Optimal scheduling of a battery-based energy storage

system for a microgrid with high penetration of

renewable sources
Jérémy Dulout, Adriana Luna, Amjad Anvari-Moghaddam, Corinne Alonso,
Josep Guerrero, Bruno Jammes

To cite this version:

Jérémy Dulout, Adriana Luna, Amjad Anvari-Moghaddam, Corinne Alonso, Josep Guerrero, et
al.. Optimal scheduling of a battery-based energy storage system for a microgrid with high
penetration of renewable sources. ELECTRIMACS 2017, Jul 2017, Toulouse, France. 6p.,
<http://www.electrimacs2017.fr/>. <hal-01533070>

HAL Id: hal-01533070

https://hal.laas.fr/hal-01533070
Submitted on 6 Jul 2017

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 ELECTRIMACS 2017, 4th -6th July 2017, Toulouse,
France

OPTIMAL SCHEDULING OF A BATTERY-BASED ENERGY STORAGE

SYSTEM FOR A MICROGRID WITH HIGH PENETRATION OF
RENEWABLE SOURCES

Jérémy Dulout1, Adriana Luna2, Amjad Anvari-Moghaddam2,

Corinne Alonso1, Josep M. Guerrero2, Bruno Jammes1
1
LAAS-CNRS, Université de Toulouse, CNRS, UPS, France
2
Department of Energy Technology, Aalborg University, Denmark

e-mail: {jdulout, alonsoc, jammes}@laas.fr {acl, aam, joz}@et.aau.dk

Abstract – A new scheduling method is proposed to manage efficiently the

integration of renewable sources in microgrids (MGs) with energy storage systems
(ESSs). The purpose of this work is to take into account the main stress factors
influencing the ageing mechanisms of a battery energy storage system (BESS) in
order to make an optimal dispatch of resources in the microgrid and enhance the
storage system lifetime while minimizing the cost of electric consumption. The load
demand and generation profiles are derived from the analysis of consumption and
renewable production (solar photovoltaic sources and wind turbines) of the Western
Denmark electric grid. Thus, the proposed microgrid is mainly fed by renewable
sources and few electricity is coming from the main grid (which helps operating costs
minimization). In this respect, a cost analysis is performed to find the optimal hourly
power output of the BESS as well as the purchased electricity from the utility.

Keywords – Battery Management System (BMS), Energy Storage System (ESS),

Microgrid (MG), Optimal scheduling, Wind and solar photovoltaic sources.

1. INTRODUCTION integration, good performances and level of maturity

[2]-[3]. In order to deal with the uncertainties of the
MG, stochastic programming is often used for
In recent years, many efforts have been spent on modelling, planning and control [4]-[6]. Several
tackling the challenge of climate change and stochastic optimization tools for microgrid planning
sustainable access to energy. To this end, a new are presented in [4]. In [5] a stochastic scheduling is
renewable decentralized energy production structure performed on a IEEE 33-bus system (including
has emerged: the microgrid (MG) [1]. A MG is a electric vehicles, wind turbines, decentralized
small-scale system (e.g., aggregation of a few generators and storage systems). In addition, [6]
houses, a university, a military base, a commercial proposes a robust and distributed energy scheduler
area, etc.) operating in a local area. The energy with wind farms, storage systems and different loads
production mainly comes from renewable sources to (some of them are fully controllable and can be time-
meet the local energy needs and the remaining shifted).
demand is fulfilled by the main grid. This system
offers many advantages, such as energy transmission Recent papers propose to optimize the operating
losses reduction, remote areas electrification, conditions of batteries in MGs to enable longer
reliability improvement, easier integration of lifetime and to minimize the cost of operation [7]-
renewable sources and much more. [10]. In this context, this paper aims at gathering
some guidelines for improving the lifetime of
However, due to the intermittent nature of renewable batteries in order to reduce the cost of the battery
sources and presence of uncertainties (e.g., energy storage system (BESS) on the long run. The
renewable power generation, load demand, cost of presented methodology defines the optimal power
electricity, cost of storage systems, etc.), the level for the BESS and for the grid (purchased
integration of energy storage systems (ESSs) in electricity) on an hourly basis, in a way to satisfy the
renewable MGs is a widespread practice. Batteries load demand and to minimize the cost of electricity
are often preferred to other ESSs in microgrid and the BESS cost (along its lifetime).
projects because of their relatively lower price, easy

1
 ELECTRIMACS 2017, 4th -6th July 2017, Toulouse,
France

The renewable generation and load demand profiles 2. ENERGY MANAGEMENT FORMULATION
of the MG are derived from the analysis of West
Denmark electric market grid data [11].
2.1. MICROGRID PRESENTATION
The remainder of the paper is organized as follows:
first, the formulation of the MG model (mainly In this paper, the MG operator is responsible for the
focused on defining proper BESS operating BESS management and scheduling of the energy
conditions) and the mission profiles (load demand exchanges with the utility in order to satisfy the
and renewable production) are presented. Then, the stability of the MG and to minimize the overall costs.
minimization problem is explained. Finally, the The examined MG architecture as well as the power
results of the proposed methodology are discussed. flows are described in fig. 1.

NOMENCLATURE

Parameters
BESS parameters of the
𝐴, 𝐵, 𝐶
lifetime model
𝐶𝑎𝑝𝑛𝑜𝑚 Nominal capacity of the BESS
Cost of electricity from the
𝐶𝑔𝑟𝑖𝑑 (𝑡)
main grid
𝐶1𝑘𝑊ℎ (𝑡) Cost of 1kWh of BESS
BESS maximum current-rate
𝐶𝑟𝑎𝑡𝑒𝐶ℎ , 𝐶𝑟𝑎𝑡𝑒𝐷𝑖𝑠
at the charge and discharge
∆𝑡 Time interval (1-hour)
Fig. 1. Typical architecture of a MG.
Charge and discharge energy
𝜂𝐶ℎ , 𝜂𝐷𝑖𝑠
efficiency of the BESS In order to ensure the MG stability, the following
Initial/Final daily state of equation must be satisfied at any time:
𝑆𝑂𝐶𝑖𝑛𝑖 , 𝑆𝑂𝐶𝑒𝑛𝑑
charge (SOC)
Minimum/Maximum daily 𝑃𝐵𝑎𝑙𝑎𝑛𝑐𝑒 (𝑡) + 𝑃𝑠𝑡𝑜𝑟𝑎𝑔𝑒 (𝑡) + 𝑃𝑔𝑟𝑖𝑑 (𝑡) = 0 (1)
𝑆𝑂𝐶𝑚𝑖𝑛 , 𝑆𝑂𝐶𝑚𝑎𝑥
SOC
where 𝑃𝐵𝑎𝑙𝑎𝑛𝑐𝑒 (𝑡) is the power balance between
𝑖 Interest rate of the project
renewable production and load demand (2),
Power balance between 𝑃𝑠𝑡𝑜𝑟𝑎𝑔𝑒 (𝑡) is the power absorbed or delivered by the
𝑃𝐵𝑎𝑙𝑎𝑛𝑐𝑒 (𝑡) production of renewables and BESS (3) and 𝑃𝑔𝑟𝑖𝑑 (𝑡) is the power exchanged with
consumption the main utility grid. Any injected power to the bus
Maximum power that can be is positive whereas the power taken out of the bus is
𝑃𝑔𝑟𝑖𝑑𝑀𝑖𝑛 , 𝑃𝑔𝑟𝑖𝑑𝑀𝑎𝑥
sold/purchased from the utility treated as a negative value.

Variables 𝑃𝐵𝑎𝑙𝑎𝑛𝑐𝑒 (𝑡) = 𝑃𝑃𝑟𝑜𝑑 (𝑡) − 𝑃𝐶𝑜𝑛𝑠 (𝑡) (2)

Total cost (according to energy

𝐶𝑇𝑜𝑡𝑎𝑙 𝑃𝑠𝑡𝑜𝑟𝑎𝑔𝑒 (𝑡) = 𝑃𝐷𝑖𝑠 (𝑡) − 𝑃𝐶ℎ (𝑡) (3)
consumption and BESS costs)
Annual cost of electric 𝑃𝑔𝑟𝑖𝑑 (𝑡) = 𝑃𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒 (𝑡) − 𝑃𝑠𝑒𝑙𝑙 (𝑡)
𝐶𝐸𝑙𝑒𝑐 (4)
consumption from the utility
𝐶𝐵𝐸𝑆𝑆 Annual cost of the BESS where 𝑃𝑃𝑟𝑜𝑑 , 𝑃𝐶𝑜𝑛𝑠 , 𝑃𝑐ℎ (𝑡), 𝑃𝑑𝑖𝑠 (𝑡), 𝑃𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒 (𝑡) and
𝐶𝑅𝐹 Capital recovery factor 𝑃𝑠𝑒𝑙𝑙 (𝑡) are the renewable production and
consumption of the MG, the charging and
BESS delta depth of discharge – discharging power of the BESS, the power
∆𝐷𝑂𝐷
Cycle depth (per day)
purchased and sold to the utility, respectively. All
𝑃𝑔𝑟𝑖𝑑 (𝑡) Power exchanged with the utility these variables are positive.
𝑃𝑠𝑡𝑜𝑟𝑎𝑔𝑒 (𝑡) Output power of the BESS
2.2. STUDY CASE – DANISH ELECTRICITY PROFILE
𝑆𝑂𝐶(t) State of charge of the BESS
𝑛 Lifetime of the BESS (years) In this study, the last three years of the Western
Denmark electric grid data [11] have been used to

2
 ELECTRIMACS 2017, 4th -6th July 2017, Toulouse,
France

define the load and renewable production profiles trip energy efficiency of 92% is considered for
presented in fig. 2. Two different profiles of lithium batteries (𝜂𝐶ℎ = 0.92 and 𝜂𝐷𝑖𝑠 = 1). The
𝑃𝐵𝑎𝑙𝑎𝑛𝑐𝑒 (𝑡) (mean and median profiles of the data BESS SOC have to be limited as described in (6) and
set) have been obtained by taking into account the (7). It has been shown in [15] that a mid-range SOC
whole generation of wind turbines and photovoltaic level and a low cycle depth are guaranteeing the
panels and half of the energy gross consumption of highest lifetime of lithium batteries (so it has been
West Denmark (in order to get a daily profile that is chosen to use the following values 𝑆𝑂𝐶𝑖𝑛𝑖 = 50%,
producing more energy than consuming (mean 𝑆𝑂𝐶𝑚𝑖𝑛 = 35%, 𝑆𝑂𝐶𝑚𝑎𝑥 = 65%).
profile), and conversely (median profile)).
𝑃𝐶ℎ (𝑡) ∙ Δ𝑡 𝑃𝐷𝑖𝑠 (𝑡) ∙ Δ𝑡 (5)
𝑆𝑂𝐶(𝑡) = 𝑆𝑂𝐶(𝑡 − 1) + 𝜂𝐶ℎ −
𝐶𝑎𝑝𝑛𝑜𝑚 𝜂𝐷𝑖𝑠 ∙ 𝐶𝑎𝑝𝑛𝑜𝑚

𝑆𝑂𝐶𝑚𝑖𝑛 ≤ 𝑆𝑂𝐶(𝑡) ≤ 𝑆𝑂𝐶𝑚𝑎𝑥 (6)

𝑆𝑂𝐶𝑖𝑛𝑖 = 𝑆𝑂𝐶𝑒𝑛𝑑 (7)

As advised by manufacturers, the charge and

discharge current-rates (expressed in [h-1]) have to be
limited for lifetime and energy efficiency reasons.
Moreover, the most efficient BESS charging method
is a constant current charge followed by a constant
voltage charge (often referred “CC-CV charge”). It
implies a power capability limitation of the BESS
when a high SOC level is reached (typically around
Fig. 2. Hourly median and mean 𝑃𝐵𝑎𝑙𝑎𝑛𝑐𝑒 (𝑡)
80% for lithium batteries). (8) and (9) ensure the
obtained from the considered West Denmark data.
power capability limitations of the BESS during
charge and discharge.
In the same way, the hourly price of electricity
evolution from the Nord Pool Elbas intraday market
0 ≤ 𝑃𝑑𝑖𝑠 (𝑡) ≤ 𝐶𝑟𝑎𝑡𝑒𝐷𝑖𝑠 𝐶𝑎𝑝𝑛𝑜𝑚 (8)
is presented in fig. 3. In this study, the mean value
over the three last years (circle markers in fig. 3),
0 ≤ 𝑃𝑐ℎ (𝑡) ≤ 𝐶𝑟𝑎𝑡𝑒𝐶ℎ 𝐶𝑎𝑝𝑛𝑜𝑚 (9)
𝐶𝑔𝑟𝑖𝑑 (𝑡), has been chosen as the hourly price of {
0 ≤ 𝑃𝑐ℎ (𝑡) ≤ 𝐶𝑟𝑎𝑡𝑒𝐶ℎ 𝐶𝑎𝑝𝑛𝑜𝑚 Δ𝑡 𝐷1 𝑒 (𝐷2 𝑆𝑂𝐶)
electricity from the main grid.
Indeed, when the BESS reaches 𝑆𝑂𝐶𝐶𝑉 the power that
can be absorbed is limited by a decreasing
exponential law [8], [13], as described in fig. 4 (𝐷1
and 𝐷2 are obtained by curve fitting of data from
manufacturers).

Fig. 3. West Danish hourly price of electricity [11].

Fig. 4. CC-CV charging method of the BESS.

2.3. ENERGY STORAGE SYSTEM MODEL The lifetime of the BESS is quantified by a capacity
and power capability fade. It is commonly accepted
Worldwide, several technologies of ESSs are that a battery is considered at its end of life when its
available, such as pumped hydroelectric storage, fuel nominal capacity reaches 80% of its initial capacity.
cells, flywheels, compressed air energy storage [3]. The lifetime of a BESS relies on two mechanisms,
In our context, electrochemical storage (more calendar and cycle ageing [13]-[15]. According to
precisely lithium batteries) is the most appropriate [14] and [15], cycle lifetime is always lower than
way to store energy. The following section describes calendar lifetime (shelf life). Considering that the
how to model a BESS based on [5], [12]-[15]. First, temperature of the BESS room will be regulated, the
the definition of the state of charge (SOC) is given in cycle depth and the SOC-level are the two main
(5). As explained in a previous work [12], a round stress factors influencing the cycling lifetime.

3
 ELECTRIMACS 2017, 4th -6th July 2017, Toulouse,
France

Performing low cycle depths will enhance the BESS

𝐶𝐸𝑙𝑒𝑐 = 365.25 ∑ 𝐶𝑔𝑟𝑖𝑑 (𝑡) × 𝑃𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒 (𝑡) × Δ𝑡 (14)
number of cycles [13]-[15], as depicted in fig. 5.

𝐶𝐵𝐸𝑆𝑆 = 𝐶1𝑘𝑊ℎ × 𝐶𝑎𝑝𝑛𝑜𝑚 × CRF (15)

A cost of 350€/kWh has been considered for lithium

batteries [18]. The capital recovery factor, CRF,
converts the initial investment into annual equally
separate payments over a given period:

𝑖(1 + 𝑖)𝑛 (16)

𝐶𝑅𝐹 =
(1 + 𝑖)𝑛 − 1

where 𝑖 is the interest rate (7.7% for this type of

Fig. 5. Number of equivalent full cycles vs. cycle
project [12]) and 𝑛 is the BESS lifetime in years.
depth of the BESS.
A last constraint concerning the main grid utilization
The degradation of the battery can be sensed with the
should be considered as follows:
rainflow counting method by following the SOC
level and counting the number of cycles with their
𝑃𝑔𝑟𝑖𝑑𝑀𝑖𝑛 ≤ 𝑃𝑔𝑟𝑖𝑑 (𝑡) ≤ 𝑃𝑔𝑟𝑖𝑑𝑀𝑎𝑥 (17)
respective cycle depth, as described in [16]. In order
to limit the number of equations while keeping a
good level of accuracy, it has been chosen to model 3. RESULTS
the lifetime of the battery with the Ah-throughput
method described in [17].
The proposed non-linear optimization problem
In this study, the cycle depth (difference between the based on equations (1)-(17) has been solved with the
maximum and minimum SOC performed in a cycle) General Algebraic Modeling System (GAMS)
has been fixed to 30% as in [12]. The maximum software. For this simulation, a 1.5MWh BESS has
energy that the BESS can exchange is: been used and the maximum power that can be
exchanged with the utility has been fixed to 200 kW.
𝐸𝐸𝑥𝑐ℎ𝑀𝑎𝑥 = 𝐴 𝑒 −100.𝐵.∆𝐷𝑂𝐷+𝐶 × 𝐶𝑎𝑝𝑛𝑜𝑚 × ∆𝐷𝑂𝐷 (10)
The results of the scheduling method are presented
where A, B and C are constants obtained by curve in fig. 6. Due to the power exchanges limitations
fitting of experimental lifetime data from [15] with the main grid that have been set, the BESS
(A=1.057e4, B=0.05459, C=455 with R²=0.9729 enables to reduce peak demand from 8:00 to 12:00
and RMSE=689.1). and from 19:00 to 22:00 for the median profile
(fig 6-A). It can be also noticed that due to the price
The exchanged energy per day is calculated as of electricity (fig. 3), the BESS is charging most of
follows: all when the tariffs are relatively low. As expected,
the BESS satisfies the constraints of SOC, cycle
(11)
depth and C-rate. Regarding the profile described in
𝐸𝐸𝑥𝑐ℎ𝐷𝑎𝑦 = ∑(𝑃𝐷𝑖𝑠 (𝑡) + 𝑃𝐶ℎ (𝑡)) × ∆𝑡
fig 6-B, the minimization of the BESS annual cost
𝑡
implies to minimize the energy exchanges between
Finally, BESS lifetime (in years) is given by: the MG and the BESS.

𝐸𝐸𝑥𝑐ℎ𝑀𝑎𝑥 (12) For both profiles, the annual cost of electricity is by

𝑛=
365.25 𝐸𝐸𝑥𝑐ℎ𝐷𝑎𝑦 far lower than the annual cost of the BESS. This is
mainly due to the capacity and the cycle depth
chosen in this example, based on a previous work on
2.4. OPTIMIZATION PROBLEM optimal sizing of a lithium battery pack [12].
Obviously, when the capacity of the BESS and the
In this study, the objective function (13) is a trade- maximum daily cycle depth are changing, the
off between the need to use intensively the BESS to optimal charge/discharge profile will be modified.
decrease the annual cost of the electricity (14) and This highlights the necessity to analyze together the
the need to improve the BESS lifetime (i.e. to reduce sizing and scheduling problems in order to get the
its annual cost) (15). best from the BESS. Indeed, in this case, it seems
that the actual sizing does not lead to the best
Min {𝐶𝑇𝑜𝑡𝑎𝑙 = 𝐶𝐸𝑙𝑒𝑐 + 𝐶𝐵𝐸𝑆𝑆 } (13) scheduling that can be achieved.

4
 ELECTRIMACS 2017, 4th -6th July 2017, Toulouse,
France

Fig. 6. MG scheduling for a 1.5 MWh BESS A) Median profile, B) Mean profile.

[4] Hao Liang, Weihua Zhuang, “Stochastic

4. CONCLUSION Modeling and Optimization in a Microgrid: A
Survey,” Energies 2014, 7, 2027-2050.
A methodology for optimal scheduling of a
microgrid (MG) supplemented by a battery energy [5] Amin Gholami, Tohid Shekari, Farrokh
storage system (BESS) was described. The BESS Aminifar, Mohammad Shahidehpour,
behaviour and lifetime models have been discussed “Microgrid Scheduling With Uncertainty: The
and implemented in order to minimize the annual Quest for Resilience,” IEEE Trans. on Smart
cost of electricity and the annual cost of the BESS Grid, vol. 7, no. 6, pp. 2849-2858, Nov. 2016.
considering operating limitations (e.g. cycle depth,
SOC level, C-rate, etc.). [6] Yu Zhang, Nikolaos Gatsis, Georgios B.
Giannakis, “Robust Energy Management for
However, it has been seen that the annual cost of the Microgrids With High-Penetration
BESS is far more expensive than the annual cost of Renewables,” IEEE Trans. on Sustainable
electricity. This is mainly due to the fact that the Energy, vol. 4, no. 4, pp. 943-953, Oct. 2013.
sizing of the BESS can be improved by changing the
nominal capacity and maximum daily cycle depth. [7] Guannan He & al., “Optimal Bidding Strategy
Further research will be conducted in this field order of Battery Storage in Power Markets
to reduce the cost of stored electricity. Furthermore, Considering Performance-Based Regulation and
the BESS is a valuable resource that enables, among Battery Cycle Life,” IEEE Transactions on
other things, the flattening of peak demand which Smart Grid, vol. 7, no. 5, pp. 2359 – 2367, Sept.
may help to reduce significant costs in 2016.
infrastructures, environmental impact and risks of
outages. [8] Jen-Hao Teng, Shang-Wen Luan, Dong-Jing
Lee, Yong-Qing Huang, “Optimal
Charging/Discharging Scheduling of Battery
REFERENCES Storage Systems for Distribution Systems
Interconnected With Sizeable PV Generation
[1] Hassan Farhangi, “The path of the smart grid,” Systems,” IEEE Transactions on Power
IEEE Power and Energy Magazine, vol. 8, no. 1, Systems, vol. 28, no. 2, pp. 1425 – 1433, May
pp. 18 – 28, January-February 2010. 2013.

[2] IRENA report, “Battery storage for renewables: [9] Khalid Abdulla & al., “Optimal Operation of
market status and technology outlook,” January Energy Storage Systems Considering Forecasts
2015. and Battery Degradation,” IEEE Transactions
on Smart Grid, DOI
[3] Xing Luo, Jihong Wang, Mark Dooner, 10.1109/TSG.2016.2606490.
Jonathan Clarke, “Overview of current
development in electrical energy storage [10] Inderjeet Duggal, Bala Venkatesh, “Short-term
technologies and the application potential in scheduling of thermal generators and battery
power system operation,” Applied Energy vol. storage with depth of discharge-based cost
137, 1 January 2015, Pages 511–536.

5
 ELECTRIMACS 2017, 4th -6th July 2017, Toulouse,
France

model,” IEEE Transactions on Power Systems, Profile,” Proceedings of the 2015 IEEE Energy
vol. 30, no. 4, pp. 2110 – 2118, July 2015. Conversion Congress and Exposition (ECCE),
pp. 14 -21.
[11] energinet.dk, “Download of market data
(Danish electricity grid),” [Online]. Available: [15] Madeleine Ecker & al., “Calendar and cycle life
http://energinet.dk/EN/El/Engrosmarked/Udtra study of Li(NiMnCo)O2-based 18650 lithium
ek-af-markedsdata/Sider/default.aspx ion batteries,” Journal of Power Sources 248
[Accessed: 20 December 2016]. (2014) 839-851.

[12] Jérémy Dulout, Amjad Anvari-Moghaddam, [16] H. Beltran & al., “Lithium ion batteries ageing
Adriana Luna, Bruno Jammes, Corinne Alonso, analysis when used in a PV power plant,” IEEE
Josep M. Guerrero, “Optimal sizing of a lithium International Symposium on Industrial
battery energy storage system for grid- Electronics (ISIE), 2012, DOI:
connected photovoltaic systems,” International 10.1109/ISIE.2012.6237330.
Conference on DC Microgrids (ICDCM), June
2017. [17] Henrik Bindner et al., “Lifetime Modelling of
Lead Acid Batteries,” Risø-R-1515 April 2005,
[13] D. Linden, T.B Reddy, “Handbook of batteries, ISBN 87-550-3441-1.
McGraw-Hill Education,” Third edition, 2001.
[18] Björn Nykvist, Måns Nilsson, “Rapidly falling
[14] Daniel-Ioan Stroe & al., “Degradation costs of battery packs for electric vehicles,”
Behaviour of Lithium-Ion Batteries based on Nature Climate Change 5, 329–332 (2015).
Field Measured Frequency Regulation Mission