Applied Energy 186 (2017) 28–45

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Optimal energy cost and economic analysis of a residential

grid-interactive solar PV system- case of eThekwini municipality in
South Africa
B.P. Numbi ⇑, S.J. Malinga
Centre for the Development of Green Technologies, Department of Electrical Engineering, Mangosuthu University of Technology, 511 Mangosuthu Highway, Umlazi, 4031,
P.O. Box 12363, Jacobs, 4026 Durban, South Africa

h i g h l i g h t s

Optimal energy cost control of a residential grid-interactive PV system is modeled.

There is potential of more than 92% cost saving and payback period of less than 14 years.

The higher the feed-in tariff (FIT) the lesser the need for a battery bank.

Use of battery storage improves the cost effectiveness with low FITs unlike with high FITs.

The higher the grid electricity price, the higher the profitability of the system.

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, an optimal energy control of a 3 kW residential grid-interactive solar PV system is pre-
Received 2 June 2016 sented. A typical two-bedroom household within the eThekwini municipality in the city of Durban,
Received in revised form 18 October 2016 South Africa, is considered. The aim is to assess the potential of energy cost saving and cost effectiveness
Accepted 19 October 2016
the system can achieve under the new residential feed-in tariff (FIT), referred to as residential embedded
Available online 5 November 2016
generation tariff. Simulation results show that a potential of 69.41% energy cost saving and payback per-
iod (PBP) of 19 years can be achieved under the current FIT. By increasing the current FIT by 58%, the
Keywords:
energy cost saving is increased by 22.8%, while the period is decreased to 14 years. When the FIT is
Grid-interactive solar PV
Optimal energy cost
increased by 71.25% from its current value, a 75% increase in the cost saving is achieved with a PBP of
Economic analysis 8.6 years. However, it is shown that the higher the FIT the lesser the need for a battery bank if grid failures
Residential electricity customers are not considered as a constraint.
Feed-in tariff The impact of battery storage and grid electricity price on the profitability of the grid-interactive solar
PV system is also analyzed. It is shown that the use of battery storage system will have a positive effect on
the system profitability only when the FIT is not attractive. For attractive FITs, the use of battery storage
will have a negative effect on the profitability of the grid-interactive solar PV system. Simulation results
also show that the higher the grid electricity price the higher the profitability of the system.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction the reliability and efficiency of coal-fired power stations, a well-

In South Africa, about 85% of the electricity demand is supplied
from coal-fired power plants [1]. Coal has been used for a consid-
erable number of years as the main energy fuel in South Africa
because of reasons such as the abundance of coal reserves in the
country, the cost effectiveness of generating electricity from coal,
⇑ Corresponding author.
E-mail address: NumbiB@mut.ac.za (B.P. Numbi).
established infrastructure for electricity production from coal, etc.
However, due to the outdated power stations, reduced coal
reserve margins and high investment cost of new coal-fired power
stations, Eskom, the main electricity supplier, is currently experi-
encing a big challenge to meet the continual growth of energy
demand in South Africa, resulting in load shedding and power
blackouts [1,2]. Being part of the Kyoto protocol, South Africa is
also committed to reduce the carbon emission from its coal-fired
power stations.

http://dx.doi.org/10.1016/j.apenergy.2016.10.048
0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.
 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45
In an attempt to meet the national energy demand, Eskom has
built open cycle gas turbines based on diesel fuel which were con-
ceptually aimed to operate for short periods, only during peak
demand times [2], usually between 7h00 to 10h00 and 18h00 to
20h00 [3]. However, because of the reduced availability level of
the current coal-fired power stations, due to the reasons men-
tioned earlier, these gas turbines are operated for longer periods
than initially planned [2]. Operating the aged coal-based power
generation equipment that require expensive maintenance, and
also the consumption of large amounts of diesel fuel required by
gas turbines, would justify the steadily increase of electricity price
by Eskom. In eThekwini municipality,1 for instance, the electricity
cost has been increased by 9.80% from 2012 to 2013, by 4.95% from
2013 to 2014, by 6.80% from 2014 to 2015 and by 12.20% from 2015
to 2016 [4]. Moreover, with the forthcoming implementation of the
carbon tax in South Africa [5], the electricity cost is expected to ramp
up owing to the fact that coal-fired power plants are one of the big-
gest greenhouse gas emission sources.
On the one hand, to reduce the demand pressure on the national
power grid, South Africa, similar to other countries in the world,
has encouraged electricity customers and independent power pro-
ducers (IPPs) to generate power from alternative energy sources,
especially those which are renewable and sustainable, such as
solar photovoltaic (PV) and wind, and sell it back to the grid. In
the first phase, connection to the national grid of small-scale
embedded generation systems not exceeding 100 kW has been
approved by the National Energy Regulator of South Africa
(NERSA). Municipalities are allowed to import electricity from
these energy systems at the same rate as Eskom’s electricity retail
price, referred to as the Mega Flex rate [6]. However, up to now,
only the eThekwini municipality has agreed on the electricity pur-
chase from its residential customers, provided that the connection
capacity does not exceed 4.6 kVA for a single-phase system and
13.8 kVA for a three-phase system [4].
On the other hand, because of the steady increase in the elec-
tricity cost, South African electricity customers are showing great
interest to install distributed generation (DG) systems such as solar
PV and wind. Grid-connected solar PV is seen to be the fastest
growing DG resource in South Africa due to the high solar resource
the country has, combined with the substantial drop in the initial
investment cost. The average daily solar radiation in South Africa
is between 4.5 and 6.5 kW h/m2/day [7], which is higher than that
of most countries in the world. In less than one month, for instance,
the Authors, through the Centre for the Development of Green
Technologies (CDGT), had the opportunity to work on the design
of about 550 kWp rooftop grid-connected PV systems to be
installed on commercial buildings in the province of KwaZulu-
Natal.
The application of grid-connected PV systems has found more
market in commercial and industrial sectors than in the residential
sector. For urban residential buildings, however, a grid-connected
solar PV system might not be cost effective owing to the fact that
most electricity usage will not be offset by the investment cost of
the solar PV system. The reason is that, unlike with industrial
and commercial customers, the need for electricity in urban resi-
dential buildings is greater during the evening than during the
day, whilst the PV panels produce power only during the day. To
solve this issue, two approaches are possible. The first one is to sell
the excess power to the national grid during the day when the res-
idential load demand is insignificant. However, selling a large
amount of PV excess power to the eThekwini municipality distri-
bution network will lead to a less cost-effective residential solar
1
EThekwini is one of the eleven districts of the KwaZulu-Natal province in South
Africa.
29
PV system because of the less attractive Eskom’s feed-in tariff (FIT),
in contract to attractive FITs provided in countries such as Canada
and Germany [8]. In South Africa, the FIT already takes into account
the interconnection fee, so that no additional cost is charged to
customers when feeding their energy to the grid. Eskom’s electric-
ity retail price is half of the price at which residential customers
import power from eThekwini municipality distribution network.
The second solution is to store the excess power during the day
and use it later in the morning and evening when the load demand
is high. This option gives residential customers the freedom and
flexibility to control, at any time, the bi-directional power flow
between their buildings and the grid (import and export), unlike
in a grid-connected PV system where the power flow is unidirec-
tional [9]. This configuration is usually referred to as grid-
interactive solar PV system. Because of energy storage, an addi-
tional advantage of the grid-interactive PV system is that residen-
tial customers can benefit from the time-based pricing, such as the
time-of-use (TOU) electricity tariff, through peak shaving, to fur-
ther reduce their electricity bill cost. In South Africa, NERSA
advised that priority to export to the grid must be given to cus-
tomers on the TOU tariff as this minimizes the national grid peak
load. However, an optimal energy control scheme is needed to
achieve minimal electricity cost by meeting simultaneously phys-
ical and operational constraints of the system.
The grid-interactive solar PV system is still a new technology
within PV systems. This was designed to combine both advan-
tages of grid-connected and off-grid solar PV systems, in the
sense that it supplies power to the user, feeds the excess to the
grid, boosts grid stability and also provides back-up power during
grid failure periods [10]. However, there are only few research
studies on grid-interactive PV systems. Research papers, such as
[11–13] have dealt with the optimal energy control of off-grid
hybrid energy systems with the aim to minimize the diesel
energy cost of load demands in remote areas. Authors in Ref.
[14] propose an operational power scheduling based on the
‘‘IF-THEN” rules, for a microgrid, including PV generators with
embedded storage units and a gas microturbine. The objective
is to meet the demand load by using the solar PV generator as
the prior energy source in such a way to minimize the power
generated by the gas microturbine. The study in Ref. [15] pro-
poses a rule-based control method for a battery energy storage
system integrated with a hybrid PV/wind energy system to make
these renewable energy sources more dispatchable. In Ref. [16],
an optimal energy management of a battery energy storage sys-
tem in a grid-connected microgrid is proposed with the aim to
balance the demand load and the renewable sources, by minimiz-
ing the cost of electricity imported from the grid. However, the
study concentrates on a large-scale grid-connected renewable
energy system. The authors in Ref. [17] present an optimal man-
agement model of a grid-connected hybrid PV/fuel/battery energy
system with the energy purchasing cost and CO2 emission as
objective functions to be minimized. However, the FIT, ToU tariff
and economic analyses are not included. In Ref. [18], an optimal
energy control of a grid-connected hybrid energy system is inves-
tigated to minimize mainly the diesel fuel energy cost on one of
South Africa’s hotels. Although the optimal energy control of a
grid-connected solar PV-battery hybrid system is studied in
[19,20], the configuration used does not match with the definition
of grid-interactive solar PV system. The grid-interactive PV sys-
tem owes its definition from its bi-directional energy transfer
capability provided by the voltage source inverter [9]. The avail-
ability of pre-assembled grid-interactive PV systems makes the
technology to be cost effective in terms of capital costs. Also,
the fact that most grid-interactive PV systems come as pre-
assembled packages adds a special constraint to the system
design. One of these constraints is the battery bank capacity that
30 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45
depends on the charge controller rating. In Ref. [21], an optimal
power flow management of a grid-interactive PV system is pro-
posed. However, a dynamic programming (DP) technique is
employed and the study does not cover the economic analysis.
The DP has a weakness of requiring a large memory for problems
studied under a long prediction horizon and discretized with a
short sampling period [22].
From the above-mentioned works, it can be seen that there
have been less research works conducted with the aim to study
the economic viability of a grid-interactive solar PV system for res-
idential applications. Hence, there is a need of extensive economic
studies of residential grid-interactive PV systems in order to boost
the PV market in residential sectors. With this in mind, the focus of
this work is to develop an optimal control model to solve the
energy management problem of a grid-interactive solar PV system.
The cost effectiveness of the system is also analyzed under the TOU
tariff and FIT of the eThekwini municipality. A two-bedroom
household within the city of Durban in South Africa is used as a
case study.
This paper is organized as follows: Section 2 presents the opti-
mal control model of the grid-interactive solar PV system; Section 3
presents data for the case study used in this work. The baseline
model is presented in Section 4. Simulation results of the case
study are discussed in Section 5 with the aim to evaluate the effec-
tiveness of the developed model. The economic analysis of the sys-
tem is given in Section 6. Conclusions and recommendations are
finally presented in Section 8.
2. Model development
2.1. System description
The configuration/power flow of a two-stage grid-interactive
solar PV system is shown in Fig. 1, where the maximum power
point tracking (MPPT) is achieved by the first converter (DC-DC
converter), and the grid interactive operation by the second con-
verter (DC-AC bi-directional inverter) [9]. The two-stage grid-
interactive inverter offers a much easier control than the single-
stage structure, although this may lead to a higher system cost
and power losses [9].
As shown in Fig. 1, the total power generated by PV panels, PPV ,
is fed to the MPPT converter for optimum PV power extraction at
different weather conditions. On the one hand, if the PV array is
producing more power than what is required, the maximum PV
power, PMP , from the MPPT converter is split into two components
namely, the DC input power to the bi-directional inverter, P þ DC , and
the battery charging power, P þ B . After converting P þ
DC to AC power,
an amount of power, P LC , is fed to the sub-distribution board
where the critical loads, such as lighting, refrigerator, television
and DStv decoder are connected, and the remaining power, Pþ
AC ,
is fed to the main distribution board, where the non-critical loads,
such as the electric water heater and stove are connected. The non-
critical demand load, PLNC , is first fed and the excess power, PEXP , is
exported to the national grid under the FIT. When the PV power
output is not sufficient to meet the load, the battery bank is dis-
charged at a rate of P 
B as a supplement. On the other hand, during
the night where the PV power output is zero, the total load and
battery bank are supplied by the grid imported power, PIMP , under
TOU tariff. In this case, an amount of P IMP is supplied to the non-
critical load and the remaining, P AC is fed to the bi-directional
inverter. P 
AC is therefore used to feed the critical load, while the
remaining power is converted into DC power, P  DC . The negative
sign () for the bi-directional inverter power means the device is
operated in the reverse mode, while for the battery bank, this sym-
bolises the discharging mode.
2.2. Optimal control model of the grid-interactive solar PV system
Optimal control approach is a powerful tool that is proposed in
the literature to solve several energy management problems. In
Ref. [23] for instance, an optimal load shifting model to reduce
the energy cost of a colliery is investigated. It is shown that about
49% energy cost saving can be achieved during 5 weekdays in a
high-demand season. In Ref. [24], the authors have shown that a
potential of 37.38% energy cost saving can be achieved on a coal
belt conveying system when optimal control scheme is applied.
In Ref. [25], optimal control is applied to a deep level mine twin
rock winder system for energy cost reduction. It is found that
scheduling the hoist under TOU tariff can lead to a potential of
30.8% energy cost saving. Authors in Ref. [26] have shown that
the application of optimal control to the energy management of
a jaw crushing process in deep mine can lead to more than 50%
energy cost saving. A systems optimization is proposed in Ref.
[27] to reduce the energy cost of a parallel HPGR crushing process.
The authors have shown that a daily energy cost saving of 41.93%
can be achieved. In Ref. [28], the energy cost of a crushing circuit
based on vertical shaft impactor is minimized. It is concluded that
about 49.7% energy cost saving is possible to be achieved.
Optimal control tool was traditionally designed to solve
continuous-time problems by using the Pontryagin’s maximum
principle [29]. However, solving a continuous-time optimal control
problem by this principle will assume both objective and con-
straints to be continuously differentiable, referred to as smooth
condition. The TOU electricity tariff being of discrete nature, this
will lead the optimal energy control problem in this work to be
continuous but not differentiable. Also, because of the complexity
of the problem, characterized by several constraints, a numerical
approach is used as an alternative.
2.2.1. Objective function
The control objective to be minimized is the net electricity cost,
J C , under a given period. This is defined as the difference between
the electricity cost due to the power imported from the grid, PIMP ,
and the electricity cost due to the power exported to the grid, P EXP .
This is expressed mathematically as follows:
N 
X 
JC ¼ pj PIMPj  cP EXPj ts ; ð1Þ
j¼1
where j is the jth sampling interval, N is the total number of sam-
pling intervals, t s is the sampling time, p is the TOU electricity tariff
and c is the FIT.
2.2.2. Constraints
2.2.2.1. Power balance. The power balance is one of the most impor-
tant constraints in electrical circuits that need to be met. By
neglecting all converter power losses, the power balance con-
straints to be satisfied at different nodes of the system, are
expressed as follows:
8
< PMPj ¼ PDC j þ P Bj ; ð1 6 j 6 NÞ;
>
PDC j ¼ PLC j þ PAC j ; ð1 6 j 6 NÞ; ð2Þ
>
:P þ P ¼P þP LNC j ; ð1 6 j 6 NÞ:
AC j IMPj EXPj
2.2.2.2. Dynamics of battery state of charge. During charging and
discharging, the state of charge, SoC, of the battery bank has to
be maintained between its minimum and maximum values,
SoC min and SoC max , respectively. The SoC of the battery bank in
Fig. 1 can be expressed in the discrete-time domain as follows:
gB ts
SoC j ¼ SoC ðj1Þ þ PBðj1Þ ; ð1 6 j 6 NÞ; ð3Þ
Cn
 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45 31

Fig. 1. Schematic diagram of a grid-interactive solar PV system.

where SoC j is the SoC at the current sampling interval, j; SoC ðj1Þ is 2.2.2.4. Power flow limitations. For equipment safety purposes, all
the SoC at the previous sampling interval, ðj  1Þ; C n is the nominal power flows should be kept within minimum and maximum limits
capacity of the battery bank in kW h and gB is the round trip effi- according to the design specifications given by the manufacturer.
ciency of the battery. However, since the SoC at the current sam- These boundary constraints are expressed as follows:
pling interval has to be expressed only in terms of the battery
8 min
power, PB , which is one of the control variables, SoC ðj1Þ has to be >
> PEXP 6 PEXPj 6 Pmax
EXP ; ð1 6 j 6 NÞ;
>
>
eliminated from Eq. (3). By a recurrence reasoning, the battery bank >
> min max
>
> PIMP 6 PIMPj 6 P IMP ; ð1 6 j 6 NÞ;
SoC at the current sampling interval can be expressed as a function >
>
>
< Pmin 6 PAC 6 Pmax ; ð1 6 j 6 NÞ;
of PB and its initial value, SoC 0 , which is constant. This results in the AC AC
ð7Þ
j
following expression: >
>
> Pmin max
DC 6 P DC j 6 P DC ; ð1 6 j 6 NÞ;
>
>
> min
>
gB ts X
j >
> PB 6 PBj 6 PmaxB ; ð1 6 j 6 NÞ;
>
> min
SoC j ¼ SoC 0 þ P Bi ; ð1 6 j 6 NÞ: ð4Þ :
Cn PMP 6 PMPj 6 PmaxMP ; ð1 6 j 6 NÞ:
i¼1

With this, the constrains to maintain the SoC dynamics within spec-
ified values are written as follows: 2.2.2.5. Exclusive power flow between grid and customer. Since the
customer cannot import and export power at the same time, the
product between P IM , and PEXP , has to be zero. This is mathemati-
gB t s X
j
SoC min 6 SoC 0 þ PBi 6 SoC max ; ð1 6 j 6 N Þ: ð5Þ cally written as follows:
Cn i¼1
PIMj  PEXPj ¼ 0; ð1 6 j 6 NÞ: ð8Þ

2.2.2.3. PV power output. At any time, the DC power handled by the

MPPT converter should be less than or equal to the available PV
power output, determined by the PV module characteristics, such
as the PV power output at maximum power point and PV power
temperature coefficient, and weather conditions, such as tempera-
ture and incident irradiation. This is expressed as:
PMPj 6 PPV j ; ð1 6 j 6 N Þ:
2.2.2.6. Fixed-final state condition. In order to allow for repeated
implementation of the optimal energy control of the grid-
interactive PV system, the remaining amount of energy in the bat-
tery bank at the end of the control horizon should be equal to the
energy amount at the beginning of the control horizon. This is
equivalent to equating the SoC at the last sampling interval,
ð6Þ SoC N , to the initial condition, SoC 0 . This requirement leads to the
following constraint:
32 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45
X
N
PBi ¼ 0:
i¼1
3. Case study description
To evaluate the effectiveness of the model developed in Sec-
tion 2, a grid-interactive PV system was installed in a single-
phase two-bedroom household in Durban. The load profile of this
household, and size of each component of the PV system are pre-
sented in this section.
3.1. Load profile
Two types of load are generally connected to grid-interactive
PV system. These are the critical and non-critical demand loads.
The critical loads are those of high priority that cannot be inter-
rupted, shifted or reduced, while the non-critical loads can be
altered without causing considerable prejudice to the household.
The critical load in the household under study consists of the
lighting, refrigerator (with freezer), computers, television, and
DStv decoder, while the non-critical load consists of a washing
machine, stove, electric water heater, referred to as a geyser in
South Africa. In this household, the non-critical load is zero
during the day since the electric water heater is controlled by a
programmable timer and is switched on only between 04 : 00 to
06 : 00, while the evening and night non-critical load is predom-
inated by the stove and washing machine energy consumption.
During the day, only a small amount of energy is consumed by
the critical load since only the refrigerator is operated, while most
critical load appliances are operated during the evening and
night.
The load profiles of both critical and non-critical loads have
been obtained by using Single-phase Efergy E2 Classic energy mon-
itors, which provides only 2% measurement error. 2 Fig. 2 shows
the load profiles of the two types of loads.
3.2. Electricity tariffs
The electricity tariff of the eThekwini municipality in the pro-
vince of KwaZulu-Natal is used [4]. The FIT incentive for residential
embedded generation used in this municipality is USD 0:46/kW h.
The introduction of the demand-side management programme in
South Africa has allowed customers to have the freedom to sub-
scribe either to flat or time-of-use (TOU) tariffs. In contrast to the
flat tariff, with a TOU tariff, customers are charged at different rates
depending on the time of day the electricity is used. Furthermore,
the TOU tariff also varies depending on the season. In the high
demand season (winter), the TOU tariff is higher than that in a
low demand season. However, this only applies to commercial
and industrial customers. For residential customers, a non-
seasonal TOU tariff is used [4].
The TOU tariff, PðtÞ, for residential customers in summer (from
1 September to 31 May) is given as follows:
8 based on the maximum of charging and discharging currents,
< USD 0:06=kW h from ð0  6h00Þ and ð22  24h00Þ;
> respectively, 25 A and 80 A, and the charge controller nominal volt-
pðt Þ ¼ USD 0:08=kW h from ð6  7h00Þ; ð10  18h00Þ and ð20  22h00Þ;
>
:
USD 0:16=kW h from ð7  10h00Þ and ð18  20h00Þ:
ð10Þ
The TOU tariff for winter (from 1 June to 31 August) is given as
follows:
2
http://efergy.com/manuals/e2classicinstructionsweb2011.pdf.
8
< USD 0:06=kW h from ð0  6h00Þ and ð22  24h00Þ;
>
ð9Þ
pðt Þ ¼ USD 0:08=kW h from ð9  17h00Þ and ð19  22h00Þ;
>
:
USD 0:16=kW h from ð6  9h00Þ and ð17  19h00Þ:
ð11Þ
In Eqs. (10) and (11), the first row denotes the off-peak tariff and
periods, the second row denotes the standard tariff and periods,
and the third one is the peak tariff and periods. USD is the United
State Dollar 3 and t is the time of any weekday in hours.
3.3. PV peak capacity and grid-interactive inverter sizing
The size of a grid-interactive solar PV system is mainly dictated
by the budget, energy saving target, available mounting space and
the maximum penetration to the utility grid. In this work, the max-
imum penetration is taken as the size selection criterion. To benefit
from the FIT incentive offered by eThekwini municipality, a grid-
interactive PV system with a capacity not exceeding 4.6 kWp is
to be installed for single-phase residential buildings. With this,
the minimum and maximum limits on power exchanged with
the grid are set, respectively, as P min min
EXP ¼ P EXP ¼ 0 kW and
max max
PEXP ¼ PEXP ¼ 4:6 kW.
The 3 kW IMEON grid-interactive inverter [30] is proposed to be
used. A PV array with peak power of 3 kWp is to be installed,
according to the IMEON inverter manufacturer. The power flow
limits on the MPPT boost converter are set to Pmin
MP ¼ 0 kW and
Pmax
MP ¼ 3 kW and for the bi-directional inverter, to
Pmin min
DC ¼ P AC ¼ 3 kW and Pmax max
DC ¼ P AC ¼ 3 kW. By choosing a
YL260P-29b (260 W) PV module [31], the total number of modules
to be installed is obtained to be 12. Since the nominal voltage of
the YL260P-29b PV panel at maximum power point (MPP) is
30.9 V, the connection of 12 panels in series to the IMEON 3.6
inverter will lead to a total voltage of 370.8 V, which satisfies the
DC input voltage range of the IMEON inverter (between 150 and
510 V).
3.4. Battery bank sizing
In grid-interactive PV systems the size of the battery bank
depends on the charge controller rating, unlike in off-grid solar
PV systems where the battery size depends mainly on the demand
load to be backed up [32]. Based on best practice, the battery bank
in grid-interactive PV systems is sized in such a way that the charg-
ing current is between 10% and 20% of the total battery capacity.
Since the maximum charging current of the IMEON 3.6 charge con-
troller is 25 A, a maximum battery capacity of 250 A h can be used.
In this study, a total battery capacity, Cn, of 200 A h is considered.
12–100 (12 V100 A h) AGM deep cycle battery from CB solar
[33] is proposed. For a nominal charging voltage of 48 V and
required battery capacity of 200 A h, 4 batteries are connected in
series and 2 strings of batteries are connected in parallel. Hence,
the battery capacity can be expressed in kW h as
200  48 ¼ 9:6 kW h. The battery power flow limits are calculated
age (48 V). Hence, the battery maximum charging and discharging
power flow limits are obtained to be P max ¼ 48 V  25 A = 1.2 kW
B
and P min
B ¼ 48 V  80 A = 3.84 kW. The round trip battery effi-
ciency, gB , is assumed to be 90%.
3.5. PV power output
3
The initial currency was in South African currency Rand ‘‘R”, and was converted in
USD: 1 USD = 13.5 R.
 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45
Critical load
Non-critical load
Power consumption [kW]
3
0
0 5
Fig. 2. Hourly power consumption.
The PV power output is predicted based on the model proposed
in Ref. [21] as follows:
It
PPV ¼ P PV;STC NPVs NPVp ½1  aðT c  25Þ;
1000
with
It
Tc ¼ Ta þ ðNOCT  20Þ;
800
where P PV is the PV power output, P PV;STC is the PV output at the
maximum power point and standard test condition (STC), NPVs
and N PVp are, respectively, the number of PV panels in series and
parallel, It is the solar irradiance on a tilted surface, a is the temper-
ature coefficient of power, T c is the cell temperature, T a is the ambi-
ent air temperature, and NOCT is the nominal operating cell
temperature. It is calculated based on the following equation [1,34]:
   
1 þ cosb 1  cosb
It ¼ Id coshb þ Idif þ qI g ; ð14Þ
2 2
where Idif is the diffuse horizontal irradiance (DHI), Id is the direct
normal irradiance (DNI), Ig is the global horizontal irradiance
(GHI), b is the tilted angle, hb is the incidence angle of solar radiation
on a tilted surface, q is the reflectance of the surrounding area.
The different values of all input parameters used to predict the
PV power output are given as follows: P PV;STC ¼ 0:26 kWp,
N PVs ¼ 12; N PVp ¼ 1; a ¼ 0:0045/°C, NOCT ¼ 46 °C [31], q ¼ 0:2; b ¼
30 (optimal average value between summer and winter seasons
for a fixed PV mounting structure installed in Durban). Data for
DHI, DNI, GHI and air ambient temperature are plotted in Figs. 3
and 4 and were collected from the weather station located at the
University of KwaZulu-Natal, Howard College (latitude:
29:87098 , longitude: 30:97695 and elevation: 150 m) [35].
4. Baseline
In order to study the effectiveness of the developed model, the
net energy cost achieved through optimal energy control of the
grid-interactive PV system is compared to the energy cost incurred
by the residential customer without solar PV system and operating
on a flat tariff of USD 0:11/kW h (Residential flat tariff in eThekwini
municipality).
33
10 15 20 25
Time [h]
5. Results and discussion
Since the model developed in Section 2 is a nonlinear program-
ming problem due to constraint 8, the case study presented in Sec-
ð12Þ
tion 3 is simulated using the fmincon optimization solver of the
Matlab optimization toolbox. The simulation is performed for a
single day of both summer (01 December 2015) and winter (01
August 2015) seasons, where the load profile is assumed to be
the same. The control horizon of 24 h is therefore considered.
ð13Þ
The annual energy cost is calculated by multiplying the daily
energy cost by the number of days in the 2015 year (273 days in
summer and 92 days in winter). A sampling time of 30 min is used,
which leads to a total number of samples per variable of
N ¼ 2460 min
¼ 48. The initial state of charge of the battery bank
30 min
is set to 50% of its nominal value, the maximum state of charge
SoC max is set to 100% of the nominal value, the minimum state of
charge SoC min is set to 20% of the nominal value, which corresponds
to a depth of discharge of 80%.
To analyze the impact of the electricity price on the cost effec-
tiveness of the residential grid-interactive solar PV system, the
simulation is performed under different FIT values. Case I is where
the actual value of the FIT is used (c ¼ 0:046 USD/kW h). Case II is a
scenario where the FIT is increased and assumed to be equal to the
flat electricity tariff (c ¼ 0:11 USD/kW h), while in case III, the FIT is
further increased and assumed to be equal to the peak TOU elec-
tricity price (c ¼ 0:16 USD/kW h).
Optimal power flow solutions are presented in Figs. 5–8 for case
I, in Figs. 10–13 for case II and in Figs. 15–18 for case III, while the
corresponding performance indicators are, respectively, presented
in Figs. 9, 14, 19 and Table 1. In all the three cases (I, II and III), the
legend of the first graph shows the time of use periods, which also
applies to the rest of the graphs.
5.1. Case I: Optimal power flow with feed-in tariff, c = USD 0.046/kW h
Simulation results under this scenario are presented in Figs. 5–
9. It is shown that during the morning off-peak period when the
energy from the sun is zero, the small amount of energy required
by the critical load from 00h to 04h00 is mainly supplied from
the battery bank (P B ). The discharge rate of the battery bank is
shown through the state of charge, SoC, in the second graphs of
Figs. 6 and 8. When the electric water heater is switched on from
4h00 to 6h00, the same figures show that the battery bank contin-
ues to discharge but does not reach its lowest SoC limit. Since only
34 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45

Global horizontal

Irradiance [W/m2 ]
1000
Direct normal
Diffuse horizontal
500

0
0 5 10 15 20 25
Time [h]

28
Temperature [ o C]

26

24

22

20
0 5 10 15 20 25
Time [h]
Fig. 3. Solar irradiance and ambient temperature during summer (01 December 2015, that is, n = 335).

1000
Irradiance [W/m2 ]

Global horizontal
800
Direct normal
600 Diffuse horizontal

400

200

0
0 5 10 15 20 25
Time [h]

20
Temperature [ o C]

15

10

0 5 10 15 20 25
Time [h]
Fig. 4. Solar irradiance and ambient temperature during winter (01 August 2015, that is, n = 213).

a limited amount of energy is drawn from the battery bank, the
remaining amount of energy is imported from the grid to meet
the load demand. However, a comparison analysis between Figs. 6
and 8 shows that the battery bank is more discharged in summer
than in winter. The reason is that since the solar energy is lower
in winter, the optimal energy controller prevents the batteries
from deeply discharging so that more solar energy is sold to the
grid by reducing the need of charging the batteries during daytime
(from 06h00 to 18h00 during summer and from 08h00 to 17h00
during winter). This allows the battery bank to have enough energy
reserve so as to meet as much the evening peak energy demand as
possible. On the other hand, due to a higher solar energy during
summer, the battery bank can still supply energy to the morning
off-peak load and be fully charged at 18h00 to reduce the evening
peak energy demand.
To reduce the peak energy demand from the grid, PIMP , the bat-
tery bank is discharged at a high rate during both summer and
winter evening peak periods, while only a small amount of evening
peak energy is provided from the grid to meet the total load
demand. However, it is shown that during the evening standard
 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45 35

off-peak standard peak

P MP [kW]
2

-1
0 5 10 15 20 25

P DC P AC
Power flow [kW]

-2

0 5 10 15 20 25

3 P EXP P IMP
Power flow [kW]

-1
0 5 10 15 20 25
Time [h]

Fig. 5. Optimal power flow for summer season – case I.

2
P B [kW]

-2

0 5 10 15 20 25

100
Battery SoC [%]

50

0
0 5 10 15 20 25
Time [h]
Fig. 6. Battery optimal power flow and state of charge for summer season – case I.

period, the battery bank is rather discharged at a much lower rate
in such a way as to meet the fixed-final state condition expressed
by Eq. (9), at the end of the control horizon.
The cost performance shown in Fig. 9 demonstrates that from
00h00 to 4h00 the baseline cost and the optimal cost are the same
since no significant load is drawn. The cost saving starts to occur
from 04h00 when the electric water heater is switched on and this
is supplied from the battery bank at no cost, while the remaining
energy is imported from the grid at a cheaper cost compared to
the flat tariff. The net energy cost is further reduced from the
36 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45

off-peak standard peak

P MP [kW]
2

-1
0 5 10 15 20 25

P DC P AC
Power flow [kW]

-2

0 5 10 15 20 25

P EXP P IMP
3
Power flow [kW]

-1
0 5 10 15 20 25
Time [h]
Fig. 7. Optimal power flow for winter season – case I.

2
P B [kW]

-2

0 5 10 15 20 25
Battery SoC [%]

100

50

0
0 5 10 15 20 25
Time [h]
Fig. 8. Battery optimal power flow and state of charge for winter season – case I.

moment the sun rises, as a result of the green energy sold to the
grid. The cost saving decreases in the evening when solar energy
is not generated. The higher daily cost saving achieved during sum-
mer because of the longer sunshine duration compared to that in
winter, is also shown in the same figure. It can be seen in Table 1
that 69.41% annual cost saving is achieved. The effectiveness of
the optimal energy controller is also shown with regard to the con-
straints on power flow, battery state of charge and fixed-final state
condition. As previously discussed, the equality between the
amount of energy in the battery bank at the beginning and at the
end of the control horizon will allow the optimal energy control
scheme to be repeatedly implemented in practice.
 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45 37

2.5
baseline
2 optimal control-summer - case I

Net energy cost [USD]

optimal control-winter - case I
1.5

0.5

-0.5
0 5 10 15 20 25
Time [h]

Fig. 9. Cumulative net energy cost – case I.

off-peak standard peak

3
P MP [kW]

2
1
0
-1
0 5 10 15 20 25

P DC P AC
Power flow [kW]

-2

0 5 10 15 20 25

P EXP P IMP
Power flow [kW]

3
2
1
0
-1
0 5 10 15 20 25
Time [h]
Fig. 10. Optimal power flow for summer season – case II.

5.2. Case II: Optimal power flow with feed-in tariff, c = USD 0.11/kW h
With the feed-in tariff being increased to USD 0:11/kW h,
Figs. 11 and 13 show that unlike in case I, during the morning
off-peak period (from 00h to 6h00), the grid is used to supply
power to both the battery bank and the morning off-peak energy
demand. In case II, the battery bank is rather charged because of
a more attractive feed-in tariff offered. With a higher feed-in tariff,
the optimal energy controller allows the batteries to charge during
the morning off-peak period at a cheaper cost so that the energy
stored can be later sold to the grid, in order to minimize the net
energy cost. During daytime, it can be seen in the second graphs
of Figs. 11 and 13 that the power generated by the solar PV system
is sold to the grid and the remaining is used to feed the daily crit-
ical load (mainly the fridge energy consumption) and charge the
battery bank as in case I. However, during summer, the battery
bank is not fully charged at the end of the day compared to case
I. One of the reasons is that the energy saving due to the evening
peak shaving is less attractive compared to that achieved through
selling a large amount of solar energy at an attractive FIT. During
38 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45

P B [kW]
0

-2

0 5 10 15 20 25

100
Battery SoC [%]

50

0
0 5 10 15 20 25
Time [h]

Fig. 11. Battery optimal power flow and state of charge for summer season – case II.

off-peak standard peak

3
P MP [kW]

-1
0 5 10 15 20 25

P DC P AC
2
Power flow [kW]

-2

0 5 10 15 20 25

3 P EXP P IMP
Power flow [kW]

-1
0 5 10 15 20 25
Time [h]

Fig. 12. Optimal power flow for winter season – case II.

winter, since the solar energy yield is low, priority is given to
charge the battery bank so as to supply the load during evening
peak and standard periods.
The effect of the increase in FIT on the seasonal power flows is
also shown when comparing the third graphs of Figs. 10 and 12.
The results in these figures show that a larger proportion of solar
energy is exported to the grid in summer. This could also be justi-
fied by the above-mentioned reasons in the sense that the larger
the available amount of solar energy the greater the energy sold
to the grid and lesser to charge the battery bank, and vice versa.
 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45 39

P B [kW]
0

-2

Battery SoC [%] 0 5 10 15 20 25

100

50

0
0 5 10 15 20 25
Time [h]
Fig. 13. Battery optimal power flow and state of charge for winter season – case II.

2.5
baseline
2
optimal control-summer-case II
1.5 optimal control-winter-case II
Net energy cost [USD]

0.5

-0.5

-1

-1.5
0 5 10 15 20 25
Time [h]

Fig. 14. Cumulative net energy cost – case II.

This means that with an attractive FIT, during periods of lower
solar PV power production, such as winter and cloudy days, a big-
ger capacity of battery bank is required as opposed to periods with
high solar PV power production.
The energy cost results are given as shown in Fig. 14 and Table 1.
When compared to case I, the results show that the increase in the
FIT by 58% (from USD 0:046/kW h to USD 0:11/kW h) leads to an
increase in the energy cost saving by 22.8% (92.2% in case II and
69.41% in case I).
5.3. Case III: Optimal power flow with feed-in tariff, c = USD 0.16/kW h
With the feed-in tariff being increased to be equal to the peak
TOU tariff (USD 0:16/kW h), simulation results presented in Figs. 15
and 16 and Figs. 17 and 18 show that during the morning off-peak
period, the charging rate of the battery bank is further increased
compared to case II. This is due to the same reason as in case II.
During the daytime, Figs. 16 and 18 show that the power to charge
the battery bank from the solar PV panels is greatly reduced com-
pared to that of case II. The reason for this is that since the FIT is
very attractive, the tendency of the optimal energy controller is
to sell all the solar energy to the grid to maximize the sales, and
hence minimize the net energy cost. This can be seen in the third
graphs of Figs. 15 and 17 where it is shown that the amount of
power exported to the grid, P IMP , during the day is almost equal
to the power generated, PMP , by the solar PV panels. Hence, the
optimal controller does not see the battery bank as a means to
reduce the energy cost through peak shaving as was the case in
previous scenarios. It is therefore worthy to note that the higher
the FIT the lesser the need for the battery bank if grid failures
are not considered as a constraint. Hence, with an attractive FIT,
the use of batteries in interactive-grid solar PV systems for residen-
tial sectors is only required for back-up power during grid failure
periods and not for energy cost saving. For countries where grid
40 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45

off-peak standard peak

P MP [kW]
2
1
0
-1
0 5 10 15 20 25

P DC P AC
Power flow [kW]

-2

0 5 10 15 20 25

P EXP P IMP
3
Power flow [kW]

-1
0 5 10 15 20 25
Time [h]
Fig. 15. Optimal power flow for summer season – case III.

2
P B [kW]

-2

0 5 10 15 20 25
Battery SoC [%]

100

50

0
0 5 10 15 20 25
Time [h]
Fig. 16. Battery optimal power flow and state of charge for summer season – case III.

failures are not common, an attractive FIT will improve the eco-
nomic viability of solar PV systems by eliminating the need of bat-
tery storage since the preference will be given to the grid to
consume the solar energy, whist customers benefit from the FIT
incentive.
Fig. 19 and Table 1 show that the increase in the FIT by 71.25%
(from USD 0:046/kW h to USD 0:16/kW h) leads to an increase in
the energy cost saving by 75% (144.4% in case III and 69.41% in case
I). An energy cost saving over 100 % means that the customer will
have a negative energy bill.
 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45 41

off-peak standard peak

P MP [kW]
2
1
0
-1
0 5 10 15 20 25

P DC P AC
Power flow [kW]

-2

0 5 10 15 20 25

P EXP P IMP
Power flow [kW]

3
2
1
0
-1
0 5 10 15 20 25
Time [h]
Fig. 17. Optimal power flow for winter season – case III.

2
P B [kW]

-2

0 5 10 15 20 25
Battery SoC [%]

100

50

0
0 5 10 15 20 25
Time [h]

Fig. 18. Battery optimal power flow and state of charge for winter season – case III.

6. Economic analysis
To analyze the cost effectiveness of a project, various economic
performance indicators can be used. These indicators include, but
are not limited to simple payback period (SPP), internal rate of
return (IRR), benefits-to-cost ratio (BCR) and life cycle cost (LCC)
[36]. On the one hand, although the SPP period has the advantage
of being the simplest economic analysis method, and also the easi-
est to understand and to calculate, the approach presents some
drawbacks. The SPP does not take into account the time value of
42 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45

3
baseline
2 optimal control-summer-case III

Net energy cost [USD]

optimal control-winter-case III
1

-1

-2

-3
0 5 10 15 20 25
Time [h]
Fig. 19. Cumulative net energy cost – case III.

Table 1
Annual energy cost saving.

Strategy Energy cost (USD/day) Energy cost (USD/year) Saving (%)

Baseline 2.417 882.205 /
CASE I: c ¼ 0:046 USD=kW h
Optimal control-summer 0.643 0.643  273 days = 175.539 /
Optimal control-winter 1.025 1.025  92 days = 94.300 /
Optimal control-total net cost / 269.839 69.41
CASE II: c ¼ 0:11 USD=kW h
Optimal control-summer 0.0413 0.0413  273 days = 11.275 /
Optimal control-winter 0.625 0.625  92 days = 57.500 /
Optimal control-total net cost / 68.775 92.20
CASE III: c ¼ 0:16 USD=kW h
Optimal control-summer 1.155 1.155  273 days = 315.315 /
Optimal control-winter 0.802 0.802  92 days = 73.784 /
Optimal control-total net cost / 389.099 144.10

money, while in reality money depreciates over time because of where PW TBav is the annual average PW TB , obtained as follows:
the inflation and other factors. Another disadvantage of the SPP
is that all cash flows beyond the payback period (PBP) are ignored PW TB
PW TBav ¼ : ð16Þ
since the project lifetime is not taken into account. With this, n
investors will not be aware of the profitability of the project. On
In Eqs. (15) and (16), PW TC denotes the PW of total costs, n is the
the other hand, methods such as IRR, BCR and LCC take into
project lifetime and PW TB is the PW of total benefits, which are
account both the time value of money and the project lifetime by
the annual cost savings less any annual costs discounted to a PW
discounting all future worth cash flows to a present worth (PW)
and incurred by the user during operation. In this work, the annual
cash flow. Since the SPP is defined as the ratio between the annual
costs are the O&M costs.
benefit (AB) and initial investment cost, a ‘‘true” PBP that takes into
The PW TC is the algebraic sum of the discounted costs associ-
account the time value of money and project lifetime can be esti-
ated with the initial investment costs, replacement costs and sal-
mated as the ratio between the PW of total costs (PW TC ) and the
vage value of the grid-interactive solar PV system. However, the
annual average of the PW of total benefits (PW TB ).
salvage values of all components are neglected since all compo-
Due to its simplicity, the ‘‘true” PBP method is proposed for the
analysis of the residential grid-interactive solar PV system in this
work. Table 2 below gives prices for components of the proposed Table 2
residential grid-interactive solar PV system. This quotation con- Bill of quantity of 3 kW grid-interactive solar PV system.
tains updated prices from one of the biggest PV system component Component description Quantity Net price Net price
suppliers in South Africa. (R) (USD)
The BoS in Table 2 is composed of components, such as wiring, Imeon 3.6 solar inverter 1 33,060 2448.88
connectors, DC and AC disconnect switches and protection devices, Solar aluminium structure 6 1200 88.88
etc. YL260P-29b Yingli solar PV 12 26,400 1955.55
panel
As previously discussed, the ‘‘true” PBP is estimated using the
AGM deep-cycle battery 8 16,416 1216
following equation: Balance of system (BoS) – 3000 222.22
PW TC Labour – R5000 370.37
“True” PBP ¼ ; ð15Þ Total initial investment cost – 85,076 6301.90
PW TBav
 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45 43

nents are assumed to be used over their lifetimes. Since in this Table 3
work the inflation rate is assumed equal to the real interest rate, Payback period for different cases.

the PW of replacement cost, C rep , of each component is calculated Parameters Case I Case II Case III
as follows (adapted from Ref. [37]): Total initial investment (USD) 6301.90 6301.90 6301.90
Project lifetime, n (years) 30 30 30
C rep ¼ C cap Nrep ; ð17Þ
PV lifetime (years) 30 30 30
where C cap is the initial capital cost for each component (given in N repPV (–) 0 0 0
C repPV ðUSDÞ 0 0 0
Table 2) and Nrep is the component number of replacements over
Inverter lifetime (years) 15 15 15
the project lifetime. N repINV (–) 1 1 1
The PW TB is calculated using the following expression (adapted C repINV ðUSDÞ 2488.88 2488.88 2488.88
from Ref. [36]): Battery lifetime (years) 7 7 7
  N repBAT (–) 3 3 3
ð1 þ rÞn  1 C repBAT ðUSDÞ 3648 3648 3648
PW TB ¼ AB n ; ð18Þ AB ðUSDÞ
r ð1 þ r Þ 549.36 750.43 1208.30
PW TC ðUSDÞ 12,439 12,439 12,439
where AB denotes the annual benefit and r is the discount or inter- PW TB ðUSDÞ 7853.7 10728 17274
est rate. It is assumed that the discount/interest rate depends only PW TBav ðUSDÞ 654.47 894 1439.5
upon the inflation rate so that the discount/interest rate is equal to ‘‘True” PBP (years) 19 14 8.6
the inflation rate. Data for South African annual inflation rate from
1997 to 2016 are presented in Fig. 20, where an average inflation
rate of 5.65% is shown. The average value is used for the economic project whose lifetime is 30 years and payback period of around
analysis in this work. 8 years means that the investor will recover his total capital within
The annual cost savings achieved are calculated from Table 1 8 years and have 22 years to generate profit. Although a payback
as USD 882:20  USD 269:84 ¼ USD 612:36=year for case I, period of 14 years seems high, the investor will still have 16 years
USD 882:20  USD 68:77 ¼ USD 813:43=year for case II and to generate profit over the project lifetime. With case I, however, a
USD 882:20  ðUSD 389:1Þ ¼ USD 1; 271:3=year for case III. grid-interactive solar PV system project with a payback period of
Because of the reduced maintenance level required by AGM deep 19 years might be risky. Some of the reasons are as follows: (1)
cycle batteries, the annual O&M cost of the whole system is assumed the sustainability of the feed-in tariff programme is not guaran-
to be 1% of the total cost, that is, USD 63/year. With this, the annual
benefit (AB) is calculated as USD 612:36  USD 63 ¼
USD 549:36=year for case I, USD 813:43  USD 63 ¼ USD
Table 4
750:43=year for case II and USD 1271:3  USD 63 ¼ USD 1208:3= Payback period without battery storage system.
year for case III.
Parameters Case I Case II Case III
The cost effectiveness of the different cases is analyzed and pre-
sented in Table 3. The project lifetime is taken to be equal to the Total initial investment (USD) 5085.90 5085.90 5085.90
Project lifetime, n (years) 30 30 30
lifetime of the PV panel, assumed to be 30 years. Since the war-
PV lifetime (years) 30 30 30
ranty of the IMEON 3.6 inverter is up to 10 years, its lifetime is N repPV (–) 0 0 0
assumed to be 15 years. The lifetime of AGM deep-cycle battery C repPV ðUSDÞ 0 0 0
bank is assumed to be 7 years under a proper maintenance plan. Inverter lifetime (years) 15 15 15
Results presented in Table 3 show that with case I, when the N repINV (–) 1 1 1
C repINV ðUSDÞ 2488.88 2488.88 2488.88
feed-in tariff is USD 0:046/kW h, the total investment cost will be
AB ðUSDÞ 232.70 766.48 1057.67
recovered over 19 years. With Case II, when the feed-in tariff is
PW TC ðUSDÞ 7574.80 7574.80 7574.80
increased to be equal to the flat tariff, which is USD 0:11/kW h,
PW TB ðUSDÞ 3326.70 10,958 15,121
the payback period is reduced from 19 to 14 years. This is further PW TBav ðUSDÞ 277.22 913.16 1260.08
reduced to 8.6 years with case III when the feed-in tariff is
‘‘True” PBP (years) 27.3 8.3 6.0
increased to be equal to the peak TOU tariff (USD 0:16/kW h). A

12

10

8
Inflation rate, r [%]

2 Annual variation
0 Average (5.65%)

-2
1998 2000 2002 2004 2006 2008 2010 2012 2014 2016
Period, n [year]
Fig. 20. South African annual inflation rate from 1997 to 2016 (Adapted from Ref. [38]).
44 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45
22
20
Payback period [Years]
18
16
14
12
10
0 5
Fig. 21. Sensitivity of electricity price on the payback period.
teed, (2) the integrity of the roof has to be guaranteed to last more
than 19 years from the time the system is installed.
7. Impact of battery storage system and electricity price on the
payback period
It has been previously shown that under the current FIT, the
installation of a residential grid-interactive solar PV system in
eThekwini municipality will result in a high payback period of
19 years. Some of the reasons are the high initial investment cost
and battery replacement cost against the relatively low grid elec-
tricity price due to coal fired power plants. To analyze the effect
of battery storage system on the payback period, simulation results
without taking into account the battery storage system are pre-
sented in Table 4. It is shown that under the current FIT (case I),
a residential grid-interactive solar PV without battery storage sys-
tem will lead to a higher payback period as opposed to cases II and
III with attractive FITs. The reason is that due to a lower FIT, most
solar energy will be stored instead of being fed to the grid. This
means that for low FITs, the self-consumption of harvested solar
energy by means of energy storage system has to be maximized
in order to improve the profitability of the system. Hence, in this
case, energy storage systems will play an important role to
improve the cost effectiveness of residential solar PV systems. This
is shown through comparison between Tables 3 and 4. The results
show that with case I where the FIT is not attractive, a residential
solar PV system without battery storage system will lead to a pay-
back period of 27.3 years, which is higher than 19 years when the
battery storage system is used.
On the other hand, with high FITs (cases II and III), it is shown in
Sections 5.2 and 5.3, that the higher the FIT, the lesser the need for
energy storage system. This implies that the use of battery storage
system when the FIT is attractive will lead to unnecessary increase
in the investment cost (initial investment and replacement costs)
while not contributing to energy cost saving, thus leading to higher
payback periods. In this case, a grid-interactive solar PV system
without battery storage, referred to as grid-connected solar PV, will
be cost effective. Comparison between Tables 3 and 4 show that a
grid-interactive PV system without battery storage leads to lower
payback periods of 8.3 and 6 years in case II and III, respectively,
compared to 14 and 8.6 years when the battery storage is used.
The sensitivity of the grid electricity price on the payback per-
iod is also analyzed under case I (with battery storage). The results
are depicted in Fig. 21. The payback period is calculated for each
10% increase in the electricity price, which corresponds to the aver-
age annual increase in electricity price in South Africa. 0% increase
10 15 20 25 30 35 40
Electricity price increase [%]
corresponds to the price in current year (2016), 10% increase corre-
sponds to the price in the year 2017, 20% increase corresponds to
the price in the year 2018, 30% corresponds to the price in the year
2019 and 40% corresponds to the price in the year 2020. The anal-
ysis of the results shows that the increase in the grid electricity
price has a huge impact on the payback period of the grid-
interactive solar PV system. It is shown in Fig. 21 that for every
10% increment in the grid electricity price, the payback period lin-
early decreases by approximately 11.5% on average.
8. Conclusions and recommendations
An optimal energy model of a 3 kW residential grid-interactive
solar PV system under the feed-in tariff (FIT) is proposed in this
work. An urban residential household within the eThekwini
municipality is considered as a case study. Since the system has
an energy storage device, the household is assumed to consume
power from the grid under the time-of-use electricity tariff. To
study the economic viability of the system, the payback period
(PBP) is used as the economic performance indicator, under various
cases where the FIT is varied.
Simulation results show that a potential of 69.41% energy cost
saving is possible when the energy cost management is applied
to the residential grid-interactive PV system, under the current
FIT, that is USD 0:046/kW h. However, this leads to a large PBP of
19 years. When the FIT is increased to USD 0:11/kW h, the energy
cost saving is increased to 92.2% and the PBP is decreased to
14 years. With a more attractive FIT of USD 0:16/kW h, the energy
cost saving is further increased to 144.4% and the PBP is decreased
to 8.6 years. Simulation results show that the higher the FIT the
lesser the need for the battery bank if grid failures are not consid-
ered as a constraint.
The impact of battery storage and grid electricity price on the
cost effectiveness of the grid-interactive solar PV system is also
reported. Simulation results show that the use of battery storage
system is only beneficial to the system when the FIT is not attrac-
tive since the self-consumption of harvested solar energy by means
of energy storage has to be maximized in order to improve the
profitability of the system. For attractive FITs, the use of battery
storage will have a negative effect on the profitability of the sys-
tem. Simulation results also show that the higher the grid electric-
ity price, the higher the profitability of the system.
To this end, the following recommendations are proposed in
order to make the application of grid-interactive solar PV system
on residential buildings in South Africa economically viable:
 B.P. Numbi, S.J. Malinga / Applied Energy 186 (2017) 28–45
– The current FIT incentive should be increased at least to be
equal to the price at which residential customers purchase elec-
tricity from the grid at a flat tariff.
– The FIT incentive programme must be sustainable. This means
that it should be operational for at least 15 years from the time
the residential customer has decided to install the system.
– The government should initiate a subsidy programme on the
initial investment of the system to help the project to be cost
effective. One of the easiest solutions is to provide a free instal-
lation of the grid-interactive solar PV system.
– Reliability of the solar PV system components should be
ensured in terms of their lifetime. The higher the lifetime of bat-
teries and bi-directional inverters, the higher the cost-
effectiveness level. This also applies to the roof and building
structure supporting the panels.
Acknowledgments
The authors would like to thank the Mangosuthu University of
Technology for financial support. We would also like to thank the
anonymous reviewers for their valuable comments.
References
[1] Giglmayr S, Brent AC, Gauché P, Fechner H. Utility-scale PV power and energy
supply outlook for South Africa in 2015. Renew Energy 2015;83:779–85.
[2] Kenny A. The rise and fall of Eskom. South African Inst Race Relations 2015;2
(18):1–22.
[3] Eskom. Tariffs and charges booklet 2014/2015; 2014. <http://www.eskom.co.
za/CustomerCare/TariffsAndCharges/Documents/TariffBrochureV8.pdf> [Last
accessed: 23 February 2016].
[4] EThekwini Municipality. Electricity tariffs 2015–2016; 2015. <http://
www.durban.gov.za/Resource_Centre/Services_Tariffs/Electricity%20Tariffs/
Electricity%20Tariffs%202015%20-%202016.pdf> [Last accessed: 26 February
2016].
[5] Department of National Treasury, Republic of South Africa. Carbon tax policy
paper-Reducing greenhouse gas emissions anf facilitating the transition to a
green economy; 2013. <http://www.treasury.gov.za/public%20comments/
Carbon%20Tax%20Policy%20Paper%202013.pdf> [Last accessed: 26 February
2016].
[6] Niekerk BV. Solar PV and embedded generation in South Africa; 2013. <http://
www.sustainable.org.za/uploads/files/file13.pdf> [Last accessed: 22 January
2016].
[7] South African Department of Energy. Clean energy. <http://www.energy.gov.
za/files/media/Pub/CleanEnergy_A5booklet.pdf> [Last accessed: 25 January
2016].
[8] Solangi KH, Islam MR, Saidur R, Rahim NA, Fayaz H. A review on global energy
policy. Renew Sustain Energy Rev 2011;15:2149–63.
[9] Ozdemir S, Altin N, Sefa I. Single stage three level grid interactive MPPT
inverter for PV systems. Energy Convers Manage 2014;80:561–72.
[10] Obi M, Bass R. Trends and challenges of grid-connected photovoltaic systems.
Renew Sustain Energy Rev 2016;58:1082–94.
[11] Tazvinga H, Xia X, Zhang J. Minimum cost solution of photovoltaic-diesel-
battery hybrid power systems for remote. Sol Energy 2013;96:292–9.
[12] Tazvinga H, Zhu B, Xia X. Energy dispatch strategy for a photovoltaic-wind-
diesel-battery hybrid power system. Sol Energy 2014;108:412–20.
[13] Tazvinga H, Zhu B, Xia X. Optimal power flow management for distributed
energy resources with batteries. Energy Convers Manage 2015;102:104–10.
45
[14] Kanchev H, Lu D, Colas F, Lazarov V. Energy management and operational
planning of a microgrid with a PV-based active generation for smart grid
applications. IEEE Trans Indust Electron 2011;58:4583–92.
[15] Teleke S, Baran ME, Bhattacharya S, Huang AQ. Rule-based control of battery
energy storage for dispatching intermittent renewable sources. IEEE Trans
Sustain Energy 2010;1:117–24.
[16] Levron Y, Guerrero JM, Beck Y. Optimal power flow in microgrids with energy
storage. IEEE Trans Power Syst 2013;28:3226–34.
[17] Ren H, Wu Q, Gao W, Zhou W. Optimal operation of a grid-connected hybrid
PV/fuel cell/battery energy system for residential applications. Energy
2016;113:702–12.
[18] Sichilalu SM, Xia X. Optimal energy control of grid tied PV-diesel-battery
hybrid system powering heat pump water heater. Sol Energy
2015;115:243–54.
[19] Sichilalu SM, Xia X. Optimal power dispatch of a grid tied-battery-photovoltaic
system. Energy Convers Manage 2015;102:81–91.
[20] Wu Z, Tazvinga H, Xia X. Demand side management of photovoltaic-battery
hybrid system. Appl Energy 2015;148:294–304.
[21] Riffonneau Y, Bacha S, Barruel F, Ploix S. Optimal power flow management for
grid connected PV systems with batteries. IEEE Trans Sustain Energy 2011;2
(3):308–20.
[22] Ceraolo M, Di Donato A, Franceshi G. A general approach to energy
optimization of hybrid electric vehicles. IEEE Trans Vehicular Technol
2008;57(3):1433–41.
[23] Middelberg A, Zhang J, Xia X. An optimal control model for load shifting – with
application in the energy management of a colliery. Appl Energy
2009;86:1266–73.
[24] Zhang S, Xia X. Optimal control of operation efficiency of belt conveyor
systems. Appl Energy 2010;87(6):1929–37.
[25] Badenhorst W, Zhang J, Xia X. Optimal hoist scheduling of a deep level mine
twin rock winder system for demand side management. Electric Power Syst
Res 2011;81(5):1088–95.
[26] Numbi BP, Zhang J, Xia X. Optimal energy management for a jaw crushing
process in deep mines. Energy 2014;68:337–48.
[27] Numbi BP, Xia X. Systems optimization model for energy management of a
parallel HPGR crushing process. Appl Energy 2015;149:133–47.
[28] Numbi BP, Xia X. Optimal energy control of a crushing process based on
vertical shaft impactor. Appl Energy 2016;162:1653–61.
[29] Pontryagin L, Boltyanskii V, Gamkrelidze R, Mishchenko E. The mathematical
theory of optimal processes. New York, United States: John Wiley & Sons Inc.;
1962.
[30] Imeon Energy. Self consumption smart grid inverter; 2015. <http://
www.domsolar.com/pdf/Tech-Details-EN-V4C-IMEON-ENERGY-3.6-9.12-Self-
Use-Smart-Grid-Inverter.pdf> [Last accessed: 06 April 2016].
[31] Yingli solar. YGE 60 Cell 40 mm series. <http://www.yinglisolar.com/assets/
uploads/products/downloads/YGE_60_Cell_Series_EN.pdf> [Last accessed: 06
April 2016].
[32] Mahmoud MM, Ibrik IH. Techno-economic feasibility of energy supply of
remote villages in Palestine by PV-systems, diesel generators and electric grid.
Renew Sustain Energy Rev 2016;10:128–38.
[33] CB Solar. DC12-100 (12 V100A h CB Solar); 2015. <http://wynlandesolar.co.za/
wp-content/uploads/2016/04/DC12-100.pdf> [Last accessed: 11 April 2016].
[34] Duffie JA, Beckman WA. Solar engineering of thermal process. 4th ed. New
Jersey and Canada: John Wiley & Sons, Inc.; 2013.
[35] SAURAN. Station details: KZH-University of KwaZulu-Natal, Howard College.
<http://www.sauran.net/ShowStation.aspx?station=2> [Last accessed: 11
April 2016].
[36] Capehart BL, Kennedy WJ, Turner WC. Guide to energy management. 5th
ed. Indian Trail, United States: The Fairmont Press, Inc.;; 2008.
[37] Ogunjuyigbe ASO, Ayodele TR, Akinola OA. Optimal allocation and sizing of PV/
wind/split-diesel/battery hybrid energy system for minimizing life cycle cost,
carbon emission and dump energy of remote residential building. Appl Energy
2016;171:153–71.
[38] World inflation data. Historic inflation South Africa – CPI inflation. <http://
www.inflation.eu/inflation-rates/south-africa/historic-inflation/cpi-inflation-
south-africa.aspx> [Last accessed: 13 April 2016].