You are on page 1of 5

An Intelligent Battery Management System for Home

Microgrids
Christopher Lebrón, Fabio Andrade, Efraín O’Neill, Agustín Irizarry
Department of Electrical and Computer Engineering
University of Puerto Rico, Mayaguez Campus
Mayaguez, Puerto Rico
christopher.lebron@upr.edu; fabio.andrade@upr.edu; efrain.oneill@upr.edu; agustin@ece.uprm.edu

Abstract— This paper presents an intelligent battery for the development of photovoltaic (PV) systems [6]. The
management system (iBMS) used in a grid-connected Microgrid. proposed microgrid in figure 1 will be a residential type
The microgrid is composed by a PV generator, Battery system composed of three homes. The microgrid will be grid
and a small residential group. This microgrid is located in a connected and it will integrate a photovoltaic system with
Caribbean zone which an average solar radiation greater than batteries for energy storage. Demand Response (DR) system is
1000 W/m2 but with large intermittency due to the movement of added because it can reduce the home power demand when the
clouds. The iBMS considers a fuzzy controller whose rules are batteries or the grid are under stress [7]. An intelligent battery
defined by the behavior of the people living in the houses, the management is needed in order to not generate disturbances to
intermittent energy available from the PV generator, the state of
the electrical grid and to optimize the battery state of charge.
charge (SOC) of the battery and the contracted power from the
electric mains. Also the iBMS considered a block of demand
respond (DR) from a non-critical electric loads available in
houses. The mail objectives of the iBMS is to supply electric
power to houses using the available energy, not to generate
disturbance to the electric mains and always to maintain the
battery SOC in the possible maximum.

Index Terms—Battery Management System, Demand Response,


Fuzzy Controller, Microgrid, Photovoltaic Generator.

I. INTRODUCTION
Microgrids are the systems that integrate distributed
generation (DG) units, energy storage systems (ESS) and
controllable loads on a low voltage network which can operate
in either grid-connected mode or stand-alone mode [1]. In the
grid-connected mode, the microgrid adjusts power balance of Figure 1: Home Microgrid working in grid-connected mode
supply and demand by purchasing power from the main grid
or selling power to the main grid [2]. The intermittency of
distributed renewable generation poses significant challenges Battery management systems (BMS) are vital in the use of
for the operation and integration of microgrids [3]. To microgrids. Over the past decade researchers have developed
alleviate the power fluctuations of non-dispatchable DG units, different strategies for battery management systems. One of
various control schemes are used in microgrids, including these strategy is the use of fuzzy logic (FL)-based expert
power regulation of each dispatchable DG unit, charging and system applied for microgrid battery scheduling. Indeed, fuzzy
discharging of ESS, and load shedding [4]. set theory can encompass such subjective decision-making
The tropical climate in the Caribbean zone facilitates an process due to its ability to define human reasoning that can
abundant supply on solar renewable energy resource [5]. The handle uncertainties regarding to the microgrid exogenous
case study is located in an island named Puerto Rico (PR) environment and the uncertainty of the forecasted parameters.
located in the Caribbean. In some regions in PR the solar The use of fuzzy logic helped improved the fuzzy expert
irradiation exceeds the 1000 W/m2 making this resource ideal system for battery scheduling decreased considerably the

978-1-5090-5167-0/16/$31.00 ©2016 IEEE


battery maintenance cost by extending the operation lifetime II. GENERAL OVERVIEW
through lower depth of battery discharge [8]. Another strategy A PV generator, Battery system and a small residential
is the SOC-based management system, in this technic consists group compose the microgrid. The microgrid normally
of a hierarchical control system containing three modules: operates in a grid-connected mode, as power can be imported
SOC computation module, battery mode identification system or exported from and to the main grid.
and closed-loop feedback mode identification system [9]. In
addition, a common strategy for BMS is to use an inverter as A. PV Array Model
an energy management system, the energy from the dc-link
can be taken into a dc–ac inverter to feed the energy in the The behavior of a photovoltaic (PV) module is largely well
grid or the household. Alternatively, this energy can be stored known. It works inside a limited voltage and current values
into the battery via a second dc–dc converter, called battery following an exponential I-V curve. This curve varies with
converter. The battery converter consists of a synchronous irradiance and temperature. Manufacturers of it usually
buck converter with two phases and a balancing unit [10]. provide the main parameters referred to reference values of
irradiance (W) and temperature (T) (Stan dard Test
The objective of this paper is to develop control strategy Conditions, STC: 1000 W/m2 and 25ºC, respectively). A
for a battery system to improve operation of a home simple-to-use PV model, proposed in [12], is used to fulfil of
microgrid. The iBMS proposed in the paper try to maintain a accurately prediction of PV characteristics (even changing
flat consumption of the mains and to compensate for the irradiance and temperature), make possible building
intermittent nature of the PV generator using the storage series/parallel module associations, configure partial
system and the demand response function. On the other hand, shadowed scenarios and just to require parameters provided by
the BMS must consider the SOC of the battery to prevent manufacturers in typical datasheets.
significant damages due to excessive charge or discharge. A
battery model appropriate for microgrid application is derived The mathematical model used, firstly the I-V curve in STC
base on the cell model presented in [11]. A full model for a is modelled and then the influence of irradiance and
microgrid, with a storage system, a PV generator and temperature is taken into account.
residential consumptions is built in MATLAB/SIMULNK and The used irradiance and temperature data is based on data
simulations are carried in a dSPACE system using the concept collected in southern Puerto Rico. These data show a highly
of Hardware in the Loop (HIL). random behavior and it has a maximum irradiance of
2000W/m2 typically between 13h and 15h. The PV system has
an array of 640 solar cells in series and 8 in parallel, fig. 2.
This generator uses the programmed MPPT algorithm. This
algorithm is combined with the model as a control loop,
starting from the measured Ipv and Vpv values, the PI controller
is designed to produce a reference current value Ipvref to be
used as an input value for the PV module model.

B. Battery Model
The used model, proposed in [11], is based on a Thévenin
circuit with a RC parallel circuit and includes a state of charge
estimator. The model simulates hysteresis and sub-hysteresis
effects. Also, this model considers variations in the internal
and external variables by modifying the values of its internal
parameters.
The SOC that is defined as the quantity of charge
remaining in the battery, is expressed in terms of percentage
with respect to the nominal capacity Cn.
The battery has a stack of 210 Ni-Cd cells (240Vdc) and
22Ah of capacity.

C. VSI Converters
The prime energy of this generator, which can be solar or
batteries, is transferred as AC power through a VSI converter
that uses an intelligent control system. The control system
uses a PLL block and inner control loops of current and
voltage to ensure the respective reference signals. The output
Figure 2: PV Array Model
current and voltage reference signals are configured by an
outer control loop of sharing power which uses the basic idea
of the synchronous generator concept to control the sharing
power by using droop curves (“P vs. f” and “Q vs. V”). Fig 3.

978-1-5090-5167-0/16/$31.00 ©2016 IEEE


D. Residential Model
The small residential group data was simulated from three
homes located in Puerto Rico through Matlab/Simulink. There
were three households in this case study. For a typical
consumption pattern of residential groups in Puerto Rico has
its high demands during the night, this is because work and
school hours are mostly during the day, from 8:00 am to 5pm
and most of the family members are not at home during this
period of time. Some significant appliances that are in an
average home in PR are: television, A/C Unit, fridge, oven,
coffee maker, washing and draying machine and illumination.
The peak demand for these three homes is approximately 20
kW.

Intelligent Battery Energy DR


SOC DR
Management System DR
Pbat Pgrid Figure 4: Fuzzy Logic Membership input functions

Ppv (-)A, Medium High level (-)MA, Medium level (-)M, Low
level (-)B. The levels for discharging the baterry are at Low
Pref2 Pref1 level (+)B, Medium level (+)M, Medium High level (+)MA
ω0

Kp LPF vα iα − vβ iβ
and High level (+)A. The outputs of the FLC are: three
Reference
Generator
ω0)
E = E0 − kq ( Q − Q LPF v β iα − vα iβ
different levels of demand response and access to the electrical
𝐸=sin(𝜔𝑡)
grid for when is needed.
0

Droop Curves Power Calculation


Reference Kp LPF vα iα − vβ iβ
Generator
v β iα − vα iβ
𝐸=sin(𝜔𝑡G) (s) E = E0 − kqG
c
( Q − Q0 )
Droop Curves
v (s)
LPF

Power Calculation
!"#$ = !&-!(-!)*+, (1)
SVM
VdcIdc
abc/αβ

abc/αβ

abc/αβ

S1-S2-S3 S1'-S2'-S3'
Gc(s) Gv(s)

+ S1 S2 S3 SVM
The FLC uses ecuation (1) to maintain the power flow
balance in the microgrid. However, the baterry has a
VdcIdc a Ial Vac Ia0
abc/αβ

abc/αβ

abc/αβ

b
S1-S2-S3 Ib
S1'-S2'-S3'
l Vbc Ib0
Vdc
Cdc c Icl Vcc Ic0
R

S1'
+ S2'
S1
S3'
S2
Li Ri
S3
L0 R0 restriction that it requiers the baterry’s satate of charge
between the values of 20% and 80%.
a Ial Vac Ia0
n b Ibl Vbc Ib0
Vdc
Cdc c Icl Vcc Ic0
R CL
i R L0 R0
i i
S1' S2' S3'

n
Ci 20 % < &'( < 80% (2)
Figure 3: iBMS for the Home Microgrid The FLC optimizes the battery’s SOC by operating in
various levels of demand response, once the batery starts to
produce power and the SOC begins to decrease the controller
III. AN INTELLIGENT BATTERY MANAGEMENT SYSTEM activates the different levels of Demand Response. If the SOC
The management system proposed in this paper is based on is at medium charged it activates the first level of DR. Each
a fuzzy logic controller (FLC). This controller is developed time the SOC level drops it activates another DR level option
within the fuzzy tool box of Matlab/Simulink. The objective until the higher level is reach and the system has no choice to
of the the FLC is to optimize the energy produced by the require additional power from the electrical grid.
photovoltaic generators and maintaining a constant flow from
IV. SIMULATIONS RESULTS
the electrical grid. The FLC in figure 4 has two inputs
function: the battery state of charge (SOC) and the real power Simulations based on Hardware in the loop (HIL), by
of the battery (Pbat). using a dSPACE ds1006, MATLAB/Simulink and dSPACE
Control Desk, are performed to evaluate the proposed method
In the Soc input function there are different leves of for study the large-signal stability. In the case of study, an
representations: Low Soc (B), Medium Low (MB), Medium inverter-based generator is connected to a Microgrid working
(M), Medium High (MA) and High (A). For the Pbat input in island-mode. Electrical setup and control systems parameter
function has 2 components when the battery is charging and are listed in table I.
discharging. The leves of charging the baterry are at high level

978-1-5090-5167-0/16/$31.00 ©2016 IEEE


Figure 5: Simulation Results with Fuzzy Logic Controller

978-1-5090-5167-0/16/$31.00 ©2016 IEEE


The simulation results are shown in figure 5, in the last dynamic behavior of the microgrid is examined. The
graph is represented the time period of the simulation that are simulation results demonstrate the effectiveness of the control
seven days. Each day is represented with the ramp plot with strategies through the fuzzy logic.
values from 0 to 24 hours. The Three Household Power
Consumption graph represents the consumption of houses in a ACKNOWLEDGMENT
24-hour period. The values of power consumption of each day
varies randomly maintaining a pattern of actual consumption. This project is funded in part by NSF, under grant # ACI-
In this graph it can be observed two mayor peaks: one is 1541106. Any opinions, findings, and conclusions or
during the morning (7 hour) approximately with a power peak recommendations expressed in this material are those of the
of 10 kW, this happens when most of the families wake up to author(s) and do not necessarily reflect the views of the
work and go to school. In the other hand, the next peak National Science Foundation.
happens at the evening (15 hour) and the peak power is
approximately 15 kW, this is when the family’s returns to REFERENCES
their houses and they start to use most of their appliances. [1] F. Katiraei, R. Iravani, N. Hatziargyriou and A. Dimeas, "Microgrids
management," in IEEE Power and Energy Magazine, vol. 6, no. 3, pp. 54-65,
May-June 2008.
TABLE I. MICROGRID PARAMETERS
Parameters [2] Q. Jiang, M. Xue and G. Geng, "Energy Management of Microgrid in
System Grid-Connected and Stand-Alone Modes," in IEEE Transactions on Power
Description Rated power
PV Generator is simulated with real Systems, vol. 28, no. 3, pp. 3380-3389, Aug. 2013.
PV Generator 25 kW
time irradiance data.
[3] Dan Wang et al., "A Demand Response and Battery Storage Coordination
Battery System 210 Ni-Cd cells (240Vdc) 15 kW Algorithm for Providing Microgrid Tie-Line Smoothing Services," in IEEE
Transactions on Sustainable Energy, vol. 5, no. 2, pp. 476-486, April 2014.
Main Grid Constant Power Grid 3 kW
[4] J. A. P. Lopes, C. L. Moreira and A. G. Madureira, "Defining control
Houses Residential group of three houses 20 kW strategies for MicroGrids islanded operation," in IEEE Transactions on
Power Systems, vol. 21, no. 2, pp. 916-924, May 2006.

The Demand Response graph represents the controllable [5] C. Sharma, S. Bahadoorsingh and A. Aiyejina, "Integrating applicable
loads of each house that are: the air conditioners and some sources of renewable energy in the Caribbean," Power and Energy Society
General Meeting, 2011 IEEE, San Diego, CA, 2011, pp. 1-6.
lights. This controllable loads represent approximately 2.5 kW
of the total power consumption. When the SOC of the [6] M. I. Mercado-Oliveras and A. A. Irizarry-Rivera, "Residential grid-tied
battery’s reaches its critical level it activates a level of photovoltaic energy system design in Puerto Rico," North American Power
response. For example, in the SOC plot its initial value is 20% Symposium (NAPS), 2015, Charlotte, NC, 2015, pp. 1-6.
and this represents the most critical point of the battery. The
[7] H. Kanchev, D. Lu, F. Colas, V. Lazarov and B. Francois, "Energy
demand response activates and disconnects the controllable Management and Operational Planning of a Microgrid With a PV-Based
loads in order to drop the demand from the houses and try to Active Generator for Smart Grid Applications," in IEEE Transactions on
maintain the power flow balance. Once the SOC starts to Industrial Electronics, vol. 58, no. 10, pp. 4583-4592, Oct. 2011.
increase the loads are connected back slowly.
[8] Chaouachi, A.; Kamel, R.M.; Andoulsi, R.; Nagasaka, K.,
The purpose of this controller is to maintain the power grid "Multiobjective Intelligent Energy Management for a Microgrid," in
on a constant value in this case 2.8 kW and the SOC at its Industrial Electronics, IEEE Transactions on , vol.60, no.4, pp.1688-1699,
maximum value. It can be seen in the Electrical Grid graph the April 2013
curve it is mostly constant except when the demand respond,
[9] Zhixin Miao; Ling Xu; Disfani, V.R.; Lingling Fan, "An SOC-Based
the Photovoltaic Generators and the battery are no enough to Battery Management System for Microgrids," in Smart Grid, IEEE
supply the load at the time and it needs more energy from the Transactions on , vol.5, no.2, pp.966-973, March 2014
grid this can be observe in the graph. Also, the opposite occurs
when the Photovoltaic Generator produces more energy than [10] Bragard, M.; Soltau, N.; Thomas, S.; De Doncker, R.W., "The Balance
the load at the time and the 15 kW limit of the battery. When of Renewable Sources and User Demands in Grids: Power Electronics for
this happens the excess of energy is delivered to the electrical Modular Battery Energy Storage Systems," in Power Electronics, IEEE
Transactions on , vol.25, no.12, pp.3049-3056, Dec. 2010
grid. However, the grid stays for the most part constant. Since,
the initial value of the SOC is 20 % it can be observe in the [11] M. García-Plaza, D. Serrano-Jiménez, J. Eloy-García Carrasco, and J.
graph that its value tends to exceed its maximum value from Alonso-Martínez, “A Ni–Cd battery model considering state of charge and
the day before. hysteresis effects,” J. Power Sources, vol. 275, pp. 595–604, Feb. 2015.

V. CONCLUSIONS
[12] C. Roncero-Clemente, E. González-Romera, E. Romero-Cadaval, M. I.
In this paper, a Fuzzy logic battery management system Milanés-Montero and V. Miñambres-Marcos, "PSCAD/EMTDC model for
has been proposed to control the battery at grid-connected photovoltaic modules with MPPT based on manufacturer specifications,"
mode. Three modes of operation were de fined for the battery Compatibility and Power Electronics (CPE), 2013 8th International
to be operated. The functionalities of the BMS proposed in Conference on, Ljubljana, 2013, pp. 69-74.
this paper are tested through case studies in HIL/Matlab the

978-1-5090-5167-0/16/$31.00 ©2016 IEEE

You might also like