each hydrogen storage alloy tank, a pressure sensor for detecting the hydrogen pressure in each hydrogen storage alloy tank, and the supply / stop A fuel cell system comprising: control means for monitoring the means, the heating means, the temperature sensor, and the pressure sensor, and instructing the supply / stop means and the heating means according to their states . 2. The fuel cell system according to claim 1, wherein the supply / stop unit, the heating unit, the temperature sensor, and the pressure sensor are to be instructed next according to states of the supply / stop unit, the heating unit, the temperature sensor, and the pressure sensor. Storage means for storing the control data set respectively,
The control data gives an instruction to the supply / stop unit and the heating unit on a simulation modeling the fuel cell system, and evaluates the state of the fuel cell, the temperature sensor, and the pressure sensor as a result. The state transition path is found by reinforce learning, and the states of the supply / stop means, the heating means, the temperature sensor, and the pressure sensor recorded in the state transition path, and then the supply / stop means and the heating are recorded. It was created based on the instructions given to the means,
The control unit monitors the supply / stop unit, the heating unit, the temperature sensor, and the pressure sensor, and performs an action to be instructed next to the supply / stop unit and the heating unit according to their states. A fuel cell system, wherein the fuel cell system reads out from the control data and instructs the supply / stop unit and the heating unit.

Images