energies

Article
Active Charge Balancing Strategy Using the State of
Charge Estimation Technique for a PV-Battery
Hybrid System
Md Ohirul Qays 1 , Yonis Buswig 1 , Md Liton Hossain 2, * and Ahmed Abu-Siada 2
1 Department of Electrical and Electronic Engineering, Faculty of Engineering, Universiti Malaysia Sarawak,
Kota Samarahan 94300, Sarawak, Malaysia; [email protected] (M.O.Q.); [email protected] (Y.B.)
2 School of Electrical Engineering, Computing and Mathematical Sciences, Curtin University, Perth 6102,
Australia; [email protected]
* Correspondence: [email protected]




Received: 30 May 2020; Accepted: 1 July 2020; Published: 3 July 2020


Abstract: Charging a group of series-connected batteries of a PV-battery hybrid system exhibits an

imbalance issue. Such imbalance has severe consequences on the battery activation function and the
maintenance cost of the entire system. Therefore, this paper proposes an active battery balancing
technique for a PV-battery integrated system to improve its performance and lifespan. Battery state of
charge (SOC) estimation based on the backpropagation neural network (BPNN) technique is utilized
to check the charge condition of the storage system. The developed battery management system
(BMS) receives the SOC estimation of the individual batteries and issues control signal to the DC/DC
Buck-boost converter to balance the charge status of the connected group of batteries. Simulation and
experimental results using MATLAB-ATMega2560 interfacing system reveal the effectiveness of the
proposed approach.

Keywords: active battery balancing; backpropagation neural network; DC/DC Buck-boost converter;
PV-battery integrated system; state of charge estimation

1. Introduction
In the last few decades, photovoltaics (PV) have been broadly used as a cost effective and reliable
renewable energy source with the aim of reducing the reliance on fossil fuel used in conventional
thermal generation [1]. In [2], a detailed PV model is investigated using an artificial neural network
(ANN). However, adopting such model in practical PV systems increases the implementation time and
complexity when compared to classical models. A maximum power point tracking (MPPT) system
is essential for PV systems to yield the maximum accessible power irrespective the irregular solar
irradiance and atmospheric condition. A combination of Adaptive neuro-fuzzy inference system
(ANFIS) and hill climbing (HC) MPPT technique is presented in [3]. Reported results outperformed the
tracking precision of the conventional MPPT technique. To increase the profit of the PV system during
partial shading events, an adaptive perturb and observation (P&O) MPPT algorithm is proposed [4].
However the proposed algorithm is only valid for monocrystalline and polycrystalline-based PV panels.
Ramp and step alterations of solar radioactivity was used to test the developed P&O MPPT technique.
Results show that the proposed technique is of quicker tracking speed than the conventional method.
During cloudy days and night-time, battery energy storage systems (BESS) play an important role
for providing off-grid power [5–7]. In [8], a 1.5 kW PV-battery integrated system is simulated using
HOMER software to assess the battery lifetime. Simulation was conducted by keeping the battery depth
of discharge (DOD) within 50~80%. The research reveals that the greenhouse gas (GHG) effect of the

Energies 2020, 13, 3434; doi:10.3390/en13133434 www.mdpi.com/journal/energies

Energies 2020, 13, 3434 2 of 17

atmosphere can be reduced by improving the battery lifespan through a proper charging/discharging
control system during the off-grid condition. The SOC of a 48 V valve-regulated lead-acid battery
(VRLA) is estimated in [9] through the sigma-point Kalman filter (SPKF) algorithm. The 2nd order
Equivalent Circuit Model (ECM) was designed in MATLAB-PIC18F2550 interfacing system and tested
via Unscented Kalman Filter (UKF) scheme.
Nevertheless, BESS comes across quite a few challenges, including battery to battery factorial
delinquencies and cell performance. Figure 1 shows that a group of series connected batteries may
encompass dissimilar levels of SOC. Thus, a minimum-labeled SOC battery is deep-discharged during
the discharging period, while a maximum-labeled SOC battery will be over-charged in the charging
period [10,11]. The least squares variable projection algorithm (LSVPA) for lithium-ion batteries
(LIB) was employed in [12] in order to improve the noise immunity and unbiased model parameter
identification. Reported results indicated a reduction of the root mean square error from 10.22 mV to
0.54 mV by using a MATLAB-Arbin BT2000 testing system. Wöhler-curve-based LIB battery functioning
life is studied in [13] for economic assessment and sizing of battery systems in railway applications.
The proposed method improved the battery lifetime compared to the semi-empirical method. In [14], a
Luenberger observer is utilized along with instrumental variable estimation and bilinear principle
combination to estimate the SOC for LIB. The proposed method resulted in 2.60% and 1.25% error in
the case of hybrid pulse experiment and Federal Urban Dynamic Schedule conditions, respectively.
In [15], four series-connected 12 V lead acid (LA) batteries were investigated for electric motorcycles.
The circuit was designed using six MOSFET switches to improve the life-cycle of the battery bank.
However, the proposed method still requires further experimentation for real-life application.

Figure 1. Different leveled SOC of individual batteries within a battery pack.

Two types of battery-cell balancing are classified as passive and active cell balancing techniques [16].
The passive balancing technique coverts the extra energy into heat through a resistor that is not
approachable to surroundings whereas the active balancing technique allocates the energy among the
batteries from which power losses are significantly reduced. A solar power-supported 48 V LA battery
was explored for electric vehicles in [17]. Eight MOSFET switches and a National Instrument (NI) DAQ
were implemented to control the balancing system for four individual 12 V LA batteries. Though the
projected model increases the battery energy from 2.1% to 3.3% in every 13.2 km, the polarization and
temperature effect of the battery was not considered. A real-time monitoring system for multiple LA
batteries using an ATmega128a microcontroller-based BQ34Z110 board is proposed in [18]. WizFi210
is integrated with several types of sensors, such as humidity, temperature, and liquid level devices, to
monitor the battery condition and calculate the SOC, which was not very accurate.
To the best of the authors’ knowledge, the active charge equalization strategy based on SOC
estimation approach using the BPNN scheme is a better solution to prolong the battery activation
life for a PV-battery hybrid system, and no such research has been investigated previously. Hence,
the foremost contribution of this paper is the presentation of an improved BMS through active cell
balancing technique for a standalone PV-Battery integrated system. Additionally, an operative SOC
estimation technique is proposed using a BPNN algorithm to prolong the functional life of a lead acid
battery along with the Wöhler-curve-based aging model. The proposed model is validated through
simulation and experimental analyses using a MATLAB-ATMega2560 interfacing system.
The rest of this paper is arranged as follows: Section 2 explains the modeling approach of
the proposed model. Numerical simulation and experimental analysis are presented in Section 3.
Key results and a discussion of the proposed methodology are deliberated in Section 4, while key
conclusions are drawn in Section 5.
Energies 2020, 13, 3434 3 of 17
𝑰𝑺𝑪 𝑫
𝑹𝒔 𝑹𝒔𝒉
2. Modeling Approach
𝑽𝑫
𝑰𝑷𝑽 = 𝒏𝑷 𝑰𝑳 − 𝑰𝑫 −
2.1. Photovoltaic (PV) Mathematical Modeling 𝑹𝒔𝒉
𝑰𝑫 combination of series or parallel
The 𝑰𝑳 coupled solar cells regulate the generated power of
𝒏𝑷
PV panels. Principally, two types of PV cells 𝑹𝒔𝒉 are categorized as double-diode
𝑽𝑫 and single-diode
modules [19,20]. Though double-diode modules exhibit high accuracy, low analytical speed is depicted
due to the structural complexity. In contrast, single-diode modules are widely adopted in numerous
𝒒(𝑽 + 𝑰𝑹 𝒔
power electronics application due 𝑰𝑫 =
to 𝒏their
𝑷 𝑰𝑶𝑺 high
𝐞𝐱𝐩 precision. Consequently,
−𝟏 this paper focuses on a
𝒏𝒔 𝑨𝑲𝑻
single-diode-based PV module which entails a current source (ISC ), diode D, and series-parallel
resistors Rs , Rsh . From Figure 2, Kirchhoff’s Current 𝑰𝑶𝑺 𝒒
law formulates the output 𝑽
current (IPV ) as:
𝑰𝒄𝒆𝒍𝒍 𝑹𝒔 𝒏𝒔 𝑨
𝑲 𝑻 V D
IPV = nP IL − ID − (1)
𝑰𝑳 Rsh 𝑰𝑶𝑺
where ID represents the diode current, IL is the light-generated current, number of parallel connected
cells is nP , shunt resistance is Rsh and 𝝍
𝑰𝑳 = diode voltage is VD . Diode current can be calculated from
[𝑰𝑳𝒓𝒆𝒇 + 𝝁𝑰𝒔𝒄 (𝑻 − 𝑻𝒓𝒆𝒇 )]
Equation (2): 𝝍 𝒔𝒉
" ! #
q(V −𝒒𝑬
+ IRs
ID =𝑰n𝑶𝑺
P I=
OS𝑫exp 𝟑
𝒇 𝑻 𝐞𝐱𝐩 ) −1 (2)
ns(AKT
𝑨𝑲𝑻
where the reverse saturation current is denoted by IOS , electron charge is 𝝍 q, output
𝝍𝒔𝒉 voltage
𝑰𝑳𝒓𝒆𝒇 is V,
cell current is Icell , series resistance is𝑻 Rs , number
𝑻𝒓𝒆𝒇 of series connected cells is ns , ideality factor is A,
𝑫𝒇
Boltzmann constant is K and the cell temperature 𝑬 is T in Kelvin.

Figure 2. PV cell structure based on a single diode module.

The light-generated current (IL ) and reverse saturation current (IOS ) is affected by solar irradiance
and atmospheric temperature and can be calculated from Equations (3) and (4):

ψ  
IL = [ILref + µIsc T − Tref ] (3)
ψsh
!
−qE
IOS = Df T3 exp (4)
AKT
where the existing irradiance and reference irradiance are represented by ψ and ψsh . ILref is the light
current at reference condition, T and Tref indicate the present and reference temperatures; respectively,
Df is the diode diffusion factor and E is the bandgap energy.

2.2. Maximum Power Point Tracking (MPPT)

The PV systems should be operated at maximum capability irrespective the disparities of
environmental and loading circumstances. The maximum probable power can be taken out from the
PV panel using a systematic prevailing appliance namely maximum power point tracking (MPPT).
Generally, a DC-DC power converter is operated in conjunction with a PV array whose duty cycle is
identified via the MPPT system. As shown in Figure 3, The determination of the MPPT is to adjust
Energies 2020, 13, 3434 4 of 17

the I–V functional point of PV modules to harvest the supreme accessible power from the electronic
converter [21]. Several types of MPPT techniques comprising fuzzy logic based MPPT, artificial
neural network (ANN), perturbation and observation (P&O), fractional open-circuit voltage (FOCV),
incremental conductance (IC), look-up table, ripple correlation control (RCC) and one cycle control
(OCC)-based MPPT can be adopted [22]. The authors of [19] reported that the fuzzy logic-based MPPT
technique is the most truthful method compared to other approaches. A modified P&O-based MPPT
technique is presented in [23] through MATLAB/SIMULINK software. Compared to the traditional
P&O method, this algorithm was capable ∆ to recognize the voltage maximum∆ power point (MPP)
exactly. Nonetheless, experimental justification of this technique was recommended by the authors
along with the temperature effect in future research.

∆ ∆

Figure 3. Block diagram of PV-MPPT governing system.

In order to improve the working prototype in real world application, a P&O-based MPPT system
as shown in Figure 4 is presented in [24]. In this method, voltage and current collected from the PV
panel are used to calculate the solar power at a time (t) which is compared with a former sample
measured at time (t − 1). Voltage perturbation stays on the similar way, if the power alteration is
positive. Negative power alteration (∆P) decreases the voltage perturbation (∆V) as can be seen in
Table 1. Following this technique, MPP can be traced over the entire PV curve. Figure 4 summarizes
the flowchart of the entire P&O algorithm.

− −

Figure 4. MPPT flowchart for the P&O algorithm.

− −
 −− −−
−− −−

Energies 2020, 13, 3434 5 of 17

Table 1. Perturbation summary for the P&O algorithm.

Perturbation ∆P Resultant Perturbation

+ve +ve +ve
+ve −ve −ve
−ve +ve −ve
−ve −ve +ve

2.3. State of Charge (SOC) Estimation

𝑛𝑒𝑡𝑖
𝑛𝑒𝑡𝑖 = = 𝑥𝑥 𝑤
𝑤 + + 𝑏𝑏
State of charge (SOC) is essential for LA batteries to declare their proper operation and to eliminate
over-charging/deep-discharging
𝑛𝑒𝑡𝑖 issues. However, the nonlinear electrochemical 𝑛 responses mark the
𝑛𝑒𝑡𝑖 𝑛 𝑥𝑥
SOC estimation𝑤a challenging subject [25,26]. In the case of nonlinear systems, a BPNN is the most𝑖often
𝑤 𝑖
employed
𝑛; technique because of its self-learning and 𝑛 feedback distinction
𝑏𝑏 [27]. Thus, a BPNN-based
𝑛; 𝑛
SOC assessment methods is adopted in this paper, as shown in Figure 5. As mentioned in Figure 6, SOC
is
𝑛 estimated for every single battery independently by collecting its voltage and current parameters
𝑛
where VT is the total voltage of the battery bank. The root mean square error (RMSE) of the SOC
11 −− 𝑒𝑥𝑝(−2𝑛𝑒𝑡𝑖 ))
estimation is minimized by adjusting ℎℎ =
= 𝑓(𝑛𝑒𝑡𝑖
the )) =
weights.
𝑓(𝑛𝑒𝑡𝑖 =In the𝑒𝑥𝑝 (−2𝑛𝑒𝑡𝑖
BPNN, the total input of hidden layers can be
11 ++ 𝑒𝑥𝑝
𝑒𝑥𝑝 (−2𝑛𝑒𝑡𝑖
(−2𝑛𝑒𝑡𝑖 ))
calculated from:
X3
𝑛𝑒𝑡𝑜
neti n ==
𝑛𝑒𝑡𝑜 = xiℎℎw𝑔 𝑔 +
+ b𝑝𝑝n
in + (5)
i=1

where neti n represents the total inputs to the hidden layer neuron n, 𝑜xi symbolizes the value of
𝑛𝑒𝑡𝑜
𝑛𝑒𝑡𝑜 𝑜
input neuron, win deliberates the weight
𝑖𝑖 ℎℎ between the input layer and hidden layer𝑖𝑖neurons i and n;
respectively, bias
𝑜𝑜 of𝑔the hidden layer neuron n is mentioned as𝑜bn . 𝑝
𝑔 𝑜 𝑝

Figure 5. Structural design of the proposed BPNN algorithm.

Figure 6. SOC estimation process from batteries.

Energies 2020, 13, 3434 6 of 17

A tangent function Equation (6) is the utilized activation function for the hidden layer neuron n.
The valuation of the output layer neuron can be found from Equation (7):

1 − exp(−2netin )
hn = f (netin ) = (6)
1 + exp(−2netin )
n
X
neto = hi gi + p (7)
i=1

where, neto represents the overall product of the output layer neuron o, the value of hidden layer
neuron to output layer neuron for i is hi , the weight between hidden layer neuron i and the output
layer neuron o is gi , and the bias of the output layer neuron o is p.
A sigmoid function is used as the applied activation function of the output layer neuron where
output yn can be calculated from Equation (8):

1
yn = f (neto) = (8)
1 + exp(−neto)

ei = fRe f − fEst (9)

∂Error ∂Error ∂ fEst

= (10)
∂wnij ∂ fEst ∂wnij
v
n
t
X
RMSE = ( e2i )/n (11)
i=1

where ei represents the output error, and fRe f and fEst are the reference output function and estimated
output function, respectively. Using a gradient descent algorithm (GDA) as given by Equations (9)–(11),
weights are restructured repeatedly to attain the expected output and reduce the error. From
Equation (11) the RMSE is considered from which the improved weights are achieved while the error
is diminished.

2.4. DC/DC Buck-Boost Converter

A circuit diagram of the used four switched synchronous DC/DC Buck-boost converter is shown
in Figure 7 in which the battery cells are charged by the PV supply voltage VPV [28,29]. Four peculiar
MOSFET switches (S1 ~S4 ) are tangled into the DC/DC converter to standardize the energy transference
of battery arrangements (VnBat ). The module can be operated in Buck mode, Buck-boost mode or boost
mode by pursuing the energy requirement of the process. To activate the Buck mode operation, S3 is
turned off while S4 is on. Additionally, S1 and S2 are activated to administer the procedure. MOSFETs
S2 and S1 are opened and closed repeatedly to charge the inductor by the battery power. Inductor
current increases while a capacitor supplies the output current for the charging duration. Furthermore,
S1 and S2 are opened and closed frequently throughout the inductor discharging period to charge the
load at this point. By altering the duty cycle to be d < 1, the average load voltage VLoad (or dVBat ) can
be coupled together with the battery voltage VBat .
For the boost mode, MOSFETs S2 and S1 are turned off and on in a complement manner. S4 and S3
are opened and closed constantly by ensuring the inductor charging requirement. The arrangements
are reversed at the discharging period where the average load voltage VLoad can be calculated as
1/(1 − d)VBat . However, the output load voltage at Buck-boost operation is VLoad or d/(1 − d)VBat which
can be adjusted by following the desired battery voltage. MOSFET switches S2 and S4 are retained on
to charge the inductor as well as S1 and S3 are preserved off for this mode. The MOSFET switches
are activated on the inverse side to discharge the particular inductor. Table 2 summarizes the four
switched synchronous DC/DC Buck-boost converter operation.
Energies 2020, 13, 3434 7 of 17

Figure 7. Four switched synchronous DC/DC Buck-boost converter.

Table 2. Summary of four switched DC/DC Buck-Boost converter operation.

MOSFET Switch Inductor Condition Buck Mode Boost Mode Buck-Boost Mode
S1 Charge ON ON ON
− Discharge OFF OFF ON −
S2 Charge OFF OFF OFF
Discharge ON ON OFF
S3 Charge OFF 𝑑𝑉
ON 𝑉
ON
𝑉 = 𝑑𝑉
Discharge 𝑉 OFF = 𝑑𝑉 𝑉 =OFF 𝑉 = OFF
1−𝑑 1−𝑑
S4 Charge ON OFF OFF
Discharge ON ON ON
Average Load Voltage VLoad = dVBat VLoad = dVBat VLoad = dV Bat
1−d
VLoad = V Bat
1−d

2.5. Battery Lifetime Estimation

A Wöhler-curve-based aging model is considered in this𝑳𝑳research
𝒊
to estimate the battery
lifetime [30,31]. This method is structured based on the battery DOD which is complementary
𝑬𝒊
to SOC. The DOD versus number of cycles for = Whöler
𝑳𝑳𝒊the curve of a lead-acid battery is shown in
𝑬𝒎𝒂𝒙
𝒊
Figure 8. In this regard, the mathematical representation for battery lifetime loss LLi is expressed as:
𝑬𝒊 𝑬𝒎𝒂𝒙
𝒊
Ei 𝑳𝑳
LLi = max (12)
Ei

where i indicates a certain DOD, Ei is the number

𝑳𝑳 = of𝑳𝑳events
𝒊
and Emax
i
is the maximum number of
events which can be withstood for the battery. The𝒊 entire loss of battery-lifetime LL for the overall
range of DOD (0~100%) can be calculated from:
𝑳𝑳
X 𝑳𝑳
LL = LLi (13)
i 𝟏
𝑳𝒊𝒇𝒆𝒕𝒊𝒎𝒆 =
𝑳𝑳

Figure 8. Wöhler curve for a lead-acid battery [30,31].

Energies 2020, 13, 3434 8 of 17

The battery is considered at its end of life when LL reaches 1. Thus, the total lifetime of the battery
can be estimated from Equation (14) as the inversion of LL by considering the SOC profile.

1
Lifetime = (14)
LL

3. Proposed Methodology

3.1. Simulation Model

To validate the proposed BMS, the predicted model is simulated using MATLAB/SIMULINK
2019a software (MathWorks). The proposed methodology controls the MPPT scheme for PV panel
and estimates the SOC of the battery bank through BMS in the designed model. The MPPT method
generates the maximum possible power for the PV panel from the available solar insolation under any
atmospheric conditions. On the other hand, the BMS receives the battery voltage and current from
measuring sensors and estimates the existing SOC using BPNN algorithm. In this way, the charging
and discharging process for the battery is performed through the PV supply and the connected load.
As stated in [32–34], beyond 70~80% SOC the stocking battery exhibits over-charging, while lower than
30~40% SOC the holding battery runs in the deep-discharging manner. Thus, 40–80% SOC is ranged
in this paper to perform smooth battery charging-discharging operation and, hence, lengthening the
battery lifetime.
To standardize the battery bank capability and lessen the SOC imbalance among individual
batteries, active battery balancing strategy is applied in the proposed model. As shown in Figure 9, the
BMS generates a balancing control signal through the DC/DC Buck-boost converter by estimating the
SOC from each individual battery. Every single battery SOC (Si ) is compared with the average SOC
(S). If the variance is in excess of a predefined threshold value (Sthr ), the BMS instructs the relevant
𝑆
converter ̅ of a certain battery to carry on charging/discharging process according to Equation (15). In
𝑆 𝑆
addition, a DC/AC power inverter is engaged to conform the validity of the proposed approach. The
entire flowchart of the proposed BMS controlling technique is shown in Figure 10.


 Si − S > Sthr Discharging
− 𝑆S̅ >

𝑆Si − 𝑆thr 𝐷𝑖𝑠𝑐ℎ𝑎𝑟𝑔𝑖𝑛𝑔

<S Charging (15)


̅

𝑆 −𝑆 <𝑆 𝐶ℎ𝑎𝑟𝑔𝑖𝑛𝑔

Others Islanding


𝑂𝑡ℎ𝑒𝑟𝑠 𝐼𝑠𝑙𝑎𝑛𝑑𝑖𝑛𝑔

Figure 9. Active charge equalization structural design.

Energies 2020, 13, 3434 9 of 17

Figure 10. Flowchart for the proposed BMS.

3.2. Experimental Model

As shown in Figure 11, an experimental prototype of the projected model is prepared in a laboratory
environment. The hardware prototype model consists of four 12 V, 7 Ah sealed gel rechargeable
lead-acid batteries, solar module PV panel (VOC = 21.6V, ISC = 3.40A, Pm = 50W) and 5W LED light
as a load. In this prototype, the PV solar panel, battery, DC/AC inverter, and loads are connected
through fuse and circuit breaker. MATLAB/SIMULINK 2019a software (MathWorks) is interfaced with
an Arduino Integrated Development Environment (IDE) ATMega2560 controller. The integration of
MATLAB Tools for Arduino IDE code generation block diagram is shown in Figure 12. Voltage and
current sensors are used to calculate the voltage and current from the corresponding components and
delivers the signal to a personal computer (PC) via the Arduino ATMega2560 electronic board. The
SIMULINK model evaluates the data and appraises the PV power, battery SOC, and load power for
the BMS controller. The DC/AC inverter converts the DC voltage into AC voltage which is formed
as a modified AC sine wave. By imitating the effective condition of the governing mechanism, BMS
directs the DC/DC Buck-boost converter to charge, discharge, or island the specific battery to balance
the battery charge status and improve the battery-life proficiency.
Energies 2020, 13, 3434 10 of 17

Figure 11. Experimental hardware setup of the proposed model.

Figure 12. Integration of MATLAB-Arduino IDE code generation block diagram.

4. Results and Discussion

The experiment was conducted in January 2020 in the Renewable Energy Lab of Universiti
Malaysia Sarawak (UNIMAS). Active battery balancing strategy accompanied by BMS for PV-Battery
integrated system are conveyed together in the MATLAB-Arduino interfacing experience. Four
series-connected lead-acid batteries are employed in this paper to validate the proposed battery
balancing methodology for which the relationship between OCV vs. SOC is as shown in Figure 13 [35].

Figure 13. OCV vs. SOC for the employed 12 V lead-acid battery.

Initially, the batteries are comprised with dissimilar SOCs, such as 76%, 77%, 78%, and 79%. The
PV supply voltage profiles at different atmospheric conditions (899~924 W/m2 solar irradiance and
27~32 ◦ C temperature) for the four series batteries are shown in Figure 14. Figures 15–18 demonstrate
the battery balancing outcomes and exemplified some comparisons with and without the proposed
strategy. Figure 15 compares the total battery bank voltage of the conventional and proposed method.
It can be observed that the overall storage voltage for the proposed model is improved compared to
the conventional method.
Energies 2020, 13, 3434 11 of 17

Figure 14. PV supply voltage at different atmospheric conditions.

Figure 15. Overall battery bank voltage using conventional and the proposed method.

(a) (b)
Figure 16. Current profile for individual batteries (a) without and (b) with the proposed method.

(a) (b)
Figure 17. Individual battery Voltages (a) without and (b) with the proposed method.
Energies 2020, 13, 3434 12 of 17

(a) (b)
Figure 18. Individual batteries SOC (a) without and (b) with the proposed method.

Figure 16a reveals that although the supply current of the four series batteries should be the same,
as a consequence of different SOC, the supply current is different without the implementation of any
BMS scheme which may result in terminal voltage divergence among the batteries as can be noticed in
Figure 17a. Implementation of the proposed BMS brings the cells current to nearly the same level as
can be observed from Figure 16b which can limit the voltage differences among the four batteries as
can be seen in Figure 17b. Figure 18 deliberates the SOC before and after the utilization of the proposed
method. Without the proposed technique, the four batteries exhibit imbalance as can be shown in
Figure 18a. With the utilization of the proposed methodology, the SOC of the four cells becomes close
to 78.5% as shown in Figure 18b.
The fundamental proficiency of the battery energy storage system depends on the SOC estimation
accuracy [22]. From Figure 19, SIMULINK toolbox trains the collected battery data in which the
overall regression for training, testing, and validation data is about 97%. Therefore, the actual and
estimated SOC is approximately the same as shown in Figure 20. SOC estimation accuracy (RMSE
0.082%) of the BPNN algorithm shown in Figure 21 and listed in the simulation results of Table 3
attest to the robustness of the proposed technique. The accuracy of the estimated SOC using the
proposed technique outperforms the estimation accuracy using the invariant-imbedding method (IIM),
Extended Kalman Filter (EKF), and Unscented Kalman filter (UKF) reported in the literature owing to
the adopted weight observing scheme.
The produced DC voltage was transformed into a modified 240 V, 50 Hz AC waveform, as shown
in Figure 22. A comparison between the simulation and experimental results of the obtained AC sine
waveform is shown in Figure 23 which reveals good agreement of the experimental and simulation
results. The key feature of the proposed BMS controller is the elongation of the battery storage
functioning life. The misbalance of battery charging/discharging affects its performance and reduces
its lifespan. Based on Wöhler-curve along with the proposed BMS, Figure 24 reveals that the projected
model enhances the operational life of the battery compared to the conventional models.
Energies 2020, 13, 3434 13 of 17

Figure 19. Training, testing, and validation results with regression.

Figure 20. Comparison between actual and estimated SOC using the proposed technique.

Figure 21. RMSE of the estimated SOC.

Energies 2020, 13, 3434 14 of 17

Table 3. SOC estimation accuracy comparison.

Reference Year Battery Type SOC Algorithm Error (%)

[36] 2016 LIB PI Observer 3
[37] 2017 LIB EKF 6.5
[38] 2017 LIB EKF 3
[39] 2018 LIB UKF 1.5
[40] 2019 LIB EKF 2
[41] 2019 LIB BPNN 1.2
Proposed Model 2020 Lead Acid BPNN 0.082

Figure 22. Experimental modified AC voltage.

Figure 23. Comparison between simulation and experimental modified AC voltage.

Figure 24. Battery lifetime comparison of the proposed and conventional models.

r
Energies 2020, 13, 3434 15 of 17

5. Conclusions
Battery storage is an essential, but expensive, element for PV-battery off-grid systems that ensures
the continuity of the power supply at night and during cloudy days. This paper deliberates a new
BMS which comprehends the PV model, P&O MPPT technique, and load and SOC estimation of
battery storage systems. An active battery balancing approach based on adaptive BPNN and SOC
estimation strategy has been proposed for to sustain the effectiveness of a battery pack system. The
proposed model is applied on a series-connected battery string to eliminate the SOC disparity among
individual batteries in the pack. The proposed BMS administers the SOC level of each battery and
issues an activation signal to the corresponding DC/DC Buck-boost converter. Experimental analysis is
conducted on the simulated model with the help of the MATLAB-ATmega2560 interfacing portion to
validate the feasibility of the proposed technique. The proposed control strategy detects the charge
status of the lead-acid battery and takes decisions to accomplish the expected operational goals and
elongate the battery lifetime with 0.082% RMSE in the estimated SOC. Further research is recommended
to replicate this study for lithium-ion batteries as well as on-grid storage systems.

Author Contributions: M.O.Q. conceived the original idea; M.O.Q. and Y.B. wrote and edited the manuscript;
and M.L.H. and A.A.-S. supervised the study and revised the final version of the paper. All authors have read and
agreed to the published version of the manuscript.
Funding: This research was funded by Research and Innovation Management Center (RIMC), UNIMAS
via Fundamental Research Grant Scheme, Ministry of Higher Education, Malaysia, grant number
FRGS/1/2017/TK10/UNIMAS/03/1.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Ghenai, C.; Bettayeb, M. Modelling and performance analysis of a stand-alone hybrid solar PV/Fuel
Cell/Diesel Generator power system for university building. Energy 2019, 171, 180–189. [CrossRef]
2. Castro, R. Data-driven PV modules modelling: Comparison between equivalent electric circuit and artificial
intelligence based models. Sustain. Energy Technol. Assess. 2018, 30, 230–238. [CrossRef]
3. Lasheen, M.; Abdel-Salam, M. Maximum power point tracking using Hill Climbing and ANFIS techniques
for PV applications: A review and a novel hybrid approach. Energy Convers. Manag. 2018, 171, 1002–1019.
[CrossRef]
4. Ahmed, J.; Salam, Z. An enhanced adaptive P&O MPPT for fast and efficient tracking under varying
environmental conditions. IEEE Trans. Sustain. Energy 2018, 9, 1487–1496.
5. Essa, M. Management of charging cycles for grid-connected energy storage batteries. J. Energy Storage 2018,
18, 380–388. [CrossRef]
6. Correa-Florez, C.A.; Gerossier, A.; Michiorri, A.; Kariniotakis, G. Stochastic operation of home energy
management systems including battery cycling. Appl. Energy 2018, 225, 1205–1218. [CrossRef]
7. Hossain, E.; Zawad, M.; Rakibul Islam, K.H.; Akash, M.Q. Design a novel controller for stability analysis of
microgrid by managing controllable load using load shaving and load shifting techniques; and optimizing
cost analysis for energy storage system. Int. J. Renew. Energy Res. 2016, 6, 772–786.
8. Akinyele, D.; Belikov, J.; Levron, Y. Battery storage technologies for electrical applications: Impact in
stand-alone photovoltaic systems. Energies 2017, 10, 1760. [CrossRef]
9. Kujundzic, G.; Vasak, M.; Matusko, J. Estimation of VRLA battery states and parameters using Sigma-point
Kalman filter. In Proceedings of the 2015 International Conference on Electrical Drives and Power Electronics
(EDPE), High Tatras, Slovakia, 21–23 September 2015; pp. 204–211.
10. Cui, X.; Shen, W.; Zhang, Y.; Hu, C.; Zheng, J. Novel active LiFePO4 battery balancing method based on
chargeable and dischargeable capacity. Comput. Chem. Eng. 2017, 97, 27–35. [CrossRef]
11. Caspar, M.; Eiler, T.; Hohmann, S. Systematic Comparison of Active Balancing: A Model-Based Quantitative
Analysis. IEEE Trans. Veh. Technol. 2018, 67, 920–934. [CrossRef]
12. Wei, Z.; Zhao, D.; He, H.; Cao, W.; Dong, G. A noise-tolerant model parameterization method for lithium-ion
battery management system. Appl. Energy 2020, 268, 114932. [CrossRef]
Energies 2020, 13, 3434 16 of 17

13. Herrera, V.I.; Martinez-laserna, E.; Milo, A.; Gaztañaga, H. Techno-economic assessment of Lithium-ion
battery lifetime estimation methods for sizing and operation conditions definition in railway applications.
In Proceedings of the 14th IEEE Vehicle Power and Propulsion Conference (VPPC), Belfort, France,
11–14 December 2017.
14. Wei, Z.; Dong, G.; Zhang, X.; Pou, J.; Quan, Z.; He, H. Noise-Immune Model Identification and State of
Charge Estimation for Lithium-ion Battery Using Bilinear Parameterization. IEEE Trans. Ind. Electron. 2020,
67, 417–426. [CrossRef]
15. Liu, V.T.; Chen, J.R. Balancing for Lead-Acid Batteries of Electric Motorcycles. Appl. Mech. Mater. 2015,
764–765, 491–495. [CrossRef]
16. Knap, V.; Stroe, D.I.; Christensen, A.E.; Propp, K.; Fotouhi, A.; Auger, D.J.; Schaltz, E.; Teodorescu, R.
Self-balancing feature of Lithium-Sulfur batteries. J. Power Sources 2017, 372, 245–251. [CrossRef]
17. Duan, C.; Wang, C.; Li, Z.; Chen, J.; Wang, S.; Snyder, A.; Jiang, C. A solar power-assisted battery balancing
system for electric vehicles. IEEE Trans. Transp. Electrif. 2018, 4, 432–443. [CrossRef]
18. Rauniyar, A.; Irfan, M.; Saputra, O.; Kim, J.; Lee, A.; Jang, J.; Shin, S. Design and development of a Real-Time
monitoring system for multiple lead–acid batteries based on Internet of things. Futur. Internet 2017, 9, 28.
[CrossRef]
19. Gopi, R.R.; Sreejith, S. Converter topologies in photovoltaic applications—A review. Renew. Sustain. Energy Rev.
2018, 94, 1–14. [CrossRef]
20. Sundareswaran, K.; Vigneshkumar, V.; Palani, S. Development of a hybrid genetic algorithm/perturb and
observe algorithm for maximum power point tracking in photovoltaic systems under non-uniform insolation.
IET Renew. Power Gener. 2015, 9, 757–765. [CrossRef]
21. Mohanty, S.; Subudhi, B.; Ray, P.K. A Grey Wolf-Assisted Perturb & Observe MPPT Algorithm for a PV
System. IEEE Trans. Energy Convers. 2017, 32, 340–347.
22. Bendib, B.; Belmili, H.; Krim, F. A survey of the most used MPPT methods: Conventional and advanced
algorithms applied for photovoltaic systems. Renew. Sustain. Energy Rev. 2015, 45, 637–648. [CrossRef]
23. Alik, R.; Jusoh, A. Modified Perturb and Observe (P&O) with checking algorithm under various solar
irradiation. Sol. Energy 2017, 148, 128–139.
24. Ishaque, K.; Salam, Z.; Lauss, G. The performance of perturb and observe and incremental conductance
maximum power point tracking method under dynamic weather conditions. Appl. Energy 2014, 119, 228–236.
[CrossRef]
25. Xiong, R.; Cao, J.; Yu, Q.; He, H.; Sun, F. Critical Review on the Battery State of Charge Estimation Methods
for Electric Vehicles. IEEE Access 2018, 6, 1832–1843. [CrossRef]
26. Bani Ahmad, A.; Ooi, C.A.; Ishak, D.; Teh, J. State-of-Charge Balancing Control for ON/OFF-Line Internal
Cells Using Hybrid Modular Multi-Level Converter and Parallel Modular Dual L-Bridge in a Grid-Scale
Battery Energy Storage System. IEEE Access 2019, 7, 131–147. [CrossRef]
27. Hannan, M.A.; Lipu, M.S.H.; Hussain, A.; Saad, M.H.; Ayob, A. Neural network approach for estimating
state of charge of lithium-ion battery using backtracking search algorithm. IEEE Access 2018, 6, 10069–10079.
[CrossRef]
28. Shiau, J.K.; Ma, C.W. Li-Ion battery charging with a buck-boost power converter for a solar powered battery
management system. Energies 2013, 6, 1669–1699. [CrossRef]
29. Chang, F.; Roemer, F.; Baumann, M.; Lienkamp, M. Modelling and Evaluation of Battery Packs with Different
Numbers of Paralleled Cells. World Electr. Veh. J. 2018, 9, 8. [CrossRef]
30. Id, E.M.; Id, V.I.H.; Gandiaga, I.; Milo, A. Li-Ion Battery Lifetime Model ’ s Influence on the Economic
Assessment of a Hybrid Electric Bus’s Operation. World Electr. Veh. J. 2018, 9, 28.
31. Kavousi-Fard, A.; Niknam, T.; Fotuhi-Firuzabad, M. Stochastic Reconfiguration and Optimal Coordination of
V2G Plug-in Electric Vehicles Considering Correlated Wind Power Generation. IEEE Trans. Sustain. Energy
2015, 6, 822–830. [CrossRef]
32. Devarakonda, L.; Hu, T. Effects of rest time on discharge response and equivalent circuit model for a lead-acid
battery. J. Power Sources 2015, 282, 19–27. [CrossRef]
33. Pavlov, D. Invention and Development of the Lead–Acid Battery. In Lead-Acid Batteries: Science and Technology;
Elsevier: Amsterdam, The Netherlands, 2017; pp. 3–32.
34. Dell, R.M.; Moseley, P.T.; Rand, D.A.J. Batteries and Supercapacitors for Use in Road Vehicles. In Towards
Sustainable Road Transport; Elsevier: Amsterdam, The Netherlands, 2014; pp. 217–259.
Energies 2020, 13, 3434 17 of 17

35. Lee, S.; Cherry, J.; Safoutin, M.; McDonald, J. Modeling and Validation of 12V Lead-Acid Battery for Stop-Start
Technology. In Proceedings of the SAE INTERNATIONAL, Michigan, MI, USA, 4–6 April 2017; pp. 1–11.
36. Zhang, C.; Jiang, J.; Zhang, L.; Liu, S.; Wang, L.; Loh, P.C. A generalized SOC-OCV model for lithium-ion
batteries and the SOC estimation for LNMCO battery. Energies 2016, 9, 900. [CrossRef]
37. Campestrini, C.; Kosch, S.; Jossen, A. Influence of change in open circuit voltage on the state of charge
estimation with an extended Kalman filter. J. Energy Storage 2017, 12, 149–156. [CrossRef]
38. Baccouche, I.; Jemmali, S.; Manai, B.; Omar, N.; Amara, N.; Baccouche, I.; Jemmali, S.; Manai, B.; Omar, N.;
Amara, N.E. Ben Improved OCV Model of a Li-Ion NMC Battery for Online SOC Estimation Using the
Extended Kalman Filter. Energies 2017, 10, 764. [CrossRef]
39. Wang, W.; Wang, X.; Xiang, C.; Wei, C.; Zhao, Y. Unscented kalman filter-based battery SOC estimation
and peak power prediction method for power distribution of hybrid electric vehicles. IEEE Access 2018, 6,
35957–35965. [CrossRef]
40. Ren, H.; Zhao, Y.; Chen, S.; Wang, T. Design and implementation of a battery management system with
active charge balance based on the SOC and SOH online estimation. Energy 2019, 166, 908–917. [CrossRef]
41. Qays, O.; Buswig, Y.; Anyi, M. Active Cell Balancing Control Method for Series-Connected Lithium-Ion
Battery. Int. J. Innov. Technol. Explor. Eng. 2019, 8, 2424–2430.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).