KR100191917B1 - The detecting method of battery charging amount for electric-automobile using neural-network - Google Patents
The detecting method of battery charging amount for electric-automobile using neural-network Download PDFInfo
- Publication number
- KR100191917B1 KR100191917B1 KR1019960009768A KR19960009768A KR100191917B1 KR 100191917 B1 KR100191917 B1 KR 100191917B1 KR 1019960009768 A KR1019960009768 A KR 1019960009768A KR 19960009768 A KR19960009768 A KR 19960009768A KR 100191917 B1 KR100191917 B1 KR 100191917B1
- Authority
- KR
- South Korea
- Prior art keywords
- battery
- charge
- charge amount
- output
- temperature
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Tests Of Electric Status Of Batteries (AREA)
- Secondary Cells (AREA)
Abstract
본 발명은 뉴랄 네트워크를 이용한 전기자동차의 배터리 충전량 측정방법에 관한 것으로 기존 알고리즘이 방전 전류에 대해 충전량을 계산하는 것과는 달리 파워에 대해서 충전량을 계산하고 Nonlinear 한 요소를 뉴랄 네트워크를 통한 학습으로 구현하여 배터리에서 출력되는 파워의 다이나믹한 변화와 온도변화에도 정확한 충전량을 구해 줄 수 있어 운전자가 언제 배터리가 방전되어 정지할 것인가를 걱정할 필요가 없다.The present invention relates to a method of measuring a battery charge of an electric vehicle using a neural network. Unlike the conventional algorithm that calculates a charge amount for a discharge current, the present invention calculates a charge amount for a power and implements a nonlinear element by learning through a neural network. It can calculate the exact charge amount even if the power output is dynamic and the temperature change, so the driver does not have to worry about when the battery will discharge and stop.
Description
본 발명은 전기자동차의 배터리 충전량 측정방법에 관한 것으로, 특히 뉴랄네트워크(Neural Network)를 이용한 전기자동차의 배터리 충전량 측정방법에 관한 것이다.The present invention relates to a method for measuring battery charge of an electric vehicle, and more particularly, to a method for measuring battery charge of an electric vehicle using a neural network.
전기자동차에서 에너지 소스인 배터리에 얼마만큼의 에너지가 남아 있는지를 충전량(State of Charge)으로 나타낸다. 그러나 배터리의 불규칙성(온도, 방전에 따라 변함) 때문에 기존 차량의 연료게이지처럼 정확히 충전량을 측정하는 것은 어렵다.The amount of energy remaining in the battery as an energy source in an electric vehicle is expressed as a state of charge. However, because of battery irregularities (varies with temperature and discharge), it is difficult to accurately measure the charge amount as in the fuel gauge of a conventional vehicle.
배터리의 충전량을 측정하는 여러 가지 방법이 다음과 같이 있으나 정확히 충전량을 측정하는 알고리즘은 없다: 전압 이용방법은 배터리의 전압이 충전량과 관련이 있다는 것을 이용하여 전압을 측정하여 충전량을 측정하는 방법으로 차량에서처럼 다이니믹한 부하에 대해서는 전압과 충전량의 관계가 불명확해지기 때문에 정확도가 매우 떨어진다. Amp-hour 방법은 충전할 때 배터리에 들어간 Ah에 대해서 실제로 사용한 Ah를 적산함으로써 얻은 값의 비를 충전량으로 하는 방법으로 실제 사용할 수 있는 Ah가 온도, 방전상태에 따라 변하기 때문에, 즉 reference 값이 변하기 때문에 부정확하다.There are several ways to measure the charge level of a battery, but there is no algorithm to accurately measure the charge level. The voltage usage method is a method of measuring a charge amount by measuring a voltage using a voltage related to a charge amount. For dynamic loads, the accuracy is very low because the relationship between voltage and charge is unclear. The Amp-hour method uses the ratio of the value obtained by integrating the Ah actually used for the Ah entered into the battery when charging. Since the actually available Ah varies depending on the temperature and the discharge state, that is, the reference value changes. Because it is incorrect.
본 발명의 목적은 상기와 같은 문제점을 해결하기 위하여 배터리가 실제 사용할 수 있는 에너지량을 온도와 방전이력에 대한 함수로서 모델링하여 다이나믹한 부하에 대해 정확한 충전량을 산출할 수 있는 뉴랄 네트워크를 이용한 전기자동차의 배터리 충전량 측정방법을 제공하고자 하는 것이다.An object of the present invention is to solve the above problems by modeling the amount of energy that the battery can actually use as a function of temperature and discharge history electric car using a neural network that can calculate the exact charge amount for the dynamic load To provide a method for measuring battery charge.
제1도는 본 발명의 뉴랄 네트워크를 이용한 배터리 충전량 측정방법을 도시하는 도면.1 is a diagram illustrating a method for measuring battery charge using a neural network of the present invention.
* 도면의 주요부분에 대한 부호의 설명* Explanation of symbols for main parts of the drawings
1 : 배터리 2 : 인버터1: battery 2: inverter
3 : 모터3: motor
본 발명을 첨부된 도면에 도시된 실시예를 참조하여 하기에 설명한다.The invention is described below with reference to the embodiments shown in the accompanying drawings.
제1도에는 본 발명의 뉴랄 네트워크를 이용한 배터리 충전량 측정방법을 실현하는 구성이 도시된다. 배터리(1)는 인버터(2)를 통하여 모터(3)에 에너지를 공급하며 배터리(1)에는 온도센서(4)가 부착되어 본 발명의 충전량 알고리즘에 온도신호를 보내준다. 본 발명의 충전량 알고리즘은 배터리(1)의 전압과 전류로부터 파워를 검출하여 사용한다.1 shows a configuration for realizing a battery charge measuring method using a neural network of the present invention. The battery 1 supplies energy to the motor 3 through the inverter 2 and the temperature sensor 4 is attached to the battery 1 to send a temperature signal to the charge amount algorithm of the present invention. The charge amount algorithm of the present invention detects and uses power from the voltage and current of the battery 1.
본 발명의 충전량 알고리즘은 2 개의 뉴칼 네트워크로 구성되는데, NN1과 NN2 이다. NN1은 배터리의 온도와 출력 파워를 입력으로 하고 출력은 배터리의 가능한 총 에너지양을 낸다. NN2는 NN1의 입력인 온도와 출력과 더불어 NN1의 출력인 가능 에너지 그리고 현재의 충전량을 입력으로 한다. 그리고 출력은 DOD(dcpth of discharge)의 변화량을 낸다. 따fk서 이 출력을 적분하면 현재의 DOD가 되고 1에서 빼면 충전량을 얻을 수 있다. 이렇게 얻은 충전량은 앞에서 설명하였듯이 NN2의 입력으로도 들어간다.The charge amount algorithm of the present invention consists of two nucal networks, NN1 and NN2. The NN1 takes the temperature and output power of the battery as inputs, and the output gives the total amount of energy available in the battery. NN2 takes the input of temperature and output which are the input of NN1, the available energy which is the output of NN1, and the current charge amount. And the output gives the amount of change in the dc of discharge (DOD). Thus, integrating this output gives us the current DOD, and subtracts from 1 to get the charge. The charge thus obtained enters the input of NN2 as described earlier.
뉴랄 네트워크는 BPN(Back Propagation Network) 구조의 학습방법을 택한다.The neural network takes the learning method of BPN (Back Propagation Network) structure.
즉, NN1 은 In other words, NN1 is
여기서 E*는 다음과 같이 구해진다.Where E * is obtained as
여기서 함수 f 는 signoid function 으로서 다음과 같다.Where function f is a signoid function:
이 뉴랄 네트워크는 잘 알려진 구조이며, 학습을 통해서 φ와δ, 그리고 ψ를 튜닝하여 원하는 입출력 함수를 얻게 되는 것이다.This neural network is a well known structure, and by learning, we tune φ, δ, and ψ to obtain the desired input and output functions.
학습방법은 NN1은 gnn1(P,T)=E 로서 배터리의 온도와 파워의 입력으로 들어 때 배터리의 가능 에너지를 E*1, P1, T1), (E2, P2, T2), ....,(E*n, Pn, Tn) 데이터 쌍을 얻은 다음 gnn1이 이 데이쌍의 관계를 갖도록 한다. 이 데이터 쌍의 합습 데이터는 다음과 같이 실험을 하여 얻는다.Learning method NN1 is gnn1 (P, T) = E a battery of the energy input into the time of the temperature of the battery and the power E * 1, P1, T1) , (E2, P2, T2 a), .... Obtain a pair of (E * n, Pn, Tn) data and let gnn1 have this data pair relationship. The pooled data of this data pair is obtained by experiment as follows.
완전 방전 때까지의 총 방전 에너지를 구한다. 즉,Find the total discharge energy until complete discharge. In other words,
여기서 timek은 방전시간이다.Where time k is the discharge time.
이와 같은 여러조건 Pk,Tk하에서 실험하여 E* k를 얻은 다음 다음과 같은 에러 함수를 정의하여 이 에러 함수가 최소가 되도록 NN1을 학습시킨다. 여기서 학습시킨다는 말을 다르게 표현하면 NN1 내의 파라미터 즉, φ, δ, ψ를 튜닝한다는 것이다.Experiment under these various conditions P k , T k to get E * k , and then learn NN1 to minimize this error function by defining the following error function. In other words, the term "learning" is used to tune the parameters in NN1, that is, φ, δ, and ψ.
φ 의 튜닝은 다음과 같이 한다. 다른 파라미터도 같은 방법으로 한다.Tuning of φ is done as follows. Do the same for other parameters.
여기서 η ψ 는 학습률이다.Where η ψ is the learning rate.
이와 같이 여러 학습 데이터 쌍에 대해서 반복하여 학습하되 에러가 설정된 리미트 값 보다 작을 때까지 수행한다.As described above, the training is repeatedly performed on the training data pairs until the error is smaller than the set limit value.
마찬가지로 δ,ψ 에 대해서도 구할 수 있다.Similarly, it can obtain | require about δ, ψ.
NN2 는 NN1 과 같으며 단지 입력수가 4 개라는 점만 다르다.NN2 is the same as NN1 except that there are four inputs.
본 발명의 뉴랄 네트워크를 이용한 전기자동차의 배터리 충전량 측정방법은 기존 알고리즘이 방전 전류에 대해 충전량을 계산하는 것과는 달리 파워에 대해서 충전량을 계산하고 Nonlinear 한 요소를 뉴랄 네트워크를 통한 학습으로 구현하여 배터리에서 출력되는 파워의 다이나믹한 변화와 온도변화에도 정확한 충전량을 구해줄 수 있어 운전자가 언제 배터리가 방전되어 정지할 것인가를 걱정할 필요가 없다.Unlike the conventional algorithm which calculates the charge amount for the discharge current, the method of measuring the battery charge amount of the electric vehicle using the neural network of the present invention calculates the charge amount for power and outputs the nonlinear element by learning through the neural network to output from the battery. Accurate charge capacity can be obtained even with dynamic power changes and temperature changes, eliminating the need for drivers to worry about when the battery will run out.
또한 본 발명의 충전량 측정방법은 기존방법이 linear 모델이기 때문에 풀 수 없었던 배터리 충전량 모델링이 가능하고 학습기능으로 스스로 그것을 구할 수 있다.In addition, the method of measuring the amount of charge of the present invention is possible to model the battery charge amount that could not be solved because the conventional method is a linear model and can obtain it by themselves as a learning function.
Claims (1)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1019960009768A KR100191917B1 (en) | 1996-04-01 | 1996-04-01 | The detecting method of battery charging amount for electric-automobile using neural-network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1019960009768A KR100191917B1 (en) | 1996-04-01 | 1996-04-01 | The detecting method of battery charging amount for electric-automobile using neural-network |
Publications (2)
Publication Number | Publication Date |
---|---|
KR970071027A KR970071027A (en) | 1997-11-07 |
KR100191917B1 true KR100191917B1 (en) | 1999-06-15 |
Family
ID=19454806
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
KR1019960009768A KR100191917B1 (en) | 1996-04-01 | 1996-04-01 | The detecting method of battery charging amount for electric-automobile using neural-network |
Country Status (1)
Country | Link |
---|---|
KR (1) | KR100191917B1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7932699B2 (en) | 2006-12-11 | 2011-04-26 | Hyundai Motor Company | Method of controlling battery charge level of hybrid electric vehicle |
KR102065120B1 (en) * | 2018-09-27 | 2020-02-11 | 경북대학교 산학협력단 | Battery charging state estimation method based on neural network |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20030092808A (en) * | 2002-05-31 | 2003-12-06 | 현대자동차주식회사 | a method for calculation a battery state of charge in electric vehicle |
CN109307852A (en) * | 2018-09-06 | 2019-02-05 | 中国电力科学研究院有限公司 | A kind of method and system of the measurement error of determining electric automobile charging pile electric energy metering device |
-
1996
- 1996-04-01 KR KR1019960009768A patent/KR100191917B1/en not_active IP Right Cessation
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7932699B2 (en) | 2006-12-11 | 2011-04-26 | Hyundai Motor Company | Method of controlling battery charge level of hybrid electric vehicle |
KR102065120B1 (en) * | 2018-09-27 | 2020-02-11 | 경북대학교 산학협력단 | Battery charging state estimation method based on neural network |
Also Published As
Publication number | Publication date |
---|---|
KR970071027A (en) | 1997-11-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6362598B2 (en) | Method for determining the state of charge and loading capacity of an electrical storage battery | |
KR100606878B1 (en) | Method for estimating polarization voltage of secondary cell, method and device for estimating remaining capacity of secondary cell, battery pack system, and electric vehicle | |
CN103907030B (en) | The estimation unit of battery charge rate and the method for estimation of battery charge rate | |
US6388450B2 (en) | Method for determining the state of charge of storage batteries | |
US8519675B2 (en) | System, method, and article of manufacture for determining an estimated battery cell module state | |
US6388447B1 (en) | Method and apparatus for battery fuel gauging | |
EP3207388B1 (en) | Battery condition monitoring | |
CN101395489B (en) | System and method for determining both an estimated battery state vector and an estimated battery parameter vector | |
RU2336618C2 (en) | Method of battery power-carrying capacity calculation using improved prediction methods based on element model | |
KR100425352B1 (en) | Apparatus for estimating charged state of battery and method for estimating degraded state of battery | |
CN101142732B (en) | Method for adjusting support vector machine, apparatus and method for estimating battery state of charge | |
US10124696B2 (en) | Estimation and compensation of battery measurement and asynchronization biases | |
US10048319B2 (en) | Estimating of the state of charge of a battery | |
KR100901252B1 (en) | Method and Apparatus for estimation of State Of Charge using sliding mode observer | |
EP2430697B1 (en) | Method for determining extent and type of capacity fade | |
CN103439668A (en) | Charge state evaluation method and system of power lithium ion battery | |
KR20040060998A (en) | Method and device for estimating remaining capacity of secondary cell, battery pack system, and electric vehicle | |
CN102282477A (en) | Cell abnormality detection circuit and power supply device | |
CN111707955B (en) | Method, apparatus and medium for estimating remaining life of battery | |
CN105223512A (en) | Based on the method for the dynamic rectification dump energy of battery behavior | |
CN116113837A (en) | Method for estimating state of charge of battery | |
CN112630661B (en) | Battery state of charge (SOC) estimation method and device | |
KR100878123B1 (en) | Method and system for battery state and parameter estimation | |
KR100191917B1 (en) | The detecting method of battery charging amount for electric-automobile using neural-network | |
Baba et al. | State of charge estimation of lithium-ion battery using Kalman filters |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
A201 | Request for examination | ||
E701 | Decision to grant or registration of patent right | ||
GRNT | Written decision to grant | ||
LAPS | Lapse due to unpaid annual fee |