 Open Access
 Total Downloads : 88
 Authors : Fan Wang
 Paper ID : IJERTV6IS070279
 Volume & Issue : Volume 06, Issue 07 (July 2017)
 Published (First Online): 26072017
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
Microgrid Capacity Optimisation with a Modified Particle Swarm Optimisation Algorithm
Fan Wang School of Management
Zhejiang University Hangzhou, China
Abstract Particle swarm algorithm the search speed is quick, simple calculation, the efficiency high characteristic, is suitable for the optimization problem of single objective function.But the basic particle swarm optimization (PSO) algorithm is easy to fall into local optimal solution has adverse effect on the correctness of the result.Between the global search and local search, this paper found a suitable inertia weight factor to omega, using linear decreasing inertia weight, make them get the optimal solution space.
This article includes the fan, photovoltaic power generation, diesel generators and energy storage device independent model of micro grid economy model, think the load constant, in a systematic and objective function is minimum total cost, consider the micro source equipment investment and operation and maintenance costs, the cost of replacement battery and diesel generators of the cost of fuel and pollution costs.According to the discontinuity of wind turbines and photovoltaic cells and consider spare battery and diesel generator operating characteristics, in view of the storage battery charging and discharging characteristics of the analysis of the working process of the storage battery as backup power supply, according to the emission characteristics of diesel generators to consider increase the cost of pollution in the objective function.Based on improved particle swarm algorithm to optimization of system, work out the optimal capacity configuration of system power supply, at last, numerical examples verify its rationality.
Keywordsmicrogrid; particle swarm optimisation;
INTRODUCTION
A microgrid describes a smallscale power generation, supply and utilisation system. Microgrids have a full suite of power generation, power utilisation and power supply functions, which can optimise the distribution of energy in a
grid. Microgrids can be classified as independent microgrids and gridconnected microgrids according to whether or not they are connected with the conventional large grid. In addition, microgrids can operate in a variety of modes. Microgrids can be flexibly accessed or connected with the surrounding large grid based on the current situation of load and power supply, making them an important component of nextgeneration smart power networks.
When operating in connection with the surrounding power grid, a microgrid and the surrounding distribution network can exchange power back and forth, which can trim peaks and fill valleys in the main grid power supply. If the main grid breaks down, a large microgrid can be switched into independent operation to supply power continuously for end users. An independent microgrid does not need to connect with the conventional large grid, because it can consistently supply power over a long time with generation from various distributed power supplies within the microgrid. Independent microgrids often used in remote areas like islands and mountainous terrain. Such locales have little access to infrastructure for delivering conventional fuel, and power transmission is inconvenient, but rich renewable energy resources are often available. Thus, microgrids not only facilitate local residents lives, but they also use and develop local renewable resources.
In an independent microgrid system, distributed power
supplies can be divided into two categories according to whether or not they can be readily dispatched. Affected by weather conditions and variation in natural resources, the output of renewable energy generators such as wind turbines and photovoltaic (PV) arrays cannot be regulated. Diesel power generation, energy storage devices, fuel cells, and the like can, on the other hand, provide ondemand power. In order to make full use of renewable energy resources, the
relatively unreliable renewable power generation units are often used preferentially in the operation of independent microgrids. Conventional power generation units can supplement renewable generation sources to balance the variability in their power output and allow integrated microgrids to make full use of renewable energy. In the case that the traditional large grid cannot support a microgrid, the influences of weather conditions, local availability of resources and load demand are more obvious. Therefore, careful analysis of how to allocate the capacity and quantity of various distributed power supplies within a microgrid in order to optimise distribution is important for the planning of an independent microgrid. This optimisation predicts the reliability and quality of power output, in addition to improving economy and rationality in the microgrid.
Optimisation Model for Microgrid
In this paper, we propose an optimisation model for an independent microgrid with diverse distributed generation resources under constant load. The process of optimising the microgrid system should be divided into the following steps: determining the objective function, determining the
B. Objective Function for a Microgrid System Optimisation Model
Optimisation problems can be divided into singleobjective and multipleobjective problems. The selection of an objective function determines the optimal scheme to a considerable extent. We choose to minimise the total cost of the microgrid system in this paper. The total cost of a microgrid system includes installation, operation, and maintenance costs. In order to ensure the reliability of the microgrid system, we introduce a cost for power outages. To ensure a degree of environmental protection around the microgrid system, an additional cost is introduced to reduce the negative impact of diesel emissions on the environment. The obtained objective function, i.e. the minimum total cost of system is:
3
minC CPi xi CAmi xi Crep x3 Ccon CDE x4 max0, set
i1
21
The above objective function includes four kinds of power
X x , x , x , x
constraint, and then performing a further analysis on the
supply in the system, where
1 2 3 4
are
specific operation parameters of the microgrid combined with the work characteristics of each microgrid component.
optimisation variables that represent the number of each microresource (wind, light, storage, and diesel, respectively);
A. Operational Description of the Microgrid Model
The distributed power supply of the microgrid system
CPi
and
CAmi
are the equipment investment cost and the
modelled in this paper includes wind generators, PV power generation stations, diesel generators and energy storage devices. Wind energy and solar energy are renewable and clean resources with wide distribution, straightforward
annual operation and maintenance costs, respectively, of the ith microsource; Crep is the annual resetting cost of the
C
collection and utilisation, so the wind and PV generation stations are used preferentially during microgrid operation. When the wind and PV output is sufficient for the load requirement, the spare power will be stored by storage batteries. When wind and PV output cannot supply loads due
energy storage unit; con
CDE is the cost of fuel; shortage punishment; set
is the cost of pollution control;
is the coefficient for power is the presupposed load supply
to factors like wind speed and light conditions the storage battery begins to suply power. At this time, wind and solar power jointly output power with the storage unit, and the energy storage unit is in the discharging state. The active capacity of storage battery exceeds the limitations imposed on the battery by operating either in charging or discharging modes. Therefore, the storage battery can be cut out from the system and connected to a diesel generator to supply power
rate, which is 0.98 in this paper; and is the actual value of
load supply rate in the microgrid power supply system.
The following functions specify the costs in the microgrid model, starting with installation cost CP:
P ci rf
C C Z ,Y
22
In the above formula, Ci is the unit price of the ith
for loads.
microsource; and
Zrf
represents the coefficient of return
on investment, whose formula is as follows: C. Constraints
Zrf
,Y
1 Y
1 Y 1
23
Next, we detail the constraints in our optimisation model. In this microgrid system composed by a variety of distributed power supplies, the power output of each microsource is
In the above formula, represents the discount rate, the
current value of which generally is 8%; and Y represents the service life, whose value is 20 a in this paper.
Operation costs are represented with CAm:
affected by not only weather and natural conditions, but also the constraints of load demand and energy storage unit specifications.
Site conditions and the overall scale of the microgrid power
CAm KOM Pj
K
24
Pj
generation system will constrain the available amounts of each microsource.
PV
0 NWT NWT max
OM is the coefficient of power cost;
is the actual
0 N
NPV max
power output of microsource.
Power outages are represented with a cost function C:
0 NB NB max
0 NDE NDE max
211
C max0, set
25
Power outages are mainly determined by the load power supply rate , calculated as follows:
In the formula, Ni represents the number of each
N
T microsource, and max
represents the specific maximum
Ploss t
T
1 t 1
PL t
number of each microsource respectively.
The energy storage unit is constrained in how it can provide and store power as follows:
t 1
26
P P t P
min

max
Plosst max0, PL t PPV t PB t PDE t
Pmin P t Pmax
SOC SOCt SOC
27
min
max
(210)
In the following formula,
Ploss t
is the power vacancy at
Pmin ,
Pmax ,
Pmin , and
Pmax
respectively represent
hour t. We model the microgrid for one year so the total time the upper and lower limits of the energy storage units
period is
T 8760 .
charging and discharging input and output.
SOCmin
and
The following formula determines the pollution cost Ccon:
SOCmax
respectively represent the minimum and
Ccon
T k
Am Bm PDE
t dt
maximum of the charging state of energy storage unit. In this
0 m1 28
paper,
SOCmin and
SOCmax
respectively is 0.2 and 1.
In (28), m is an index for the type of pollutant, k is the
total number of kinds of pollutants, Am represents the
pollution control cost of each pollutant, and Bm is the emission coefficient for each pollutant.
The coordinated output index for wind and solar power is an important index for microgrid operation. It reflects whether the wind and PV power generation can meet load demands or not. Coordinating the respective shares of wind and PV output can minimise energy waste and save power generation costs. The formula for the coordination index is as following:
365 24 t
by vi
v v
v
id , and the spatial positions of the
i i
U m1 t 1
8760
(211)
particles can be expressed by xi
xi1
, xi 2
,…, xin
, where
1, P
t P
t P t
i is an index for each particle. Equation (41) describes the
t 0
PV WT L
position of particle from time t to time (t + 1).
0, PPV
t PWT t
PL t
212
t 1 t
t t t t
IMPLEMENTATION OF MICROGRID OPTIMISATION
vid
vid c1r1
pid xid

c2r2
pgd xid
31
MODEL

Particle Swarm Optimisation (PSO)
Note that PSO is an optimisation algorithm based on group intelligence. PSO was proposed by Craig Reynolds in 1987 for the simulation and study of the social system of birds. PSO simulates the behaviour of birds foraging in nature. In a
In the formula, is the inertial weight coefficient, which can affect the search ability of the ith particle. Large values
of enhance the overall search ability of particle, and small
values enhance the partial search ability of particles.
When = 1, we refer to the problem as a basic particle algorithm; if 1, we refer to the problem as the standard
given region, a group of birds searches for a randomly placed food resource. Each bird does not know the exact location of
particle algorithm. r1
and r2
are random numbers between
the food, but each bird does know the distance between the target food and their own position. The best method for
0 and 1. c1
and c2
are learning factors.
finding this piece of food is for each bird to share its own distance from the food so that all the birds can concentrate
Equation (42) describes the velocity adjustment of particle
i from time t to time (t + 1).
their search around the bird closest to the food. Each bird in this model can be considered as a particle, and these particles
t 1 t t 1
x x v
id id id
32
are given random initial positions. Each particle has its own fitness determined by the optimisation constraints and its movement speed.
Solutions to the PSO problem usually conform to the principles of proximity, quality, stability, adaptability and diversity. PSO is an algorithm of heuristic overall search, theoretically derived from observations of group intelligence. In the behaviour of foraging birds, the distance between each bird and food and flight speed is shared between companions. Similarly, each particle in the model can affect its surrounding particles through information sharing. With constant dynamic adjustments, the search of the overall area can be reduced to the search of the region where particles are closest to the target. In the process of solving an optimisation problem each particles position is a solution of the optimisation problem, and the positions of the whole particle swarm represents the entire solution set. By comparing the distance between each particle and the optimal solution, i.e., the fitness function, the optimal solution in the space can be determined by emergent dynamic adjustments to the whole particle swarm to realise the optimisation of problem. In PSO, the velocity of the particle swarm can be expressed

Improvements on PSO
The basic procedure of PSO is charted in Fig. 32:
overall search, and a linearly decreasing inertia weight is used to obtain the optimal solution within the search space. In this paper, we use a linearly decreasing inertia weight factor to
add two parameters
min
and
max
. Formula (43)
represents the linearly decreasing inertia weight:
max
t
T
max
in
43
In our model,
max = 0.9 min = 0.4, which represent the
Fig. 32. Flow chart of PSO.
PSO has the characteristics of fast search speed, simple calculation and high efficiency, and it is suitable for solving optimisation problems with a singleobjective function. However, the basic PSO algorithm is easily trapped by local
optimal solutions, which has a negative effect on the overall
correctness of the result. The inertia weight factor has a great influence on the convergence speed of the PSO
algorithm. We consider that a suitable inertia weighting factor should be found between a local search and the
inertia weight of the maximum number of iteration and the initial inertia weight, respectively; t represents the current iteration; and T represents the maximum number of
iterations. In this paper, the number of particles is 50, and the maximum number of iterations is 80.

Optimisation strategy
we propose the following allocation hierarchy, in view of the costeffectiveness, reliability and environmental compliance of microgrid systems:
Wind and PV power generation should be allocated first. Energy storage systems should provide power in the event that wind and PV power is unavailable.
Diesel generator should provide power as a last resort, only when the combined output of renewable power generation and the storage battery cannot meet demand.

Optimisation Process
Due to annual cycles of wind speed, sunshine and other meteorological factors, a year will be used as the scheduling cycle for microgrid system. Surf the Internet to collect microgrid location in terms of light and wind speed and other meteorological data, and because of wind speed and illumination change are more slowly, so we set for the unit step of 15 min and we believe that wind turbines, PV output constant within 15 min.
Calculate the unit power output of wind and light; Set the upper and lower limit and the state of charge of the charging and discharging power of the energy storage unit; Set the power output of the diesel generator to 0.
Initialise the particle swarm and calculate the power output of the microsource of the scenery and wood storage in 8,760 hours.
The target function is calculated, and the optimal capacity allocation of the distributed power supply is obtained by
using the improved PSO algorithm. Optimisation flow chart is shown in Fig. 33:
Fig. 33. Flow chart of optimisation of PSO algorithm.
RESULTS
This paper models a microgrid system including wind power units, PV power generation units, diesel power generation units and energy storage units. A year serves as the dispatching cycle. Since meteorological conditions such as wind speed and illumination change slowly, we use 15 min as the time step. In each 15min time period, the output of the
wind turbines and PV arrays remain constant. According to the optimal scheduling strategy, the optimisation target is found with the particle swarm algorithm modified as described above.

Microgrid model and parameters
In this paper, a microgrid system including wind turbine, PV power generation, diesel generator and energy storage units is simulated. The microgrid system is isolated from the surrounding distribution grid and the load is considered to be approximately constant. The parameters of our model are displayed in Table 41.
Component Type
Rated Powe r,
P,
Kw
Installati on Cost
Cp, Ten Thousand Yuan
Operation And Maintenan ce Cost Cam, Ten Thousand
Yuan/A
Fuel Expense Cde, Ten Thousan d Yuan/
Kwh
Replaceme nt Cost Crep, Ten Thousand Yuan/A
Turbine, set
20
50
0.5
0
0
PV
conversion cell, set
10
35
0.2
0
0
Diesel
engine, set
25
20
0.8
0.0004
0
Storage battery,
piece
45
15
0.15
0
7
TABLE 41. PARAMETERS OF INDEPENDENT MICROGRID MODEL
For the wind turbine modelled in this paper, the cutin wind speed is
3.5 m/s, the rated wind speed is 11 m/s, and the cutout wind speed is 22 m/s.
Fig. 41. Annual wind speed data.
Fig. 42. Year of illumination data.
Fig. 43. Annual temperature data.
Fig. 44. Annual load vs. time.
According to the data we collected, the wind speed changes rapidly and it is generally unpredictable within a year. In
general, lighting conditions vary predictably, but they are also affected by local weather patterns. This simulated data confirms our expectation that renewable resources are relatively unreliable.

Optimal configuration results and analysis
We chose several independent parameters in the model are set to represent the operation of the microgrid system. The renewable resource coordination output index is set to 0.8, and the load rate is set 0.97. According to the parameters of each generation unit discussed above and the gathered wind speed and light and temperature data, we calculated the power output of a single wind turbine and a single PV cell.
Fig. 45. Power output of a single wind turbine.
Fig. 46. Power output of a single PV cell.
By improving the analysis and calculation of particle swarm algorithm, the optimal configuration results and optimisation indexes of microgrid were obtained, as shown in the following table:
Wind Driven Generator, Set 
PV Module, Set 
Diesel Generator, Set 
Storage Battery, Piece 
Total System Cost For The Year, Ten Thousand Yuan 
19 
206 
16 
2 
69.436 
TABLE 42. Optimised Configuration Results Of The MicroGrid System Of Renewable And NonRenewable Resources
TABLE 43. MicroGrid Optimisation Index
Annua l Rate Of Load Power Suppl y 
Annual System Total Cost, Ten Thousan d Yuan 
Renewable Output Index 
Electricit y Deficienc y Rate 
Use Ratio Of Renewable Resources 
0.9928 
69.43 
0.8221 
0.0009 
0.8174 
From the results of system optimisation we can see that the microgrid system allows is a large number of configurations for the PV arrays and energy storage devices. From the historical data of local solar irradiance we can see that residential solar energy is very rich in reserves at the modelled site. The solar irradiance data shows that this particular location is ripe for development of PV generation resources, as days are long and weather is generally clear. From the wind speed historical data we can see that the wind speed in the region is more intermittent, and even much of the time the wind is completely still. Moreover, the power curve of the wind generator unit shows that the turbine operates intermittently and an increase of turbine quantity can only increase the system power output for a brief period of time. If we blindly increase the quantity of the fan, we will greatly increase the equipment costs and need to provide more battery energy storage devices to store surplus energy. With relatively high costs for batteries currently, adding more turbines will therefore be expensive. Therefore, in the configuration of microgrids, the number of wind turbines should be reduced accordingly.
Analysis of the output ofdiesel generator
Due to local weather changes reasons or seasonal patterns, the power generation units of the microgrid system sometimes fail to provide sufficient power to support the load. When that happens, because the battery is incapable of supplying electricity since it is charging, and in order to guarantee the reliability of the system, a diesel engine can operate as a backup generator to reduce the risk of power outages. From the irradiance data comparison in Fig. 47 and 42 we can see that diesel engines work less frequently in the summer and mainly provide power supply when days are short. The diesel generator mainly supplements the PV arrays in our optimised model.
Fig. 47. Annual output curve of diesel engine.
Wind Driven Generator, Set 
Pv Module, Set 
Diesel Generator, Set 
Storage Battery, Piece 
The Total System Cost Of The Year, Ten Thousand Yuan 
19 
189 
17 
0 
69.542 
2) Improved PSO algorithm compared with traditional PSO TABLE 44. Optimises The Result Of Traditional PSO
Annual Rate Of Load Power Supply 
Annual System Total Cost, Ten Thousand Yuan 
Scenery Synergy Output Index 
Electricity Deficiency Rate 
Use Ratio Of ReGenerable Resources 
0.9919 
69.542 
0.8006 
0.0013 
0.8023 
TABLE 45. Results From The Optimisation Of Traditional PSO
Computing Time (S) 
Iterations Convergence 
To 

Improved optimisation 
particle 
swarm 
394100 
99 

Traditional optimisation 
particle 
swarm 
133740 
51 
TABLE 46. Computational Cost Comparison Between Improved And Traditional Particle Swarm Algorithms
From improved particle swarm algorithm of the comparison of optimised result of in Table 42 we can see that the configuration of distributed power supply results is roughly the same, but total cost when using our improved particle swarm algorithm for the optimisation is slightly less than it is when using the traditional PSO algorithm. At the same time, the load supply rate and renewable energy utilisation are slightly higher with the improved algorithm.
It can be seen from Table 46 that the calculation time of particle swarm algorithm is great and the convergence number is relatively high, so this method needs further improvement to achieve computational efficiency.
V. CONCLUSION
Models of independent microgrids combine distributed power supplies with a variety of renewable resources, which exploits novel energy sources sufficiently while meeting load demands. For the remote mountains or islands which are rich in natural resources but isolated from powerdistribution infrastructure, optimisation is of great significance to the research and development of microgrids. In microgrid systems, the scheduling model of distributed power supply power output is inherently unpredictable and unstable, which affects the economy and reliable operation of the entire distribution network, increasing the importance of installing rationally designed microgrids. This paper models wind turbines, PV power generation, diesel generators and energy storage devices in an independent microgrid. We used a modified implementation of the PSO algorithm to determine optimal allotments of distributed generation in a microgrid at a particular location. The main results of the study are as follows:
Establish the economic model of distributed generation units
in microgrid, including installation cost and running maintenance cost, and analyses its working characteristics.
The objective function of microgrid optimisation is determined, including the cost and operation maintenance cost of each microsource, and consider the introduction of power failure penalty and the cost of diesel emission. According to the actual operation condition of the microgrid and distributed generation unit, the constraint conditions are determined. Finally, the operation strategy of microgrid optimisation is formulated.
(2) Improve the particle swarm algorithm, use linear decreasing inertia weight PSO algorithm, and consider
between the global search and local search to find a suitable
inertia weight factor , and use linear decreasing inertia weight to reach a space to the optimal solution. Combined
with the example, the optimisation model of the microgrid is calculated by improving the particle swarm algorithm, and the optimal configuration results of the microgrid system are obtained and analysed. Analyse the changing situation of the scenery through the annual output force of the diesel engine; Analyse and compare the difference between PSO and traditional particle swarm algorithm from the perspectives of calculation results, operation time and number of iterations; According to the load power supply rate, scene coordination, annual system total cost and other indicators, the influence of different load power rate on cost is analysed.
The optimisation design of research in the network configuration meets basic expectations. When considering distributed generation capacity, we often make the load power supply rate as large as possible for convenient analysis, but in practice we find that the load demand is not a linear relationship with the system total cost. The choice of power supply rate should be considered with the total cost and landscape indices. Choosing an appropriate scale for microgrid generation units is more economical, more reliable, and more environmentally friendly than increasing the operational benefits of the microgrid.
Our study also suffered some disadvantages that we plan to improve upon in future work. We could perform a more robust comparison of the improved particle swarm algorithm advantages and traditional optimisation algorithms. The network structure in our model is simple, and does not take into account the processes of transmission line loss and other factors. We plan to further improve the algorithm detailed above, as well as incorporate line loss in our model to deliver more realistic results.
REFERENCES

Ding Ming, Wang Bo, ZHAO Bo, Chen Zinian.An Capacity Optimization of Independent Firewood Micronetwork System [J]. Power System Technology, 2013, (03): 575581. ,,,
. [J].
,2013,(03):575581.

Qin Yuejin, Wang Youchun, Fan Li, Wen Shui flash, Liu Xin Optimization of distributed power supply in microgrid[J]. Journal of Wuhan University (Engineering Science),2014,(03):359364. ,,,,.
[J]. (),2014,(03):359364. 
He Yiren, Teng Huan, Gao Hongjun. Optimal allocation of multimicrosource independent micro grid capacity considering seasonal characteristics [J]. Modern Electric Power, 2017, (01): 814.
,,.
[J]. ,2017,(01):814. 
Gao Feng, Yan Tao, Tang Wei, Zhang Xiaohui, Zai Hongbin, Qiao Xiaodong. Optimal allocation of multitarget power capacity for independent wind / light / storage hybrid microgrid [J]. Smart Grid, 2016, (08): 802810.[4],,,,,.
/ / [J].
,2016,(08):802810.

Zhang Youbing, Bao Kankan, Yang Xiaodong, Ren Shuaijie, Qi Jun, Xie Luyao.Multiobjective optimization of independent microgrids [J]. Journal of Zhejiang University of Technology, 2016 (06): 619623.[5],,,,,.
[J]. ,2016,(06):69623. 
Jing Zhaoxia, Hu Rongxing, Yuan Zhaoxin, Zhu Jisong, Wu Qinghua. Optimal allocation of island microgrids with wind / light / pump storage and load response [J]. Power System Automation, 2017, (01): 6572 +116.[6],,,,. //
[J].,2017,(01):6572+116.

Zhou Xiaoyan, Liu Tianqi. Micropower grid micropower combination and capacity allocation optimization [J]. Power System and Its Automation, 2013, (05): 149155.[7],.
[J].
,2013,(05):149155.

Zhou Zeyuan, Su Dawei, Wang Zhicheng, Zhu Weiping, Li Peng.Economic and Environmental Protection of Independent Multivariate Complementary Micrographs Based on Adaptive Variation Particle Swarm Optimization [J]. Power Grid & Clean Energy, 2015, (04): 814.[8],,,,.
[J].,2015,(04):814.

Mao Luming, Yang Yuqun.Optimization of power supply capacity of microgrids considering the importance of load [J]. Industrial Instrumentation and Automation Devices, 2016, (06): 6265 + 90.[9]
,. [J].
,2016,(06):6265+90.

Yand Peipei, Ai Xin, Cui Mingyong, LEI Zhili.Analysis of micronetwork economic operation based on particle swarm optimization algorithm with multiple energy supply systems [J]. Power System Technology, 2009, (20): 3842.[10],,,
.
[J]. ,2009,(20):3842. 
Liu Shu, Li Zhengli, Wang Yi, Ma Rui, Lu Dan, Liu Hao Ming.Capacity Optimal Configuration of Energy Storage Device in Microgrid with Distributed Power Generation.[J]. Power System Protection and Control, 2016, (03): 7884.[11],,,,
,. [J].
,2016,(03):7884.

Dou Xiaobo, Yuan Jian, Wu Zaichun, Ni Yimin, Fan Chen, Xiaoyu.Optimization and Optimization of Grid Capacity Optimization for Gridconnected Microgrid [J]. Electric Power Automation Equipment, 2016, (03): 2632.[12],,,,,
. [J].
,2016,(03):2632.

Li Peng, Xu Weina, Zhou Zeyuan, Li Rui. Optimization of microgrid optimization based on improved gravitational search algorithm [J].
Proceeding of the CSEE, 2014, (19): 30733079.[13],,,. [J]. ,2014,(19):30733079.

Cao Bei, Yang Yue.Considering the optimal configuration of micro – network capacity of scenery water storage[J]. Journal of Hydroelectric Energy Science, 2015, (06): 209212 + 150.[14],.
[J].
,2015,(06):209212+150.

Kang Xiaoming, Liu Qingzhen.Optimization of power capacity of hybrid microgrid based on improved particle swarm optimization [J]. Power and Electric Engineering, 2012, (04): 36 + 25.[15],
. [J].
,2012,(04):36+25.