Artificial intelligent fuel cell system

Patent metadata

Field	Value
Publication number	US-2019018067-A1
Application number	US-201716067000-A
Country	US
Kind code	A1
Filing date	Sep 14, 2017
Priority date	Sep 26, 2016
Publication date	Jan 17, 2019
Grant date	—

Abstract

An artificial intelligent fuel cell system according to an exemplary embodiment of the present invention may include: a fuel cell stack in which a plurality of unit cells is combined for generating electric energy with an electrochemical reaction; a sensor unit which measures in real time data about each of the unit cells forming the fuel cell stack, temperature, pressure, humidity, and flow rates of reaction gases, and cooling water, and current and voltage data during an operation of a fuel cell; an artificial intelligent unit which collects the data measured by the sensor unit with a predetermined time interval, generates a model for predicting and controlling performance of the fuel cell through the learning and analysis of the collected data, compares the generated model with the data measured in real time and diagnoses a state of the fuel cell stack, and generates a control signal for changing an operation condition of the fuel cell stack; and a control unit which changes the operation condition of the fuel cell stack according to the generated control signal.

First claim

1 . An artificial intelligent fuel cell system, comprising:
a fuel cell stack in which a plurality of unit cells for generating electric energy with an electrochemical reaction is combined; a sensor unit which measures in real time data about each of the unit cells forming the fuel cell stack, temperature, pressure, humidity, and flow rates of reaction gases, and cooling water, and current and voltage data during an operation of a fuel cell; an artificial intelligent unit which collects the data measured by the sensor unit with a predetermined time interval, generates a model for predicting and controlling performance of the fuel cell through the learning and analysis of the collected data, diagnoses a state of the fuel cell stack by comparing the generated model with the data measured in real time, and generates a control signal for changing an operation condition of the fuel cell stack; and a control unit which changes the operation condition of the fuel cell stack according to the generated control signal.

2 . The artificial intelligent fuel cell system of claim 1 , wherein the artificial intelligent unit includes:
a data collecting unit which collects the data about the temperature, the pressure, the humidity, the flow rates, the current, and the voltage measured by the sensor unit in real time with a predetermined time interval; a data learning and model generating unit which learns and analyzes the data collected with the predetermined time interval by using machine learning and a time-series analysis and generates a model for predicting and controlling performance of the fuel cell; and a performance predicting and diagnosing unit which compares the generated model with the measured data, distinguishes a change in performance over time of the fuel cell stack into the first state and the second state, and diagnoses the performance change state, and generates a control signal for changing an operation condition of the fuel cell stack according to the diagnosed state of the fuel cell stack and makes the control unit change the operation condition of the fuel cell stack, and the first state is a temporary and short-term performance degradation state, and the second state is a long-term and irreversible performance degradation state.

3 . The artificial intelligent fuel cell system of claim 2 , wherein the data learning and model generating unit includes:
a machine learning unit which generates the data collected with the predetermined time interval as a model for predicting performance through a machine learning algorithm, and makes the performance predicting and diagnosing unit compare a prediction value from the generated model with a measurement value of the measured data and diagnose a state of the fuel cell stack; and a time series analyzing unit which performs a time-series trend analysis analyzing a time trend pattern by using the prediction value and the measurement values, and makes the performance predicting and diagnosing unit distinguish the change in performance over time of the fuel cell stack into the first state and the second state and diagnose the performance change state, and the first state is a temporary and short-term performance degradation state, and the second state is a long-term and irreversible performance degradation state.

4 . The artificial intelligent fuel cell system of claim 3 , wherein the machine learning unit feeds a variance that is a difference between the prediction value and the measurement value back to the generated prediction model and corrects the generated prediction model, and makes the performance predicting and diagnosing unit diagnose the performance change state of the fuel cell stack according to time by using the corrected prediction model.

5 . A method of controlling an artificial intelligent fuel cell system, the method comprising:
measuring in real time data about each of the unit cells forming a fuel cell stack, temperature, pressure, humidity, and flow rates of reaction gases, and cooling water, and current and voltage data during an operation of a fuel cell; collecting the measured data with a predetermined time interval and generating a model for predicting and controlling performance of a fuel cell through the learning and analysis of the collected data; comparing the generated model with the data measured in real time and diagnosing a state of the fuel cell stack; and generating a control signal for changing an operation condition of the fuel cell stack according to the diagnosed state; and changing the operation condition of the fuel cell stack according to the generated control signal.

6 . The method of claim 5 , wherein the generating of the model for predicting and controlling the performance includes learning and analyzing the data collected with the predetermined time interval by using machine learning through a machine learning algorithm and a time-series analysis through a time-series trend analysis and generating a model for predicting and controlling performance of the fuel cell,
the diagnosing of the state of the fuel cell stack includes comparing the generated model with the measured data, distinguishing a change in performance over time of the fuel cell stack into the first state and the second state, and diagnosing the performance change state, and the first state is a temporary and short-term performance degradation state, and the second state is a long-term and irreversible performance degradation state.

7 . The method of claim 6 , further comprising:
feeding a variance that is a difference between a prediction value of the generated model and a measurement value measured in real time back to the generated prediction model and correcting the generated prediction model; and distinguishing the change in performance over time of the fuel cell stack into the first state and the second state by using the corrected prediction model, and diagnosing the performance change state, and wherein the first state is a temporary and short-term performance degradation state, and the second state is a long-term and irreversible performance degradation state.

8 . A computer-readable recording medium in which a program for implementing the method of claim 5 is recorded.

Assignees

• Lg Chemical Ltd

Inventors

• Kong Chang Sun
• Yang Jae Choon

Classifications

• H01M8/04492

  Humidity; Ambient humidity; Water content · CPC title

• H01M8/0438

  Pressure; Ambient pressure; Flow · CPC title

• H01M8/04544

  Voltage · CPC title

• B60L2260/46

  by self learning · CPC title

• H01M8/04992

  characterised by the implementation of mathematical or computational algorithms, e.g. feedback control loops, fuzzy logic, neural networks or artificial intelligence · CPC title