2010
年
3
月
第
17
卷第
2
期
控制工程
Control
Engineering
of
China
Mar.
2010
Vo
l.
17
,
No.
2
文章编号:
1671-7848
(2010)
02
-0
169
-0
6
On
Flywheel Battery Power Conversion Control
Based on
BP
Neural Network
JIANG
Yu
1
,
TANG
Shu
α
ng-qini
,
LI
Zhi-xioni
, LIU
J(,
αn3
(1.
College
of
Information
Engineering
,
China
Huangshan
University
,
Huangshan
245013
,
China;
2.
College
of
Mechanical
and
Material
Eng
,
Three
Gorges
University
,
Yichang
443002
,
China;
3.
College
of
Electrical
&
Electronics
Engineeri
吨,
Hunan
Unive
田间,
410082
,
China)
Abstract:
An adaptive PID
BP
neural network
(NN)
controller is developed for flywheel battery
pow-
er
conversion. The self-study ability of the BP
NN
and the global asymptotic stability of PID control are
integrated
, by which the high performance control is realizable in flywheel battery control through on-
line optimizing the
PID
gains
by
using the
BP
NN.
The variable learning rate strategy is applied ,
by
which the learning speed is adaptive adjusted according
to
the convergence error. The genetic algorithm
is employed
to
optimize the initial PID gains
to
enhance lhe
BP
NN
learning and training speed and
to
a-
void converging into local optima.
SVPWM
is used
to
enhance the power conversion efficiency and
巾,
crease the torque ripple. The simulation results show that
good
performances , such as faster dynamic
response
, smaller overshoot , more accurate and better robustness compared with conventional PID con-
troller
, are achieved in both charging and discharging processes
by
using the proposed controller.
Key
words:
flywheel battery; BP neural network; proportional-integral-derivative controller;
GA
CLC
mumber
, TP 27
Document
code, A
基于
BP
网络的飞轮电池电力转换控制研究
蒋宇汤双清李志雄刘侃
3
(1.黄山学院信息工程学院,安徽黄山
245013;
2.
三峡大学机械与材料学院,湖北宜昌
443002;
3.
湖南大学电气工程学院,湖南长沙
410082)
摘
要:飞轮电池储能用集成电机时交非线性特点使得传统
PID
控制难以得到理想的控制性能,为此基于
BP
神经网络
研究了一种新颖的飞轮电池电力转换器。该控制器结合
BP
神经网络自学习能力和
PID
控制的全局渐近稳定性能,通过神经
网络在线优化调节目
D
参数,以实现对飞轮电池的高性能控制。其中,采用变学习这率的神经网络学习算法,学习速率随
收敛过程误差的大小而自适应地进行调整,同时使用遗传算法
(GA)
优化得到
PID
参数的初始值,这可加快神经网络学习训
练的收敛速度并避免陷入局部最小,进一步提高控制性能;另外,
PWM
采用
SVPWM
技术以增强能量转换效率和减小转矩
脉动。数字仿真表明,基于所提出的
BP-PID
控制的电力转换矢量控制系统能够使飞轮电池在充放电两端都具有较快动态响
应,较小超调,较高稳态精度以及较强的鲁棒性,控制效果明显比传统
PID
好。
关键词:飞轮电池
BP
神经网络
PID;
遗传算法
1 Intruduction
The
energy
density
and
power
density
of
flywheel
hatteIγare
greater
than
chemical
hatteries
[ 1.3] ,
which
make
the
flywheel
hattery
more
useful
for
application
in
wind
power
,
EVs
,
UPS
,
aerospace
,
etc.
And
sev-
eral
recent
advanced
improvements
in
material
,
mag-
Received
date: 2008-11-14;
Revised
date:
2008-12-25
netic
hearings
and
power
electronics
make
it
more
com-
petitive
in
these
applications.
As
one
of
the
key
issues
,
the
power
converter
plays
an
important
role
in
confirming
the
high
peno
口
n
ance
of
the
energy
exchange.
Referenoe
[ 4-5 ]
employ
a
PI
controller
to
actualize
the
charging
and
dischar-
gr
吨,
and
reference
[ 6 ]
uses
a fuzzy
methodology
just
Foundation:
Supported
by
Educational
Department
of
Hubei
Province
Natural
Science
Program
(2033AOOI
);
.
Hubei
Province
Natural
Science
Foundation(2005AB
A2
94
)
Author introduction:
Jiang
Yu(
1982-) ,
female
,
Liaoning
Fushun
,
master
,
her
reasearch
interests
include
mechainical
intelligence
and
its
sim-
ulation
,
ets.
蒋
宇
(1982-)
,女,辽宁抚顺人,研究生,主要研究方向为机械智能仿真等。