扩展蚁群优化算法在电力电子电路设计中的应用
105 浏览量
更新于2024-07-15
收藏 1.05MB PDF 举报
"Extended Ant Colony Optimization Algorithm for Power Electronic Circuit Design"
这篇研究论文主要探讨了扩展的蚁群优化算法（Extended Ant Colony Optimization, eACO）在电力电子电路设计中的应用。蚁群优化算法（Ant Colony Optimization, ACO）是一种受到蚂蚁寻找食物路径行为启发的全局优化算法，通常用于在图中寻找最短路径。然而，传统的ACO算法并不适用于电力电子电路（Power Electronic Circuits, PECs）的设计，因为这些电路中的组件如电感器通常是连续值而非离散值。
电力电子电路设计中，组件如电阻和电容可能有固定的离散值选择，而电感等其他元件则通常具有连续的参数范围。论文指出，这使得直接使用标准ACO算法来寻找电路中最佳组件值变得困难。为解决这个问题，作者提出了eACO算法，它能够处理既有离散值又有连续值的组件优化问题。
eACO算法的核心思想是结合正交设计方法（Orthogonal Design Method, ODM）。通过ODM，算法可以动态更新具有连续值的组件数据库，从而在优化过程中更有效地搜索到最佳值。这种方法提高了ACO在处理电力电子电路这类问题时的适应性和效率。
电力电子电路设计是一个复杂的问题，涉及到多个变量的优化，包括电路的性能、效率、成本以及尺寸等。eACO算法提供了一种新颖的优化策略，能够在保证电路性能的同时，考虑实际制造中组件值的限制，从而为设计师提供更优的解决方案。
通过实证研究，论文验证了eACO算法的有效性，并将其与传统ACO和其他优化方法进行了比较。结果显示，eACO在找到最优解的速度和精度上具有显著优势，这表明该算法对电力电子领域的电路设计具有重要的理论和实践价值。
这篇论文为电力电子领域提供了一个强大的优化工具，即扩展蚁群优化算法，它能应对具有离散和连续组件值的电路设计挑战，有望推动电力电子电路设计技术的发展和创新。
IEEE
Proof
4 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 00, NO. 00, 2008
Similarly, c
qj
points to a parameter set stored in a compo-213
nent database C. If there are S
c
j
choices for c
qj
, there are S
c
j
214
parameter sets in C (i.e., c
qj
∈ [1, S
c
j
]). The structure of C215
is (5) as shown at the bottom of the this page, where c
j
vn
is216
the value of the nth choice for c
qj
,c
j
pn
is the pheromone of217
the choice c
j
vn
, and c
j
tol
is the tolerance of the jth component.218
The variables c
j
v1
,c
j
v2
...,c
j
vk
..., and c
j
vS
c
j
∈ [C
j
min
,C
j
max
]are219
generated throughout the algorithm, as will be described in220
Section III, C
j
min
and C
j
max
are the lower and upper limits of221
the search space for the jth component.222
B. Fitness Function223
Each ant has a fitness value [i.e., Φ
q
in (1)] showing the degree224
of attainment on the optimization objectives. The multiobjective225
functions used in [22] for optimizing PCS and FN are adopted.226
Their definitions are described as follows.227
1) Fitness Function for Optimizing PCS: The fitness func-228
tion for evaluating the ants for optimizing PCS is based on the229
following considerations, including:230
1) the steady-state error of the output voltage v
o
within the231
required input voltage range v
in
∈ [V
in,min
,V
in,max
] and232
output load range R
L
∈ [R
L,min
,R
L,max
];233
2) the operation constraints on circuit components, such as234
the maximum voltage and current stresses, ripple voltage,235
and ripple current;236
3) the steady-state ripple voltage on v
o
;237
4) the intrinsic factors associated with the components, such238
as the total cost, physical size, etc.239
Each of the aforementioned objectives is mathematically de-240
scribed by an objective function. The formula of each objective241
function i s given in [22]. Φ
q
is the total sum of three compo-242
nents, including Φ
q
−1
, Φ
q
0
, and Φ
q
+1
. Φ
q
−1
is the fitness value of243
ant q when the component values are all reduced by the corre-244
sponding tolerance stored in the last columns in (4) and (5). Φ
q
0
245
is the fitness value of ant q when the components are in their 246
corresponding nominal values. Φ
q
+1
is the fitness value of ant q 247
when the component values are increased by the corresponding 248
tolerance stored in the last columns of (4) and (5). Thus 249
Φ
q
=
Φ
q
−1
+Φ
q
0
+Φ
q
+1
3
. (6)
2) Fitness Function for Optimizing FN: The fitness function
250
for optimizing FN is based on the following considerations: 251
1) the steady-state error of v
o
within the required input volt- 252
age range v
in
∈ [V
in,min
,V
in,max
] and output load range 253
R
L
∈ [R
L,min
,R
L,max
]; 254
2) the maximum overshoot and undershoot, and the settling 255
time of v
o
during the startup; 256
3) the steady-state ripple voltage on v
o
; 257
4) the dynamic behaviors as in 2) during the input voltage 258
and output load disturbances. 259
Again, the objective functions of these objectives are defined 260
in [22], and the tolerances of the components will be taken into 261
calculations as in (6). 262
III. OPTIMIZATION PROCEDURES 263
With the help of the flowchart in Fig. 3, the optimization 264
procedures consist of eight steps. 265
Step 1—Initialization: The values of d
i
vm
(for i = 266
1, 2,...,M and m =1, 2,...,S
d
i
) in (4), and the tolerances 267
d
i
tol
in (4) and c
j
tol
in (5) are initialized with the available 268
values provided by the manufacturers, and that of c
j
vn
(for 269
j =1, 2,...,N and n =1, 2,...,S
c
j
) in (5) are initialized 270
evenly over the search space and the number of choices as- 271
signed (i.e., S
c
j
). Thus 272
c
j
vn
=(n − 1)
(C
j
max
− C
j
min
)
S
c
j
+ C
j
min
. (7)
D =
{d
1
v1
,d
1
p1
}, {d
1
v2
,d
1
p2
}, ··· {d
1
vm
,d
1
pm
}, ··· {d
1
vS
d
1
,d
1
pS
d
1
},d
1
tol
{d
2
v1
,d
2
p1
}, {d
2
v2
,d
2
p2
},
.
.
. {d
2
vm
,d
2
pm
}, ··· {d
2
vS
d
2
,d
2
pS
d
2
},d
2
tol
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
{d
i
v1
,d
i
p1
}, {d
i
v2
,d
i
p2
}, ··· {d
i
vm
,d
i
pm
}, ··· {d
i
vS
d
i
,d
i
pS
d
i
},d
i
tol
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
{d
M
v1
,d
M
p1
}, {d
M
v2
,d
M
p2
}, ··· {d
M
vm
,d
M
pm
}, ··· {d
M
vS
d
M
,d
M
pS
d
M
},d
M
tol
(4)
C =
{c
1
v1
,c
1
p1
}, {c
1
v2
,c
1
p2
}, ··· {c
1
vn
,c
1
pn
}, ··· {c
1
vS
c
1
,c
1
pS
c
1
},c
1
tol
{c
2
v1
,c
2
p1
}, {c
2
v2
,c
2
p2
},
.
.
. {c
2
vn
,c
2
pn
}, ··· {c
2
vS
c
2
,c
2
pS
c
2
},c
2
tol
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
{c
j
v1
,c
j
p1
}, {c
j
v2
,c
j
p2
}, ··· {c
j
vn
,c
j
pn
}, ··· {c
j
vS
c
j
,c
j
pS
c
j
},c
j
tol
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
{c
N
v1
,c
N
p1
}, {c
N
v2
,c
N
p2
}, ··· {c
N
vn
,c
N
pn
}, ··· {c
N
vS
c
N
,c
N
pS
c
N
},c
N
tol
(5)
剩余16页未读，继续阅读
2010-10-17 上传
2021-02-08 上传
2021-02-09 上传
2021-06-01 上传
2013-04-18 上传
2021-02-08 上传
2021-05-30 上传
weixin_38636577
- 粉丝: 4
- 资源: 935
我的内容管理
展开
