Академический Документы
Профессиональный Документы
Культура Документы
Published in The Journal of Engineering; Received on 6th October 2017; Accepted on 2nd November 2017
Abstract: Existing methods to determine the battery bank capacity of micro-grid composed of photovoltaic (PV) generation and energy
storage device are classified and summarised. In the viewpoint of peak load filling and implementing continuous power supply by independent
and grid-connected micro-grid composed of PV generation and energy storage device, the method for optimal configuration of battery bank
capacity is researched. A method of optimising the storage capacity of DC micro-grids considering the randomness of PV and load is proposed
to maximise the local absorption of renewable energy. Based on the K-means algorithm, the engineering selection basis of PV and storage
capacity in DC micro-grid system is given, and its economic benefit is evaluated. Finally, the efficiency and economy of the energy
storage capacity selection method are verified by practical application of 31.5 MW PV power station in Badaling, Beijing.
J. Eng., 2017, Vol. 2017, Iss. 13, pp. 859–864 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0452 Attribution License (http://creativecommons.org/licenses/by/3.0/)
When the battery is discharged, PSB (t ) ≥ 0, the remaining cap-
acity at time t is
This is an open access article published by the IET under the Creative Commons J. Eng., 2017, Vol. 2017, Iss. 13, pp. 859–864
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0452
3.2 Operating mode of PV and energy storage (i) Light intensity and temperature values change with the seasons,
that is, the PV output power characteristics are significantly dif-
The PV unit has two modes of operation: maximum power point
ferent. Therefore, considering the effect of the season on the
tracking (MPPT) mode and constant voltage mode. MPPT is a tech-
output characteristics of PV power supply, this paper takes k
nology to adjust the equivalent input impedance to obtain the
as 4.
maximum output power through a series of control device and
(ii) The selection of the initial clustering centre has a great influ-
control strategy [12]. The efficiency of a PV system is equal to
ence on the clustering results. Once the initial value selection
the product of the photoelectric conversion efficiency and the
is not appropriate, effective clustering results may not be avail-
maximum power point tracking efficiency of the panel. The
able. Therefore, this paper randomly selected 1 day as the initial
energy storage unit bi-directional DC–DC converter operates in
clustering centre in each quarter, taking into account the influ-
constant voltage mode and stop operating mode. In order to
ence of the season and reducing the effect of the initial value on
protect the battery operating life and improve economic efficiency.
the effectiveness of the clustering results.
The SOC is set in the controller to the upper and lower limits to
prevent the battery overcharge and over discharge.
4.2 Method of capacity allocation for energy storage devices
4.2.1 Model simplification and assumptions
4 Capacity allocation of energy storage device (i) This paper studies the allocation method of energy storage
capacity from the perspective of engineering practice, in which
4.1 Data processing method of load power and output power of PV output power and load demand power are given.
PV power based on K-means algorithm (ii) The DC micro-grid does not deliver energy to utility grid, but
Considering the actual situation, the PV power and load power only absorbs energy from AC grid in grid-connected mode.
are uncertain and real time. If the PV output power data and the (iii) Energy storage batteries cannot produce energy, only as the
load power data are selected by random sampling method, energy buffer of the intermediary. There is no over-imbalance
although convenient and simple, but with a certain sampling of power supply in the micro-grid system.
error. In order to obtain statistical and practical significance of
the storage capacity configuration results, dimension models of
daily load capacity data and daily PV power generation capacity 4.2.2 Objective functionThe objective function is to minimise inter-
data are established firstly. The K-means algorithm is used to active power between the micro-grid and the AC grid and the costs
cluster the above data within 1 year to obtain representative of energy storage device, as follows:
k group data.
K-means algorithm is one of the most classical clustering algo- (i) The lower limit of energy storage capacity based on battery op-
rithms. Owing to its simplicity and fast computation, K-means algo- erating limits is studied from the perspective of continuous
rithm is widely used to cluster and analysis large-scale data [13]. At power supply. Micro-grid consumes local renewable energy
present, the new clustering algorithm is based on the K-means as much as possible and operate with maximum renewable
method to expand and improve. The K-means algorithm divides n energy consumption to reduce the amount of abandoned light
objects into k clusters with k as the parameters, which makes the
data in the cluster have high similarity degree and the inter-cluster ELOAD + ESB
data has low similarity. max RPV = ∗100% (5)
EPV
Assuming that the original data set is D (x1 , x2 , . . . , xn ), each xi
is a d-dimensional vector, the purpose of K-means clustering is to
divide
the original data into K class given in advance, In the formula, RPV is the self-absorption rate of PV, EPV is
S = S1 , S2 , . . . , Sk . The following formula is used to calculate the total daily PV power generation, ELOAD is the daily con-
the minimum in the numerical model: sumption of PV power by the load, ESB is the energy storage
battery consumption of PV power.
k
2
(ii) At the same time, it is necessary to control investment costs and
arg min xj − mi (4) operating maintenance costs of energy storage equipment in the
S i=1 xj [Si economic point of view
J. Eng., 2017, Vol. 2017, Iss. 13, pp. 859–864 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0452 Attribution License (http://creativecommons.org/licenses/by/3.0/)
where Smin is the minimum lithium battery capacity, Smax is the than the load power
maximum capacity of lithium battery capacity.
(iii) Charge and discharge power constraints
288 5
S= PPVi − Ploadi × (13)
The charge and discharge power of battery is limited to i[p1
60
prolong the service life of lithium battery, generally related
to the total capacity of lithium batteries. The battery charge
and discharge power PSB (t) at any time must satisfy the limi- (iv) Calculate the amount of electricity difference S ′ that PV output
tation of the following equation: power less than load power. The set p2 is used to represent all
sets of sampling points where the PV output power is less than
the load power
PSBmin ≤ PSB (t) ≤ PSBmax (9)
288 5
S′ = Ploadi − PPVi × (14)
where PSBmin is the lithium battery charging power lower i[p2
60
bound, PSBmax is the lithium battery discharge power limit
constraints.
(iv) Maximum charge rate constraint [10]: (v) Determine the value of S and S ′ ,when S ≥ S ′ , the storage cap-
acity is calculated as S, that is, energy needs to be fully stored
at the peak of the PV power generation. The capacity of
− PSB (t ) ≤ Smax − S (t ) aC (10) storage battery is calculated as follows:
4.2.4 Calculation processFirst, the interactive power optimisation when S , S ′ , the storage capacity is calculated as S ′ , that is, it
target is transformed into the balance between storage capacity is necessary to provide sufficient energy to the load without
and PV load capacity. Secondly, the energy storage cost optimisa- the maximum number of sunny days. The capacity of
tion target is transformed into K-means clustering algorithm to storage battery is calculated as follows:
obtain the statistic minimum capacity cost.
S′ × D × T × g
The basic calculation steps for the optimal capacity are as follows C2 = (16)
(Fig. 3): h×a×U
In (13) and (14), D is the maximum number of days without
(i) Process PV daily output power PPV data, assuming that there sunshine, T represents the temperature correction factor
are 288 sampling points a day (every 5 min a sampling point) (when the temperature 0°C ≤ t, T = 1; when the temperature
t ≤ 0°C, T = 1.1), g is a reliable coefficient, h represents the
288 depth of discharge, a represents the decay rate, U represents
5
SPV = PPVi × (11) the discharge termination voltage.
i=1
60 (vi) Obtained storage battery capacity through the introduction of
reliable coefficient K
where i is the sampling point, PPVi is the ith sampling point PV
power. C′ = K · C (17)
(ii) Process daily load power Pload data, assuming that there are
288 sampling points a day (every 5 min a sampling point)
5 Case study
288
5
Sload = Ploadi × (12) 5.1 Raw data description
i=1
60
In this paper, the distributed PV power and load in DC micro-grid
are calculated according to the output power of a PV power plant in
where i is the sampling point, Ploadi is the ith sampling point the table and the electricity consumption curve of the station.
load power. K-means algorithm is used to calculate the historical data of four
(iii) Calculate the amount of electricity difference S that PV output seasons in PV power station within 1 year. First, randomly selected
power more than load power. The set p1 is used to represent all 1 day at each season as an initial value. After three iterations, the
sets of sampling points where the PV output power is greater four centre data are obtained, that is, as a representative 4 days
PV output power and load power curve shown in Fig. 4. It can be
seen from Fig. 4, installed capacity of PV station is 100 kW.
Features include PV centralised output power between 6:00 and
18:00; summer sunshine time is long, winter sunshine time is
short; the maximum output power appears in 12:00–14:00, up to
98 kW in summer, up to 57 kW in winter.
This is an open access article published by the IET under the Creative Commons J. Eng., 2017, Vol. 2017, Iss. 13, pp. 859–864
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0452
value that power consumption is lower than electricity consump-
tion. If the monthly electricity is surplus, there will be no need to
configure the battery capacity, the algorithm cannot match the
load changes fluctuations at each time. The minimum capacity
method based on battery model is 14 935 kWh, which is greater
than the value determined by the first two methods. The
minimum capacity method simply studies the lower limit of the
storage capacity based on the battery operation limit from the per-
spective of continuous power supply, with the storage capacity
minimisation as the objective function. However, this method
requires a very large battery capacity at the expense of cost. The
estimated cost of the battery accounts for about 80% of the total
cost of micro-grid system.
In summary, different algorithms will give different energy
storage battery configuration recommendations, the specific appli-
cation need to combine micro-grid system size, capital data, eco-
nomic and technical requirements to determine. For the purposes
of this study, it is finally recommended to use the battery capacity
based on the K-means algorithm of the minimum capacity method.
J. Eng., 2017, Vol. 2017, Iss. 13, pp. 859–864 This is an open access article published by the IET under the Creative Commons
doi: 10.1049/joe.2017.0452 Attribution License (http://creativecommons.org/licenses/by/3.0/)
7 Acknowledgement [6] Xiu X., Li J., Hui D.: ‘Capacity configuration and economic evalu-
ation of energy storage system for grid peak load shifting’, Electr.
This research was founded by the Science and Technology Project Power Constr., 2013, 34, pp. 1–5
of SGCC (Research on Network Architecture and Key [7] Xie S., Yang L., Li L.: ‘A chance constrained programming based
Technologies of DC distribution Network) (PDB 17201600116). optimal configuration method of hybrid energy storage system’,
Power Syst. Technol., 2012, 36, pp. 79–84
[8] Niu M., Huang W., Guo J., ET AL.: ‘Research on economic operation
8 References of grid-connected microgrid’, Power Syst. Technol., 2010, 34, pp.
38–42
[1] Zang H.: ‘Key basic researches of photovoltaic planning with prac- [9] Ai H.: ‘Control strategy of hybrid cascaded power conversion system
tical principle’, Southeast University, 2014 based on battery energy storage for PV power plant’, Beijing Jiaotong
[2] Zou X.: ‘The research of energy management technology in DC University, 2015
micro-grid’, Southwest Jiaotong University, 2016 [10] Fu J.: ‘Research on photovoltaic power generation system based
[3] Cai Y., Feng J., Zhang S., ET AL.: ‘Allocation method for schedulable on ultracapacitor energy storage’, Southwest Jiaotong University,
micro-grid energy storage capacity considering wind/PV power gen- 2012
eration’, East China Electr. Power, 2014, 42, pp. 1821–1826 [11] Wang W.: ‘Research on optimal allocation of hybrid energy storage
[4] Cui H., Guo Y., Chengjun X.: ‘Research on optimal allocation of dis- system in microgrid’, Southeast University, 2014
tributed generation by considering environmental benefits’, East [12] Li Y.: ‘Research on operation control of DC micro-grid based on
China Electr. Power, 2010, 38, pp. 1968–1971 photovoltaic generation and hybrid energy storage’, Shandong
[5] Zhang J., Yu L., Liu N., ET AL.: ‘Capacity configuration optimization University, 2015
for island microgrid with wind photovoltaic diesel storage and sea- [13] Zhao L., Hou X., Hu J., ET AL.: ‘Improved k-means algorithm based
water desalination load’, Trans. China Electrotech. Soc., 2014, 29, analysis on massive data of intelligent power utilization’, Power
pp. 102–112 Syst. Technol., 2014, 38, pp. 2715–2720
This is an open access article published by the IET under the Creative Commons J. Eng., 2017, Vol. 2017, Iss. 13, pp. 859–864
Attribution License (http://creativecommons.org/licenses/by/3.0/) doi: 10.1049/joe.2017.0452