Journal of Energy Storage 32 (2020) 101717

Contents lists available at ScienceDirect

Journal of Energy Storage

journal homepage: www.elsevier.com/locate/est

Optimal peer-to-peer energy management between grid-connected T

prosumers with battery storage and photovoltaic systems
K. Kusakana
Department of Electrical, Electronic and Computer Engineering, Central University of Technology, Free State Bloemfontein, South Africa

A R T I C LE I N FO A B S T R A C T

Keywords: This paper proposes a model for the optimal Peer-to-Peer energy sharing of grid-connected Prosumers. Current
Energy sharing models within available literature do not include the constraint which forbid small renewable energy producers
Microgrid, Peer-to-Peer, Prosumer, Optimal to share their excess power through the national grids; this leads to prosumers dissipating their excess power
energy management, Time of Use generated or using costly energy storage systems. Additionally, most of these models look at the savings made by
the Peer-to-Peer community, without computing the cost reduction or benefit of each prosumer taken in-
dividually. Thus, these models are not suitable or directly applicable to many countries such as South Africa.
For these reasons, the main aim of this paper is to develop a Peer-to-Peer energy sharing model that considers
the above mentioned gaps. The proposed system consists of two prosumers; a residential prosumer that employs
a roof mounted photovoltaic system with energy storage capabilities, and commercial prosumer with a dual-
tracking photovoltaic system. The prosumers are connected to each other by power lines for P2P operation. The
developed model minimizes both prosumers’ operation costs by maximizing the use of the power from the
renewable energy sources; optimally managing the internal power sharing between the prosumers; and mini-
mizing the use of the electrical utility operating with the Time-of-Use rate. The study uses the dissimilarity and
complementarity of the load patterns in the South African residential and commercial energy sectors as an asset
to implement the Peer-to-Peer energy sharing between the prosumers operating in these two sectors.
Using a case study in South Africa, the results have demonstrated that, as compare to using the grid as sole
power source, the residential prosumer can achieve a total daily cost reduction of 62.71% in summer and
68.99% in winter; while the commercial can achieve 81.31% and 31.69% in winter. For a 20 years’ project, using
the grid as baseline for the comparison, the lifecycle cost analysis have projected the residential prosumer to
break-even after 16.35 years and save 19.5%; while the commercial prosumer can break-even after 5.3 years and
save 55.2%. Moreover, this study shows the ability of the developed model to simulate systems with different
components, demands, resources as well as costs. The model developed in this study may assist researchers and
engineers interested in the optimal operation of prosumers in Peer-to-Peer energy sharing schemes.

1. Introduction small scale RESs producers to export energy into the national grid,
which force them either to store or dissipate their excess energy on the
All over the world, renewable energy systems (RESs) are being site of production [4]. Therefore, innovative ways of dealing with this
promoted and implemented as alternatives to conventional energy excess energy are needed; these can reduce the wastage or the cost of
sources; supporting the Sustainable Development Goals (SDG) [1]. energy storage devices involved while increasing the penetration of
However, most small scale RESs users are suffering from the variable renewable energies.
nature of the renewable resources which in turn reduce the systems’ Taking advantage of the fact that it is very unlikely that two load
sustainability and profitability; unless they include some type of sto- demands can display the same profile at the same time; several small
rage, or if they are connected to the national grid [2]. RESs producers can assist each other in balancing their excess energy
Looking at the increasing number of households or industries that generated with other consumers that are in deficits or those who would
are becoming “prosumers” by producing their own power from small like to reduce their reliance from the utility [4]. This novel concept is
RESs; a challenge occurs when there is excess power produced from defined as Peer-to-Peer (P2P) energy sharing. Similar trading arrange-
these systems [3]. In South Africa for example, the grid code prevents ments are applied in concepts such as Bitcoin, Airbnb or Uber.

E-mail address: [email protected].

https://doi.org/10.1016/j.est.2020.101717
Received 23 June 2020; Received in revised form 19 July 2020; Accepted 22 July 2020
2352-152X/ © 2020 Elsevier Ltd. All rights reserved.
K. Kusakana Journal of Energy Storage 32 (2020) 101717

Nomenclature O&M Operation and maintenance

P2P Peer-to-Peer
BEP Break-even point PV Photovoltaic
Cexc % of ToU used as incentive to prosumers Pi Power flows (control variables)
CP Commercial prosumer PLi Prosumer's load demand
Cn Capacity of the selected battery RES Renewable energy sources
Cinitial Initial cost of each component; RP Residential prosumer
Csalvage Salvage cost linked for each component; SoC State of charge
Crep Replacement cost of each component; STC Standard test condition
COM Operation and maintenance cost of each component; Ta Ambient temperature
CEC Net energy cost. TC Cell temperature
DC Direct current ToU Time of Use
fi Objective function ZAR South African Rand
Id (DNI) Direct normal irradiance ρj Cost of energy linked to the applicable residential ToU
Idif (DHI) Diffuse horizontal irradiance Ψj Cost of energy linked to the applicable commercial ToU
Ig (GHI) Global horizontal irradiance Δt Sampling time
It Solar irradiance ηCh Battery charging efficiency
j jth optimization sample interval ηDisch Battery discharging efficiency
LLC Life cycle cost α Power temperature coefficient
N Number of sampling intervals β Tilted angle for the stationary system
NOCT Nominal operating cell temperature θβ Incidence angle of solar radiation on a tilted surface
NPVs and NPVp Series and parallel numbers of PV modules connected ρ Reflectance of the environment around the system.

However, in many countries, this P2P arrangement can only be residential microgrid using PV with storage systems was considered.
realized if the power generation and consumption is happening on the Results showed that the operating cost was reduced by 30% in com-
same municipally registered land [3]. Furthermore, for those producers parison to the case where the prosumers are only allowed to feed power
who are permitted to feed power into the grid, the monetary incentive back to the grid. Using a similar methodology, comparable results were
is insignificant; therefore, it is more beneficial for them to look at other achieved in Ref. [21], where the minimum considered portion of pro-
alternative, such as the P2P, to deal with their excess power generated. sumers with PV should be at least 40% of all the participants. In Ref.
In some of the developed countries, different P2P energy sharing [22] an optimal P2P energy sharing model of domestic prosumers, in a
platforms and projects have been implemented. The key ones that can grid-connected DC microgrid, is developed to maximize the usage of the
be cited are such Piclo [5]; Vandebron [6]; PeerEnergyCloud [7]; Smart power generated locally and minimize the households’ operation cost.
Watts [8]; Yeloha and Mosaic [9]; SonnenCommunity [10]; Lichtblick In Ref [23], the P2P energy management has been modelled and stu-
Swarm Energy [11] and TransActive Grid [12]. However; all these died for an off-grid system including three prosumers in which two are
projects have been implemented at large scale and involve the use of equipped with PV systems and the last one has a hydrogen storage
the main utility distribution network for the different P2P energy system.
transactions. When looking closely at the studies cited above, they do not re-
Several other major research works, related to the P2P energy present the actual situation in South Africa where small prosumers are
sharing between prosumers, have been reported in the literature. In Ref. limited by the applicable Electricity Regulation Act of 2006, restricting
[13], the P2P energy trading was defined as a transaction characterized the consumption of power generated on site and low allowing ex-
by direct energy exchange among peer prosumers. Through a case portation of energy through the grid [24]. Furthermore, the savings in
study, the authors have demonstrated that P2P energy trading can these studies are computed as the total P2P energy sharing for the
allow for an improvement of the power flows between the generations community, not highlighting the unique benefit for each prosumer
and the load demands. In Ref. [14], it was revealed that there is a taken individually.
growing trend for residential homes to become prosumers with PV Very few studies have been conducted on the P2P energy sharing
systems; and using electric vehicles as means of storing energy. The concept applicable to South Africa. In Ref. [24], the state of the re-
research further looked at minimizing the cost of energy in these smart search and development on P2P electricity sharing are reviewed. Im-
homes connected in a P2P energy sharing arrangement. The authors of portant data on South Africa, that may assist researchers interested in
Ref. [15] discussed a wide-ranging survey of existing optimization the P2P research field are provided. In Ref. [25] a research was done
models, power dispatch technologies, challenges, enabling technologies with the aim to develop a model to manage the operation of a charging
as well as applicable control algorithms for the implementation of P2P station that can allow P2P energy sharing between electric vehicles
energy trading. In Ref. [16], the motivation, issues, design, guidelines while minimizing the grid cost. The station is supplied by a hybrid
and potential implementation of new P2P platforms were considered. In system composed of a PV, wind as well as grid connection. In Ref [4], a
Ref. [17], a new concept of P2P energy sharing has been discussed; model is developed to optimize the P2P power exchange taking place
enabling a market for prosumers selling electricity to consumers who between the two prosumers using grid-connected PV systems and
are unable to produce their own power, or who might have access to having different load patterns. In Ref. [26], the system is composed of
electricity but need more at certain times of the day. Ref [18] presents one consumer acquiring energy from a grid-connected prosumer with a
an innovative P2P energy arrangement between two groups of electric PV and a battery storage system. In Ref. [27], the P2P trading is taking
vehicles, which decreases reliance of the charging process on the grid, place in a standalone system, isolated from the grid, where the involved
allowing a potential cost saving of up to 71%. In Ref. [19], a model is prosumers generate electricity using a hydrokinetic, a diesel generator
developed for the optimization of the energy sharing among microgrids and a pumped hydro storage scheme.
formed of PV prosumers, which revealed to be more beneficial than When looking at Ref. [4,24–27], their results showed that sub-
operating prosumers in isolation when the feed-in tariff is considered. stantial energy costs can be saved. However, it was noticed that a cost
In Ref [20], a two-stage aggregated control for P2P energy sharing in allocation mechanism for the internal power sharing between

2
K. Kusakana Journal of Energy Storage 32 (2020) 101717

prosumers is not included. This means that the P2P energy transactions instance, a block of apartments, townhouse complex or estates [4]. P2P
were considered to be free of charge; not giving a true representation of energy sharing models specify at which prices prosumers will trade
the cost saved by prosumers. Additionally, the cost savings were com- energy with each other and how to calculate the energy bill for the
puted for the community, not giving more details of the share allocated prosumers.
to each prosumer.
From the papers cited above, it can be clearly seen that there is a
lack of studies related to P2P applied to South Africa. Therefore, this 2.2. Proposed system configuration
paper presents the development of a model for the optimal power dis-
patch of prosumers operating in a P2P energy sharing arrangement and In this study, two prosumers generating electricity using PV systems
connected to each other by power lines for P2P operation. The devel- are considered. They display dissimilar load demands while being on
oped model minimizes the involved prosumers’ operation costs by op- the same registered municipal land. For this case, it is considered that;
timally managing the power flow between the RESs, the storage system, the prosumers are connected to each other by power lines and only one
the variable load demands as well as the power supplied from the of them possesses a battery storage system. This is a scheme derived
electrical utility, operating with the Time-of-Use (ToU) rate. Given the from the P2P scheme given in Fig. 2, with the constraint that feeding
fact that the interaction between the different components operating excess energy from the prosumers to the grid is not permitted.
under the different respective constraints are non-linear; the simula- Fig. 3 shows the different power flows applicable in the considered
tions are performed using “fmincon” from MATLAB optimization P2P energy sharing arrangement involving prosumers with RESs and
toolbox. The simulations are performed for a case study, in South grid connection. In this figure, the residential prosumer (RP) has a
Africa, from which the renewable energy sources, load demand as well stationary PV with a battery storage system for self-power production
as grid tariffs are used as input data. and storage in case of excess/deficit. The reason why the RP has a
The main contributions of this work can be summarized as follows: battery bank, is due to the fact that the PV production does not match
the residential demand profile and there is a wide margin to store this
• A new optimal P2P solution is proposed to minimize the energy energy during the low demand periods.
The first strategy in meeting the RP's load demand is to maximize
costs of grid-connected prosumers using RESs and storage systems. A
mechanism for the cost allocation to the internal energy transactions the use of the PV power, then using the battery, and finally the grid
between the prosumers is considered. which needs to be minimized. But to achieve the lowest cost of energy
• Unlike in many studies, the different operation costs as well as the used from the grid during the considered control period, the operation
sequence can change given the variables involved such as the solar
lifecycle costs are computed for each prosumer, not as a community.
This gives a better understanding of what is the P2P benefit for each resources, the pricing period, the battery state of charge (SoC) as well as
prosumer. the load demand.
• The study uses the dissimilarity and complementarity of the load Considering that the PV generation usually matches the commercial
demand profile; a dual-tracking system is used to capitalize on the PV
patterns in the South African residential and commercial sectors (as
well as the different ToU tariffs applicable in these sectors) as an resource. A similar operation scenario, as for the RP, is observed on the
asset to implement the P2P energy sharing between prosumers op- commercial prosumer (CP) where the PV, has priority over the grid in
erating in these sectors. These load profiles also favor the use of supplying the load.
battery systems for the residential prosumer and dual axis tracking However, when the P2P energy exchange is considered between the
system for the commercial prosumer.

These aspects have not been yet considered together in South Africa,
and other similar cases, where this concept cannot be blandly applied
like in developed countries.
The remaining part of this paper is organized into seven sections.
Section 2 explains the P2P energy sharing concept with the current
advances on this research field reported in the literature. Section 3
describes the selected architecture and the components of the P2P
scheme to be modelled in Section 4. The simulation results linked to the
optimal power flows of the system, with the P2P, are explained in
Section 5. The economic analysis with the life cycle cost and break-even
point is presented in Section 6; followed by the study conclusions in
Section 7.

2. Description of the proposed P2P system

2.1. Basics of P2P energy sharing and prosumers concepts

P2P energy sharing is quite a new model for trading energy invol-
ving “prosumers” who are consumers having the capabilities of pro-
ducing their own power [19]. In the traditional trading model, the local
energy provider buys electricity from the utilities in bulk, then, sells it
to the consumers, as shown on Fig 1. In this arrangement, the prosumer
can sell its excess energy only to the local energy provider and this
happens at a marginal cost.
With the P2P energy trading arrangement, prosumers can exchange
energy among themselves, as indicated in Fig. 2. In South African, as
well as in some other countries, this can only be applicable in the case
where the involved prosumers are located on the same land, for Fig. 1. Conventional Model.

3
K. Kusakana Journal of Energy Storage 32 (2020) 101717

• P : Power from the RP's PV for battery charging purpose.

3

• P : Power from the grid to source the RP's load.

4

• P : Power from the RP's PV to source the CP's load.

5

• P : Power from the RP's battery to source the CP's load.

6

• P : Power from the CP's PV for the CP's self-consumption.

7

• P : Power from the CP's PV to source the RP's load.

8

• P : Power from the CP's PV for RP's battery charging purpose.

9

• P : Power from the grid to source the CP's load.

10

• P : Power from the grid for RP's battery charging purpose.

11

3. Mathematical model development

3.1. Objective function

The developed objective function “f” aims to minimize the net en-
ergy cost for each of the prosumers involved. To make it easy to un-
derstand, this function can be divided in two where the first part “f1”
deals with the energy cost linked to the power procured from the grid
(to supply the loads and recharge the battery). This is modelled as:
N
f1 = min ∑ [ρj (P4(j) + P11(j) ) + Ψj (P10) ] Δt
j=1 (1)

Where: f1 is the function to be minimized, linked to the sum of power

procured from the grid;
ρj: is the cost of energy linked to the applicable residential ToU;
Ψj: is the cost of energy linked to the applicable commercial ToU;
Fig. 2. P2P energy sharing model.
j is the selected jth sample interval where the optimization is taking
place;
N is the overall number of sample intervals;
Δt is the duration of each sampling interval.
Defined in the second part of the objective function “f2”, each
prosumer should maximize on the usage of its own energy produced
before sharing the energy with the other prosumer. A simple pricing
mechanism for the internal energy sharing between the prosumers is
considered. The prosumers will be paid a fixed percentage “Cexc” of the
variable ToU, similar to the incentive currently applicable in the case
where the prosumers are allowed to feed power into the grid [28]. This
means that when the RP gets power from the CP, a ratio of the com-
mercial ToU is applied to the transaction; and vice versa. For the dif-
ferent demand sectors, this pricing mechanism based on the ToU, insure
that the Cexc varies with the different pricing periods as well as seasons.
This can be modelled as:
N
f2 = min ∑ [Cexc (ρj (P5(j) + P6(j) ) + Ψj (P8(j) + P9(j) ))] Δt
j=1
N
+ max ∑ [(P1(j) + P7(j) )] Δt
j=1 (2)

3.2. Load demand and supply balances

For each considered sampling interval, power balances between the

supply and the demand, linked to both the RP and CP are respectively
Fig. 3. P2P energy sharing between prosumers.
modelled as:

prosumers’ generation, demands and storage; additional power flows PL1(j) = P1(j) + P2(j) + P4(j) + P8(j) (3)
are included, making the operation of the entire system more complex.
PL2(j) = P5(j) + P6(j) + P7(j) + P10(j) (4)
Each prosumer has to look at the load balances of the community, as a
microgrid, while minimizing the total cost of energy procured from the Where PLi(j) is the prosumer's load demand.
grid during the considered operation time. These equations basically indicate from which sources each load can
The various power flows, or control variables, from Fig. 3 may be get power before the power dispatch is optimized.
defined as follows:
3.3. Power sources operation constraints
• P : Power from the RP's PV for self-consumption.
1

• P : Power from the battery for the RP's self-consumption.

2 For any considered optimization sample (j), the sum of powers

4
K. Kusakana Journal of Energy Storage 32 (2020) 101717

supplied by a source to the demand, battery or shared with another 3.5. Power flow boundaries
prosumer; should be less or equal to the power generated by the con-
sidered power source. For the RP and CP's PV, the maximum power Each generator must be operated between a maximum and a
depends on the design characteristics as well as on the available solar minimum value. The maximum value depends on the design specifi-
resources; these are given by Eqs. (5) and (6) respectively. cation as well as from the resources in case of the RESs. For all the
max
involved power flows as Pi, this can generally be modelled as:
P1(j) + P3(j) + P5(j) ≤ PPV 1(j ) (5)
0 ≤ Pi (j) ≤ Pimax
(j ) (10)
max
P7(j) + P8(j) + P9(j) ≤ PPV 2(j ) (6) For the specific case of the SoC, this can vary between a maximum and a
minimum value which depend also on the type of battery used. This can
For the battery and the grid, the upper limit of the power is not time
be modelled as:
dependent and is only linked to a rating or imposed constant maximum
value. These constraints are given in Eqs. (7) and (8) respectively. SoC min ≤ SoC(j) ≤ SoC max (11)
Rated
P2(j) + P6(j) ≤ PBat (7)
3.6. Exclusive power flows
max
P4(j) + P10(j) + P11(j) ≤ PGrid (8)
In certain operation strategies, selected simultaneous power flows
are not allowed. In the case of the developed P2P energy sharing model,
3.4. Battery state of charge the simultaneous power exchange between the two prosumers is not
permissible. Additionally, the charging and discharging processed
The variation of the SoC depends on the power flows entering and cannot happen simultaneously. These two conditions can be respec-
exiting the battery. Looking at the charging and discharging processes tively modelled as:
taking place in the proposed arrangement; the SoC can be modelled as: (P5(j) + P6(j) ) × (P8(j) + P9(j) ) = 0 (12)
Δt
SoC(j) = SoC(0) + (P2(j) + P6(j) ) × (P3(j) + P9(j) + P11(j) ) = 0
Cn (13)
N N
⎡ ∑ j=1 P2(j) + P6(j) ⎤
× ⎢ηCh × ∑ (P3(j) + P9(j) + P11(j) ) − ⎥
⎢ j=1 ηDisch ⎥ 3.7. Fixed-final state condition
⎣ ⎦ (9)

Where: SoC(0)is the considered SoC at the starting of each optimization For the control algorithm to be implemented to the developed
interval; model consecutively, the input and output powers to the battery should
Cn is the capacity of the selected battery in Wh; be balanced for the considered simulation horizon. This condition will
ηCh and ηDisch are the efficiencies linked to the charging and dis- insure that the SoC is the same at the beginning of each simulation
charging processes. horizon (after N sampling intervals). This can be modelled as:

Fig. 4. Residential and commercial prosumers’ load profiles.

5
K. Kusakana Journal of Energy Storage 32 (2020) 101717

N
temperature coefficient, TC is the cell temperature, Ta is the tempera-
∑ (P2(j) + P6(j) − P3(j) − P9(j) − P11(j) ) = 0 ture of the air around the cell, and NOCT is the nominal operating cell
j=1 (14)
temperature.
The solar irradiance can be modelled as follows [31]:
3.8. Solver selection
1 + cosβ ⎞ 1 − cosβ ⎞
It = Idcosθβ + Idif ⎛ + ρ Ig ⎛
⎝ 2 ⎠ ⎝ 2 ⎠ (17)
Eqs. (1) to (14) describing the developed model, make the optimi-
zation problem of the nonlinear programming type. This can be solved Where, Idif is the diffuse horizontal irradiance (DHI), Id is the direct
using “fmincon” solver from MATLAB (R2019b) optimization toolbox normal irradiance (DNI), Ig is the global horizontal irradiance (GHI), β
[29]; running on a computer with Intel core i7 processor with 8 GB of is the tilted angle for the stationary system, θβ is the incidence angle of
RAM. solar radiation on a tilted surface and ρ is the reflectance of the en-
vironment around the system.
4. Case study selected in South Africa The size of a PV system to be implemented on the prosumers’ sides
depends on the space available, the capital funds as well as the saving
4.1. Prosumers’ load profiles and renewable resources targets. Based on the demand and available solar resources, for the
stationary PV system on the residential side, the different parameters
For the selected case, the RP and CP's demand profiles are given in used to compute the output power are: PPV,STC = 0.31 kWp , NPVs = 6,
Fig.4. It can be clearly seen that the two profiles are dissimilar in terms NPVp = 2 , α = 0.0045 ∘C , NOCT = 46 ∘C , ρ = 0.2 , β = 30∘ For the dual-
of peak demand occurrence. This can be used as a motivation to im- tracking PV system on the commercial side, the parameters used to
plement the P2P energy sharing between the power generated by pro- compute the output power are: PPV,STC = 0.31 kWp , NPVs = 5, NPVp = 2 ,
sumers in these two sectors. α = 0.0045 ∘C , NOCT = 0.0046 ∘C . For each prosumer, a 5 kVA 48 V in-
For simulation purposes, the worst case is selected where the re- verter, with maximum power on the Maximum Power Point Tracking
sources are low and the load demands are high. This is a winter day (MPPT) is proposed. This can allow the potential for future expansion of
where the grid tariff is also considered as high season. A summer case is the system.
also selected to compare the technical and economic performances of
the prosumers in different conditions imposed by the seasons. The solar 4.2.2. Battery bank sizing
data for a day selected in the two extreme season are plotted in Figs. 5 From the best practice on battery charging, the size of the battery
and 6 respectively. The data was collected for the year 2019, available linked to a PV system should allow for a charging current between 10%
for each 15-minute interval, in the city of Bloemfontein located at la- and 20% of the battery capacity. Given the fact that the maximum
titude: −29.11074, longitude: 26.18503 and elevation: 1491 m [30]. current of the charge controller is 10A; a 9.6 kWh battery system with a
Looking at Figs. 5 and 6, it should be noted that most summer days charging voltage rated at 48 V battery is proposed. The selected battery
in Bloemfontein are overcast. This may be attributed to the fact that maximum charging and discharging currents are 25 A and 80 A.
Bloemfontein obtains its rain in the summer months, unlike most winter Therefore, the maximum charging and discharging power flow powers
days, when clear skies are apparent. However, the 2nd of January was a are calculated to be PBmax = 48V × 25A = 1.2kW and
perfect day, with no cloud cover. Moreover, it may clearly be seen that PBmin = −48V × 80A = −3.84kW .
the irradiance representing the two seasons vary in magnitude.
4.3. Cost of power from the grid and internal power exchange
4.2. Prosumers’ PV sizing
The South African ToU structure for the considered low and high
4.2.1. PV capacity seasons is given by Eq. (18) and (19) respectively. Where: ρk, ρ0, ρs are
The power generated by the PV system can be modelled as [30]: the costs of energy in the peak, off-peak and standard pricing intervals
It respectively [32]. It has to be noted that this pricing structure is ap-
PPV = PPV,STC NPVs NPVp [1 − α (TC − 25)]
1000 (15) plicable for both the residential and commercial sectors. However, the
costs of energy applicable in each pricing period are sector dependant.
With:
Also these costs and pricing period change depending on the seasons
It [33].
TC = Ta + (NOCT−20)
800 (16)
⎧ ρk ; tεTk , Tk = [7, 10) ∪ [18, 20)
Where PPV is the instantaneous power generated, PPV, STC is the power at
ρ (t ) = ρ ; tεT0; T0 = [0, 6) ∪ [22, 24)
standard test condition, NPVs and NPVp are the series and parallel ⎨ 0
⎪ ρs ; tεTs; Ts = [6, 7) ∪ [10, 18) ∪ [20, 22) (18)
numbers of modules connected. It is the solar irradiance, α is the power ⎩

Fig. 5. Summer solar irradiance.

6
K. Kusakana Journal of Energy Storage 32 (2020) 101717

Fig. 6. Winter solar irradiance.

⎧ ρk ; tεTk , Tk = [6, 9) ∪ [17, 19) CP's demand ins constantly low. Therefore, Fig. 9 shows that the RP's
ρ (t ) = ρ ; tεT0; T0 = [0, 6) ∪ [22, 24) demand is principally met by battery, through P2; as well as from the
⎨ 0 grid, through P4. Towards the end of this pricing period (between
⎪ ρs ; tεTs; Ts = [6, 7) ∪ [9, 17) ∪ [19, 22) (19)
⎩ 05h00 and 06h00), the PV systems start generating; therefore, the RP's
As an adopted pricing mechanism, the internal power exchanges PV is used to supply the load in conjunction with the grid, while the
between prosumers are linked to a fixed percentage “Cexc” of the ap- energy of battery is kept for later use.
plicable ToU which is comparable to the feed in tariff available in South Fig. 10 shows that, the CP's demand, is met by the power from the
Africa. Hence, in this study, the fixed percentage is taken 65% of the RP's battery (P6) as well as a contribution from the grid (P10). However,
utility retail price [28]. More details on the energy costs and other si- between 05h00 and 06h00, the CP's PV starts generating power.
mulation parameters are given in Table 1. Therefore, the CP's demand is met by its PV generation (P7) as well as
from the grid (P10).
5. Simulation results and discussion Fig. 11 shows how the output power from the battery to the RP (P2)
and to the CP (P6) decrease the SoC.
Simulations are performed to assess the effectiveness of the devel-
oped P2P model to minimize the operation cost of each prosumer while
5.2.2. P2P operation during the standard pricing interval (yellow):
optimizing the energy exchanges between them. The simulations re-
06h00–07h00
ported are for a day in summer; the same methodology is applicable for
The first standard pricing period starts from 06h00 to 07h00. It can
any other season using the applicable input data.
be observed from Fig. 4 that the RP's demand decreases while there is a
The exclusive supply of the residential and commercial demands by
small increase in the CP's demand. Figs. 7 and 8 show that both the PV
the grid is used as baseline for comparison with the developed P2P
systems generation is increasing. Therefore, Fig. 9 shows that the RP's
scheme. The impacts of stationary and tracking PV systems on the P2P
load is solely met by its own PV generated power (P1); there is no power
energy sharing scheme are also highlighted.
from the battery (P2), no power from the grid (P4) and no power
sourced from the CP's PV (P8).
5.1. Baseline: loads exclusively supplied from the grid Fig. 10 shows that, during this standard pricing period, the CP's
demand is successfully met by its own double tracking PV system (P7).
Figs.4 can also be used to show how the supply profiles look when There is no power sourced from the RP's PV (P5), battery (P6) or from
the grid is used as sole energy supply for the two demands (considered
not having any RESs on site). This baseline system is a true reflection of Table 1
the effect that the grid will have on both demands as well as on the Simulation parameters.
power required throughout the different pricing periods, for both loads.
Item Figure

5.2. Optimal P2P sharing between the grid-connected prosumers Sampling time (Δt) 15 min
Commercial PV rated power 3.5 kWp
Residential PV rated power 5 kWp
The stationary PV system on the RP's side is mounted at a tilt angle
Battery nominal capacity 9.6 kWh
of 30°. This was selected based on the case study's location, for an ρk residential (summer) 183.60c/kWh
optimum energy yield. It may be clearly observed in Fig.7; that the PV ρ0 residential (summer) 129.50c/kWh
system starts generating power at the earliest hours of the day which ρs residential (summer) 142.50c/kWh
steadily increases until the system reaches its peak, between ρk commercial (summer) 239.85c/kWh
ρ0 commercial (summer) 124.21c/kWh
11h00–13h00. ρs commercial (summer) 133.84c/kWh
The output power from the dual-tracking PV system on the CP's side ρk residential (winter) 186.43c/kWh
is shown on Fig.8. It can be observed that, due to the tracking cap- ρ0 residential (winter) 178.75c/kWh
abilities, the system starts generating very early and the peak power ρs residential (winter) 323.51c/kWh
ρk commercial (winter) 283.6c/kWh
output is also reached before 10h00, then operates at close to its peak as
ρ0 commercial (winter) 129.50c/kWh
long as there is enough daylight. ρs commercial (winter) 142.50c/kWh
SoC0 82%
5.2.1. P2P operation during the off-peak pricing interval (green): 00h00- SoCmax 100%
SoCmin 50%
06h00
ηCh 85%
In this pricing interval, Fig 4. Shows that there is a peak power ηDisc 90%
demand from the RP towards the end of this pricing period while the

7
K. Kusakana Journal of Energy Storage 32 (2020) 101717

Fig. 7. Stationary RP's PV system output power mounted at a tilt angle of 30°.

the grid (P10). 5.2.3. P2P operation during the peak pricing interval (red): 07h00–10h00
To prepare for the coming pricing interval, Fig. 12 shows that the During this peak pricing period, Fig. 4 shows that the RP's demand is
excess power from the RP's PV not used by the load, as well as a con- low but the CP's one is high. Therefore, Fig. 9 shows that the RP's load is
tribution from the grid are used to recharge the battery through (P3) successfully met by its own PV generated power (P1); contribution is
and (P11), respectively. needed from other sources.
Fig. 10 shows that the CP's demand is mainly supplied by all the

Fig. 8. Dual-tracking CP's PV system output power.

8
K. Kusakana Journal of Energy Storage 32 (2020) 101717

Fig. 9. Optimal power flow seen from the residential prosumer's demand.

production from the dual-tracking PV (P7) which is supplemented by all operation strategy better than rerouting the excess power to the CP's
the excess power generated from the RP's PV (P5). The balance of power demand.
needed by the CP's load is supplied from the grid (P10), as a last al-
ternative due to the high cost of energy during this period.
5.2.4. P2P operation during the standard pricing interval (yellow):
Fig. 12 shows that towards the end of this peak pricing period, there
10h00–18h00
is excess power from the RP's PV not used by the load used to recharge
Throughout the second standard pricing period, Fig. 4 shows that
the battery through (P3). The optimization algorithm has found this
the RP's demand is still low while the CP's remains high. Like for the

Fig. 10. Optimal power flow seen from the commercial prosumer's demand.

9
K. Kusakana Journal of Energy Storage 32 (2020) 101717

Fig. 11. Battery output power flows and SoC's dynamic.

previous pricing period, Fig. 9 shows that the RP is supplied by its own Fig. 12 shows that the battery is being charged throughout this
PV system (P1) and there is no need for power from other sources. pricing period using the excess power from the RP's PV (P3) as well as
As for the previous pricing period, Fig. 10 shows that the CP's de- using the grid (P11). The corresponding SoC increases as shown on
mand is mainly supplied by all the production from its own PV system Fig. 11.
(P7), from a portion of the excess power generated from the RP's PV (P5)
as well as from the grid (P10).

Fig. 12. Battery input power flows.

10
K. Kusakana Journal of Energy Storage 32 (2020) 101717

Table 2 Table 3
Prosumer cost savings for a summer and winter day. Residential prosumer Bill of quantity.
Supply option Residential Prosumer daily Commercial Prosumer daily Component Cost (ZAR) Life expectancy
cost (ZAR) cost (ZAR) (years)

Summer Winter Summer Winter 3.41 kWp stationary PV system (12 Panels 19,772.5 20
Peimar Monocrystalline 310 W PV)
Grid only (Baseline) 13.68 20.64 50.46 63.11 Pylon 9.6 kWh Lithium battery 68 046.72 20
Grid used in P2P 7.74 17.49 6.79 32.02 Connector 606.97 20
Sold to peer −2.83 −11.2 −0.19 −0.11 Solar Cable 100m 1425.71 20
Purchased from 0.19 0.11 2.83 11.2 Lynx Axpert 5 kVA 48 V 11,923.51 10
peer Total 118 607
Net cost in P2P 5.1 6.4 9.43 43.11
Daily savings 8.58 14.24 41.03 20
Table 4
Commercial prosumer Bill of quantity.
5.2.5. P2P operation during the peak pricing interval (red): 18h00-20h00
Component Cost (ZAR) Life expectancy
Throughout the second peak pricing interval, Fig. 4 shows that the
(years)
RP's demand is high while the CP's one is decreasing. Also, the outputs
of both PV systems are decreasing due to the decrease in solar resource. 3 kWp, dual axis solar tracking (10 Panels 80,000 20
Fig. 9 shows that, from 18h00 to 19h00, the RP's demand is suc- Peimar Monocrystalline 310 W PV)
cessfully met by the PV (P1). There is some excess power from the PV Lynx Axpert 5 kVA 48 V 11,923.51 10
Connector 606.97 20
used to recharge the battery (P3) as shown on Fig. 9. From 19h00 to Solar Cable 100m 1425.71 20
20h00, the RP's load is successfully met by the remaining power from Total 93 956.19
the PV, with the support of the battery (P2) and a contribution from the
grid (P4) as a last option to avoid paying a high cost of energy.
Fig. 10 shows that, between 18h00 and 19h00, the CP's demand is 6. Economic analysis
principally met by its own PV power. This is due to the tracking cap-
abilities of the PV that allows for prolong power generation with low The economic analysis is performed to quantify the benefit of the
incoming solar resources. From 19h00 to 20h00, the CP's demand is met optimal P2P energy sharing in the proposed arrangement in terms of
by a combination of its PV's power (P7), the battery from the RP (P6) potential daily energy cost saving as well as life cycle cost (LCC). The
and the grid (P10). power supplied from the grid is used as baseline for the comparison
Fig. 11 shows that, between 18h00 and 19h00, there is no power with the proposed prosumers operating under the developed optimal
coming out of the battery to supply any of the demands, while the same P2P energy sharing model. The different components making up the net
time, Fig. 12 shows that the battery is being recharged by the RP's PV daily energy cost are also analysed for each prosumer. Unlike the
(P3) and the grid (P11). This is also reflected by an increase in the SoC as methodology used in many P2P papers available in the literature where
seen on Fig. 11. From 19h00 to 20h00, the battery is used to supply the total costs savings are computed for the community; the economic
both demands through P2 and P6. The corresponding reduction in the analysis done in this paper looks at the benefit of each prosumer taken
SoC is shown on Fig. 11. individually.

6.1. Daily cost comparison

5.2.6. P2P operation during the standard pricing interval (yellow): 20h00-
22h00
Table 2 shows the potential daily cost saving that each prosumer
Fig. 4 shows that the RP's demand experiences a second peak, while
partaking in the P2P energy sharing scheme can achieve for the selected
the CP's demand is at the lowest. Unfortunately, there is no solar re-
day (and condition). The Table also shows the cost linked to each en-
sources to produce energy from the two PV systems. Therefore, Fig. 9
ergy transaction in the P2P arrangement. As compare to using the grid
shows that the RP's demand is supplied by a small contribution from the
as sole power source, the RP can achieve a total daily cost reduction of
battery (P2) and from the grid (P4) because the cost of energy is not that
62.71% in summer and 68.99% in winter; while the CP can achieve
high.
81.31% and 31.69% in winter. Part of these cost reductions are due to
Fig. 10 shows almost the same operation strategy for the CP as for
the saving brought by the internal power exchanged between prosu-
the RP, where the demand is met by the power shared from the RP's
mers. The same methodology can be performed for each day of the year
battery (P6) and a contribution from the grid (P10).
to compute the annual potential cost saving per prosumer.
As shown on Fig. 11, it is clear that for the two prosumers, the
These cost savings are mainly dependent on the initial capital cost
battery is used to decrease the operation cost and the reliance from the
determining the sizes of the system to be installed. Therefore, further
grid when the PV system are not in operation. The battery is supplying a
analyses, using the LLC, are needed to determine the overall economic
portion of the two prosumers’ demands trough P2 and P6 while its
benefit of each prosumer taking part in the project.
corresponding SoC is slightly decreasing. The battery could have been
used to supply the entire demands, without the use of the grid, but the
fixed-final state condition influences the operation to keep the final SoC 6.2. Life cycle cost analysis
value equal to the initial one (Eq. (14)).
With the purpose to estimate the overall costs of each prosumer in
the studies P2P energy sharing arrangement, the LLC analysis is per-
5.2.7. P2P operation during the off-peak pricing interval (green): 22h00- formed for a 20 years’ project. This analysis is also used to determine
23h59 break-even point (BEP) of each prosumer as compared to the grid used
During this off-peak period, there is no energy produced by the PV as baseline.
systems and the battery has reached its fixed-final state condition linked For each prosumer in the P2P arrangement, taking into account the
to the SoC. Given the fact that both the load demands and the electricity different components’ costs, the LCC can be calculated as:
price are low, the grid is used to successfully supply both the RP and CP
through P4 and P10, respectively shown on Fig. 9 and Fig. 10. LCCi = Cinitial (i) + Crep (i) + CEC (i) + COM (i) − Csalvage (i) (23)

11
K. Kusakana Journal of Energy Storage 32 (2020) 101717

Fig. 13. Break-even point between the Residential Prosumer and the grid.

Fig. 14. Break-even point between the Commercial Prosumer and the grid.

Where, for each component: countries.

Cinitial is the initial cost of each component; A case study was selected in South Africa where the residential
Csalvage is the salvage cost linked for each component that has not prosumer is using a roof mounted photovoltaic system with a battery
reach its replacement time; energy storage while the commercial prosumer uses a dual-tracking
Crep is the replacement cost of each component; photovoltaic system. Using the available solar resources, the prosumers’
COM is the operation and maintenance cost of each component; load demands, the applicable grid tariffs per sectors, the costs of in-
CEC is the applicable net energy cost. ternal power exchanged, as well as the system component size; the
More details on the bill of quantity applicable for each component optimization model finds the optimal power flows of the solar systems,
can be found in Tables 3 and 4. For each year, an increase of 10% is of the grid, of the internal power exchanged, and optimal operation
applicable in electricity price supplied by the grid. The yearly operation pattern (charging-discharging regime) of the battery storage system
and maintenance cost is considered to be 1% of the capital cost initially that minimizes the operation cost. The results have demonstrated that
invested; this increases using a yearly average inflation rate of 5.3% the P2P scheme assists the prosumers to benefit from their own energy
[29]. generated and substantial cost can be saved as compare to using the
Using the grid as baseline, Fig. 13 shows that the break-even point grid as sole power source. The residential prosumer has achieved a total
of the proposed RP in the P2P arrangement will occur at 16.35 years daily cost reduction of 62.71% in summer and 68.99% in winter; while
corresponding to R221 800. After the break-even point occurrence, the the commercial prosumer has achieved 81.31% and 31.69% in winter.
savings computed for the remaining system lifetime, as compared to For a 20 years’ project, using the grid as baseline for the compar-
using the grid, is R65 100 or 19.5%. ison, the lifecycle cost analysis have projected the residential prosumer
For the CP, Fig. 14 shows that the break-even point occurs after 5.3 to break-even after 16.35 years and save 19.5%; while the commercial
years corresponding to R150 900. For the considered project's lifetime, prosumer can break-even after 5.3 years and save 55.2%.
a saving of R639 300 or 55.2% can be achieved by the CP in the pro- The model developed in this study shows the ability to simulate
posed operation environment. systems with different components, load demands, resources as well as
costs. Therefore, it may assist researchers and engineers interested in
7. Conclusion the optimal operation of prosumers in Peer-to-Peer energy sharing
schemes applicable to many countries in the same situation as South
In this paper, a model to optimize the operation of grid-connected Africa. It can be used to analyze the cost incurred using different sys-
prosumers in a Peer-to-Peer energy sharing scheme is developed. The tems configuration; analyze the impact of different battery storage
developed model minimized both prosumers’ operation costs by max- types; analyze the effect of different cost structures allocated to the
imizing the use of the power from the renewable energy sources; op- internal power flows between prosumers or even to conduct sensitivity
timally managing the internal power sharing between the prosumers; analysis on the different parameters of the system.
and minimizing the use of the electrical supplying energy under the The main aim of this work was to minimize the net energy cost of
Time-of-Use rates. Additional constraints restricting the power flows each prosumers through the optimal economic power dispatch of the
from the prosumers to the grid were also included in the model to re- different components in the system submitted to the dynamic resources
plicate the environment in which several prosumers operate, in several and pricing structure. The results are dependent on the size of the

12
K. Kusakana Journal of Energy Storage 32 (2020) 101717

different components. Therefore, the multi-objective optimal sizing and [14] T. Sousa, T. Soares, P. Pinson, F. Moret, T. Baroche, E. Sorin, Peer-to-peer and
operation of the different components of such system will be part of the community-based markets: a comprehensive review, Renew. Sustain. Energy Rev.
104 (2019) 367–378.
scope of future work. [15] W. Inam, D. Strawser, K.K Afridi, R. J. Ram, D.J. Perreault, Architecture and system
analysis of microgrids with peer-to-peer electricity sharing to create a marketplace
CRediT authorship contribution statement which enables energy access, 2015 9th International Conference on Power
Electronics and ECCE Asia (ICPE-ECCE Asia), IEEE, 2015, pp. 464–469.
[16] R. Alvaro-Hermana, J. Fraile-Ardanuy, J Pedro, L.K. Zufiria, J. Davy, Peer to peer
K. Kusakana: Resources, Data curation, Writing - original draft. energy trading with electric vehicles, IEEE Intell. Transport. Syst. Maga. 8 (3)
(2016) 33–44.
[17] N. Liu, X. Yu, C. Wang, C. Li, L. Ma, J. Lei, Energy-sharing model with price-based
Declaration of Competing Interest demand response for microgrids of peer-to-peer prosumers, IEEE Trans. Power Syst.
32 (5) (2017) 3569–3583.
[18] M.S. Rahman, A.T. Oo, Distributed multi-agent based coordinated power manage-
The authors declare that they have no known competing financial
ment and control strategy for microgrids with distributed energy resources, Energy
interests or personal relationships that could have appeared to influ- Convers. Manage. 139 (2017) 20–32.
ence the work reported in this paper. [19] C. Zhang, J. Wu, Y. Zhou, M. Cheng, C. Long, Peer-to-Peer energy trading in a
Microgrid, Appl. Energy 220 (2018) 1–12.
[20] C. Long, J. Wu, Y. Zhou, N. Jenkins, Peer-to-peer energy sharing through a two-
References stage aggregated battery control in a community Microgrid, Appl. Energy 226
(2018) 261–276.
[1] K. Kusakana, H.J. Vermaak, Hydrokinetic power generation for rural electricity [21] C. Long, J. Wu, Y. Zhou, N. Jenkins, Peer-to-peer energy sharing through a two-
supply: case of South Africa, Renew. Energy 55 (2013) 467–473. stage aggregated battery control in a community Microgrid, Appl. Energy 226
[2] K. Kusakana, Energy management of a grid-connected hydrokinetic system under (2018) 261–276.
Time of Use tariff, Renew. Energy 101 (2017) 1325–1333. [22] V.S.K.V. Harish, N. Anwer, A. Kumar, Modelling of Peer to Peer sharing of power
[3] M.S. Ismail, M. Moghavvemi, T.M.I. Mahlia, K.M. Muttaqi, S. Moghavvemi, within solar based DC microgrids, Trend. Mech. Eng. Technol. 8 (3) (2018) 44–48.
Effective utilization of excess energy in standalone hybrid renewable energy sys- [23] H. Mehrjerdi, Peer-to-peer home energy management incorporating hydrogen sto-
tems for improving comfort ability and reducing cost of energy: a review and rage system and solar generating units, Renew. Energy 156 (2020) 183–192.
analysis, Renew. Sustain. Energy Rev. 42 (2015) 726–734. [24] E.A. Jordan, K. Kusakana, L. Bokopane, Prospective architecture for local energy
[4] K. Kusakana, Impact of time of use tariff and demand profiles on prosumers in Peer- generation and distribution with Peer-to-Peer electricity sharing in a South African
to-Peer energy sharing scheme, 2019 Advances in Science and Engineering context, 2018 Open Innovations Conference (OI), IEEE, 2018, pp. 161–164.
Technology International Conferences (ASET, IEEE, 2019, pp. 1–7. [25] L. Bokopane, K. Kusakana, H.J. Vermaak, Energy management of a grid-intergrated
[5] C. Park, T. Yong, Comparative review and discussion on P2P electricity trading, hybrid Peer-to-Peer renewable charging station for electric vehicles, 2018 Open
Energy Procedia 128 (2017) 3–9. Innovations Conference (OI, IEEE, 2018, pp. 275–280.
[6] C. Zhang, J. Wu, M. Cheng, Y. Zhou, C. Long, A bidding system for peer-to-peer [26] K. Kusakana, Modeling an optimal Peer-to-Peer energy sharing between prosumers
energy trading in a grid-connected microgrid, Energy Procedia 103 (2016) in a South African context, 2019 IEEE Milan Powertech, IEEE, 2019, pp. 1–6.
147–152. [27] K. Kusakana, Optimal Peer-to-Peer energy sharing between prosumers using hy-
[7] Y. Zhou, J. Wu, C. Long, M. Cheng, C. Zhang, Performance evaluation of peer-to- drokinetic, diesel generator and pumped hydro storage, J. Energy Storage 26 (2019)
peer energy sharing models, Energy Procedia 143 (2017) 817–822. 101048.
[8] T. Sousa, T. Soares, P. Pinson, F. Moret, T. Baroche, E. Sorin, Peer-to-peer and [28] B.P. Numbi, S.J. Malinga, Optimal energy cost and economic analysis of a re-
community-based markets: a comprehensive review, Renew. Sustain. Energy Rev. sidential grid-interactive solar PV system-case of eThekwini municipality in South
104 (2019) 367–378. Africa, Appl. Energy 186 (2017) 28–45.
[9] H. Le Cadre, P. Jacquot, C. Wan, C. Alasseur, Peer-to-peer electricity market ana- [29] K. Kusakana, Energy management of a grid-connected hydrokinetic system under
lysis: from variational to generalized nash equilibrium, Eur. J. Opera. Res. 282 (2) Time of Use tariff, Renew. Energy 101 (2017) 1325–1333.
(2020) 753–771. [30] P.A. Hohne, K. Kusakana, B.P. Numbi, Optimal energy management and economic
[10] S. Löbbe, A. Hackbarth, The transformation of the German electricity sector and the analysis of a grid-connected hybrid solar water heating system: a case of
emergence of new business models in distributed energy systems, Innovation and Bloemfontein, South Africa, Sustain. Energy Technol. Assess. 31 (2019) 273–291.
Disruption at the Grid's Edge, Academic Press, 2017, pp. 287–318. [31] P.A. Hohne, K. Kusakana, B.P. Numbi, Scheduling and economic analysis of hybrid
[11] C. Burger, J. Weinmann, The Decentralized Energy revolution: Business Strategies solar water heating system based on timer and optimal control, J. Energy Storage 20
For a New Paradigm, Springer, 2012. (2018) 16–29.
[12] P. Siano, G.D. Marco, A. Rolán, V. Loia, A survey and evaluation of the potentials of [32] K. Kusakana, Optimal operation scheduling of grid-connected PV with ground
distributed ledger technology for peer-to-peer transactive energy exchanges in local pumped hydro storage system for cost reduction in small farming activities, J.
energy markets, IEEE Syst. J. 13 (3) (2019) 3454–3466. Energy Storage 16 (2018) 133–138.
[13] J.A. Abdella, K. Shuaib, An architecture for blockchain based Peer to Peer energy [33] S.P. Koko, K. Kusakana, H.J. Vermaak, Optimal power dispatch of a grid-interactive
trading, 2019 Sixth International Conference on Internet of Things: Systems, micro-hydrokinetic-pumped hydro storage system, J. Energy Storage 17 (2018)
Management and Security (IOTSMS), IEEE, 2019, pp. 412–419. 63–72.

13