Power Hardware-In-the-Loop Testing of a Peer-to-Peer Energy Trading Framework

Abstract

This paper demonstrates the value of power hardware-in-the-loop (PHIL) testing for the study of peer-to-peer (P2P) energy trading. P2P has emerged as a promising candidate for coordinating large numbers of distributed energy resources (DER) that pose a risk to network operations if left unmanaged. The existing literature has so far relied on pure software simulations to study DER and distribution networks within this context. This requires the development of simplified models for complex components due to the computational limitations involved. Issues that arise through the operation of physical hardware in real-world applications are therefore neglected. We present PHIL testing as a solution to this problem by exhibiting its ability to capture the complex behaviors of physical DER devices. A high-fidelity PHIL test environment is introduced that combines key hardware elements with a simulated network model to study a P2P trading scenario. The initial findings reveal several underlying challenges of coordinating DER that are not typically discussed in prior works.

Full text

ACCEPTED FOR PRESENTATION IN 11TH BULK POWER SYSTEMS DYNAMICS AND CONTROL SYMPOSIUM (IREP 2022),
JULY 25-30, 2022, BANFF, CANADA 1
Power Hardware-In-the-Loop Testing of a
Peer-to-Peer Energy Trading Framework
Thomas Perrau, Maksim Stojkovic and Gregor Verbiˇ
c
School of Electrical and Information Engineering
The University of Sydney
Sydney, Australia
Abstract—This paper demonstrates the value of power
hardware-in-the-loop (PHIL) testing for the study of peer-to-
peer (P2P) energy trading. P2P has emerged as a promising
candidate for coordinating large numbers of distributed energy
resources (DER) that pose a risk to network operations if left
unmanaged. The existing literature has so far relied on pure
software simulations to study DER and distribution networks
within this context. This requires the development of simplified
models for complex components due to the computational limita-
tions involved. Issues that arise through the operation of physical
hardware in real-world applications are therefore neglected. We
present PHIL testing as a solution to this problem by exhibiting
its ability to capture the complex behaviors of physical DER
devices. A high-fidelity PHIL test environment is introduced that
combines key hardware elements with a simulated network model
to study a P2P trading scenario. The initial findings reveal several
underlying challenges of coordinating DER that are not typically
discussed in prior works.
Index Terms—Power Hardware-In-the-Loop, Distributed En-
ergy Resources, Peer-to-Peer Energy Trading, Real-Time Digital
Simulation, DER Coordination, Energy Prosumer
I. INTRODUCTION
Distributed energy resources (DER) are quickly becoming
a critical component of modern electricity networks. The
Australian Electricity Market Operator (AEMO) projects that
the installed capacity of DER will double or triple by 2040
and form between 13% to 22% of all electricity generation
in the National Electricity Market [1]. This level of growth
is also being observed in many other jurisdictions around the
world [2].
High penetrations of DER lead to violations of network
constraints when not managed effectively [2]–[4]. This has
sparked research into methods of coordinating large numbers
of these devices that preserve system integrity and incentivize
efficient usage. Such methods aim to strike a balance between
maximizing social welfare and minimizing network instability.
There are several leading approaches towards solving this
emerging problem. Among these is peer-to-peer (P2P) energy
trading, which employs a market-driven strategy to encourage
efficient DER operation. Under this strategy, energy can be
transacted directly between participating energy prosumers and
traditional consumers. Excess generation from DER can be
routinely auctioned to potential buyers who wish to purchase
energy at lower prices than the retail value. Buyers and
sellers are matched during predetermined trading periods,
with transacted energy then being dispatched to fulfill the
conditions of each executed trade.
This structure brings added efficiency to several key areas
of electricity networks. Local supply-demand balancing is
encouraged through the market clearing process which reduces
overall system losses [3]. P2P trading reliably converges to
optimal pricing in each trading window which has a positive
effect on social welfare [5], [6]. Other benefits include the
improved utilization of excess generation (or reduced energy
spillage) and clear incentives for new DER installations [7].
A key limitation of current P2P trading frameworks is the
exclusion of network constraints. Trading without oversight
from a network operator can lead to violations of voltage,
frequency and thermal operating limits [5]. Methods of inter-
vention by an entity such as a Distribution System Operator
(DSO) have subsequently been proposed in [7] to address this.
A review of existing literature has revealed that the testing
of P2P DER coordination methods has largely been confined to
simulations and real-world trials. While such approaches do
yield valuable insight into overall performance, each comes
with distinct limitations. Pure software simulations allow for
a wide range of operating scenarios to be tested but often lack
the ability to model complex system elements accurately. In
contrast, real-world trials provide the most accurate test setting
but at the cost of flexibility to model a range of potential
operating scenarios.
Marrying the two approaches - explicit hardware device
testing and software simulations - can yield the benefits of both
testing methods [8]. This has led to the development of hybrid
Power Hardware-In-the-Loop (PHIL) testing techniques which
have garnered significant attention in this research space.
The complex behavior exhibited by DER devices presents a
challenge that is ideal for the application of this technology.
This observation motivated us to apply state-of-the-art
PHIL techniques to the study of P2P energy trading. A full
hardware prosumer model - containing local PV generation,
battery storage and variable load - was interfaced through a
Spitzenberger & Spies linear power amplifier to a RTDS real-
time digital simulator running a simple distribution network
model. Equipment was selected to maximize the accuracy of
operating data transferred between the software and hardware
test components. A wireless communication network was also
implemented through the use of Raspberry Pi devices to
replicate real-world data collection conditions.
arXiv:2207.08009v1 [eess.SY] 16 Jul 2022

ACCEPTED FOR PRESENTATION IN 11TH BULK POWER SYSTEMS DYNAMICS AND CONTROL SYMPOSIUM (IREP 2022),
JULY 25-30, 2022, BANFF, CANADA 2
This paper aims to demonstrate the efficacy of P2P trading
for coordinating multiple prosumers in a high-fidelity envi-
ronment. Specific attention is given to the effects of hardware
devices on the overall performance of the scheme.
The remainder of this paper is structured as follows. In
Section II, a detailed description of the P2P energy trading
framework is provided. Section III summarizes the challenges
and design considerations of PHIL test environments. Section
IV then gives an overview of the PHIL test setting used
to study P2P energy trading. Preliminary results are then
presented in Section V to demonstrate the performance of the
P2P trading approach on a simple radial distribution system.
Concluding remarks are provided in Section VI.
II. P2P ENERGY TRADING
P2P energy trading differs from traditional electricity market
structures by allowing prosumers to trade energy directly with
neighboring users [7]. This has been found to introduce many
efficiencies to both the transactive environment and physical
generation of electricity [3], [5], [7], [9]–[13]. In particular,
P2P facilitates:
•improved incentives for prosumers compared to tradi-
tional retail arrangements,
•reduced spillage of DER generation due to improved
prosumer coordination,
•lower grid losses by encouraging shorter transmission
distances between consumers and generation sources, and
•less reliance on intermediaries to perform transactions.
These benefits can be largely attributed to a core outcome of
P2P energy trading: the transition from a centralized architec-
ture to a decentralized architecture. While this in particular
applies to the way electricity is generated and transacted,
it also extends to other components of electricity networks.
Recent work has identified the need for decentralized control,
computation and communication infrastructure to overcome
issues of scalability when coordinating large numbers of DER
[7], [9], [14].
From a market perspective, there have been several candi-
date methods proposed in the literature for hosting P2P energy
trading. These largely differ in the overall market structure
used and the market operating strategy employed. Market
structures range from discrete-time call auctions to continuous
double auctions (CDAs), with each method offering different
levels of scalability, awareness of network constraints, and
computational overheads [9]. Operating strategies range from
centralized architectures that employ cloud-based infrastruc-
tures to decentralized approaches that use distributed ledger
technologies (DLTs) such as blockchains [15].
The use of a CDA has gained particular attention for
being the chosen format for stock and commodity markets
around the globe [16]–[18]. In addition, CDAs composed
of rational participants are Pareto-improving and result in a
highly efficient allocation of resources [7], [16]. This makes
them an attractive option for optimizing DER usage in a P2P
trading environment.
Fig. 1. Illustration of P2P energy trading among participating prosumers [10].
Under a CDA P2P format, buyers and sellers of energy
are matched according to a set of pre-defined trading rules.
Participants lodge either buy orders (bids) or sell orders (asks)
for electricity at any time during each trading period. Sellers
of energy are solely composed of prosumers with dispatchable
DER. Priority is always given to the best standing offer
(highest bid or lowest ask), with matches occurring when the
best bid exceeds the best ask at any given time [17]. If no
match is found for a new order, it is stored until either a
match is found or the current trade period ends, in which
case, all outstanding orders are canceled. All executed trades
are settled immediately after each trading interval, with sellers
dispatching energy according to each contract. Any energy
not secured in the P2P market to satisfy a trader’s demand is
typically purchased through a traditional retailer.
Market participants are frequently modeled as Zero-
Intelligence Plus (ZIP) traders to represent human decision-
making in an auction setting [17], [18]. This assumes that
orders are submitted randomly by participants without any
detailed underlying strategy. Bid and ask prices are instead
adjusted by each trader based on the most recently matched
orders. For example, traders with a bid price lower than the
last executed order will decrease their margins by raising the
previous bid amount.
Despite demonstrating excellent performance for improving
social welfare in energy markets, several drawbacks of P2P
energy trading schemes have been identified. Foremost among
these is the absence of network constraint considerations and
the spillage of DER generation that is not traded in the
P2P market [3]. Current research is being conducted into
overcoming these challenges and transforming theoretical P2P
approaches into real-world solutions.
III. PHIL TES TI NG
PHIL testing seeks to gain the benefits of pure software sim-
ulations and real-world hardware experimentation by combin-
ing elements from both methods [8]. Traditional pure-software
simulations offer a high degree of flexibility for power system
studies by allowing the analysis of a wide range of scenarios.
However, this comes at the cost of accuracy due to the
need to develop software models for highly-complex hardware
devices. In contrast, real-world testing offers the reverse; the
presence of physical hardware yields high accuracy but lowers
testing flexibility due to the risks of damaging equipment

ACCEPTED FOR PRESENTATION IN 11TH BULK POWER SYSTEMS DYNAMICS AND CONTROL SYMPOSIUM (IREP 2022),
JULY 25-30, 2022, BANFF, CANADA 3
and endangering personnel. PHIL testing embeds complex
hardware devices within a software environment capable of
performing simulations in real time, resulting in both a highly
accurate and flexible test environment.
Interest in PHIL testing methods has been increasing within
the field of power systems engineering [8], [19]–[28]. Tech-
nological advancements in Real-Time Simulation (RTS) and
power interfacing devices have improved the fidelity of PHIL
test environments for studying the complex dynamic behavior
of electricity networks [29], [30]. This has led to recent appli-
cations of PHIL techniques for the study of DER integration
due to the complexity of the hardware interactions involved.
The two primary challenges of PHIL testing are the preser-
vation of system stability and simulation accuracy [8], [24],
[25], [27], [30]–[34]. As PHIL consists of high-power signals,
instability can quickly lead to equipment damage and the
endangerment of personnel [31]. The ideal test environment
should therefore be capable of capturing system behavior with
high accuracy while avoiding unbounded growth in any of the
signals involved.
This is difficult in practice due to the inherent errors
introduced by the equipment used. In particular, the interface
between the hardware and software components of the test sys-
tem has a significant impact on overall stability and accuracy
[27]. Errors are introduced through sensor noise, time delays,
finite signal sampling rates and the response characteristics
of the power amplifier device [31]. These culminate in lower
quality test results and an increased risk of system collapse
when amplified through the power interface device.
Recent work has been done on quantifying the effects
of individual interface elements on overall PHIL stability
[30], [32], [33]. This has allowed the time delays introduced
by different power amplifiers, communication methods, RTS
devices and other components to be directly contrasted when
making design decisions. Improved stability and accuracy
analysis are gained, which can be used to establish concrete
operating boundaries for a particular PHIL test environment.
Simulation errors can be mitigated through the use of more
advanced equipment. However, this comes with additional
cost and may not always be possible within the budget of
a development team. The trade-off between cost and system
accuracy is, therefore, one of the major considerations in the
field of PHIL testing.
IV. PH IL T ES TI NG SE T-UP
The PHIL test environment at the University of Sydney was
designed to study DER coordination methods. The general
concept was to model a typical residential prosumer entirely
in hardware for connection to a real-time simulated distribu-
tion network model. This would allow all complex hardware
interactions to be captured explicitly within the simulation
environment. Their effects on the wider network could then
be observed under different operating scenarios.
The typical prosumer was assumed to possess on-roof PV
generation, local energy storage, and a variable load. This
Fig. 2. Typical arrangement of a PHIL test system [25].
Fig. 3. Basic signal block diagram of a PHIL interface using the Ideal
Transformer model [27].
was realized in hardware using the following key pieces of
equipment:
•1 × Spitzenberger & Spies APS5000 Four-Quadrant Am-
plifier
•1 × SMA Sunny Boy Solar Inverter
•1 x Spitzenberger & Spies PVS3000 Photovoltaic Simu-
lator
•1 × SMA Sunny Island Battery Inverter
•1 × LG Chem RESU10H Lithium-Ion Battery
•1 × Chroma 63803 Electronic Load Bank
The communications required between the market platform1
and each prosumer were also modeled explicitly in hardware.
To do this, four other neighboring prosumers were simulated
on Raspberry Pi devices and interfaced wirelessly to a central
computer hosting the CDA trading platform. This allowed
the effects of communication delays and data loss on system
operations to be directly incorporated.
An RTDS Technologies Mid-Size Cubicle was used to
implement the simulated distribution model. The model design
was done using the corresponding RSCAD graphical user
interface. The real-time monitoring and control of network
simulations were also done through this software.
The general arrangement of the PHIL test environment is
shown in Figure 4.
1The market platform serves the sole purpose of matching buyers and
sellers. This role can be viewed as that of an auctioneer in a local energy
market. It is oblivious to any overarching market structures, and so can be
considered truly decentralized.

ACCEPTED FOR PRESENTATION IN 11TH BULK POWER SYSTEMS DYNAMICS AND CONTROL SYMPOSIUM (IREP 2022),
JULY 25-30, 2022, BANFF, CANADA 4
Fig. 4. University of Sydney PHIL test environment design.
A key goal of this test environment was to obtain a very high
level of accuracy for observing prosumer behavior. As a result,
the interface between the software and hardware components
were made to minimize time delays and noise as much as
possible.
A linear power amplifier was chosen to achieve higher
dynamic performance than switched-mode and synchronous
generator amplifiers seen in other PHIL installations [30].
This amplifier operates with a slew rate of 52V/µs, which
corresponds to time delays under 5µs for most operations
[35]. In contrast, the switched-mode amplifier used in [33]
was found to introduce a time delay of 66.67µs. Improved
system accuracy and stability was gained as a result.
The analogue connections between the RTDS and power
amplifier were replaced with a fiber-optic Aurora link after
initial testing. This is known to be faster, more efficient
and does not require anti-aliasing filters when compared to
analogue interfaces [33]. The work in [32] found a time delay
of <1µs for the Aurora protocol and 20µs for a standard
analogue connection.
The installation of the Aurora link was done in response
to stability and accuracy issues uncovered during the initial
testing. The results in Section V were therefore obtained using
analogue communications, with a discussion on the potential
impact presented in Section V-B
V. PHIL TES TI NG RE SU LTS
Preliminary testing of a basic P2P market framework was
conducted on the PHIL test environment to verify its perfor-
mance. Tests were done prior to the installation of the Aurora
link using a standard analogue interface as shown in Figure 4.
The P2P market operating algorithm was modeled off the
approach developed in [7]. A CDA with hourly trading periods
was used as the market structure and market participants
were assumed to be ZIP traders. The matching of orders
by the market platform was simulated using Algorithm 1.
Participating ZIP energy traders updated their asks and bids
in response to market events using Algorithm 2.
Any energy not secured through the P2P market to satisfy
demand was purchased at the time-of-use (ToU) rates of a
typical retailer. Likewise, surplus PV generation not able to
be sold in the market was instead settled at a standard retail
feed-in-tariff (FiT). The pricing used in the simulation was
adapted from the 2021 residential standing offer of a large
Australian energy retailer (AGL) [36].
A simple low-voltage distribution line served as the sim-
ulated network model (Figure 5). The line was connected
through an ideal 11/0.4kV Dyn1 distribution transformer to
the wider network. Line conductors were assumed to be 7/4.50
single-core aerial bare aluminum, with impedances derived
from [37]. The five single-phase market participants used in
the simulation were connected to the network as shown in
Figure 5. Connections to each household were staggered over
the three phases to reflect real-world residential connection

ACCEPTED FOR PRESENTATION IN 11TH BULK POWER SYSTEMS DYNAMICS AND CONTROL SYMPOSIUM (IREP 2022),
JULY 25-30, 2022, BANFF, CANADA 5
Algorithm 1 CDA Order Matching Process
1: while market is open during time slot do
2: randomly select a new ZIP trader
3: .Process orders with given (price, quantity, time)
4: if trader is buyer then
5: add new order ob(pb, qb, tb)to order book
6: else
7: add new order os(ps, qs, ts)to order book
8: end if
9: .Match orders based on price-time priority
10: while pbest
s≤pbest
bdo
11: if tbest
b≤tbest
sthen
12: trade min(qbest
b, qbest
s) at pbest
b
13: else
14: trade min(qbest
b, qbest
s) at pbest
s
15: end if
16: subtract traded quantity from obest
band obest
s
17: remove obest
band/or obest
sfrom book if filled
18: update trader bidding strategies
19: end while
20: update trader bidding strategies
21: end while
Algorithm 2 ZIP Trader Bid Update Process
1: for trader in market during time slot tdo
2: .Update buyer profit margins
3: if trader is buyer then
4: if olast was matched at price ptrade then
5: if pb≥ptrade then
6: decrease bid price pb
7: else if olast is sell order and pb≤ptrade then
8: increase bid price pbif pb≤ptrade
9: end if
10: else if olast is buy order and pb≤ptrade then
11: increase bid price pb
12: end if
13: .Update seller profit margins
14: else
15: if olast was matched at price ptrade then
16: if ps≤ptrade then
17: increase bid price ps
18: else if olast is buy order and ps≥ptrade then
19: decrease ask price ps
20: end if
21: else if olast is sell order and ps≥ptrade then
22: decrease ask price ps
23: end if
24: end if
25: end for
TABLE I
IMPEDANCES OF DISTRIBUTION LINE SEGMENTS
Line Segment Distance (m) Impedance (Ω)
Line 1 75 0.0239 + 0.0218j
Line 2 40 0.0128 + 0.0116j
Line 3 40 0.0128 + 0.0116j
Line 4 40 0.0128 + 0.0116j
Line 5 40 0.0128 + 0.0116j
TABLE II
BATTE RY AND P V CAPACITIES BY HOUSEHOLD
Household
Number
Hardware
Model
Battery Capacity
(kWh)
PV Capacity
(kW)
HEMS
1 Yes 7.5 3 Yes
2 No 7.5 5 Yes
3 No None 5 No
4 No None None No
5 No 7.5 5 Yes
TABLE III
VALUE C AP TUR ED B Y EACH M AR KET PA RTIC IPAN T
Buyer Seller Total
Daily Value ($) 2.62 0.19 2.81
Expected Annual Value ($) 958.07 68.41 1026.48
Expected Annual Value per House-
hold ($)
191.61 13.68 205.30
Proportion of Value Captured (%) 93.34 6.66 100.00
standards [38]. Line impedances separating each model ele-
ment are shown in Table I.
Three of the households had on-site PV generation, battery
storage and a home energy management system (HEMS). The
HEMS was modeled as a simple linear programming (LP)
optimization problem with an objective function aimed at
maximizing the savings of the prosumer. The forecasting of
demand and PV generation at the beginning of the simulated
period was assumed to be deterministic for the sake of simplic-
ity. The effect of forecasting errors was, therefore, outside the
scope of this study. The expected income from energy trading
was also not included in the HEMS, with only retail pricing
being used to form the objective cost function.
Of the two remaining households, one had only on-site PV
generation and the other was a regular electricity consumer.
The chosen battery and PV capacities for each household are
shown in Table II.
The P2P market was operated over a one day period to
verify its performance. Each prosumer was allocated an hourly
PV generation profile and load profile to align with the chosen
trading period duration. The profiles were adapted from actual
meter data from Australian residences published by a large
Australian distribution network operator (Ausgrid) [39].
A. P2P Trading Results
The trading activity resulting from the simulation is shown
in Figure 7. The total value gained over a traditional retail
arrangement for each household is shown in Table III.

ACCEPTED FOR PRESENTATION IN 11TH BULK POWER SYSTEMS DYNAMICS AND CONTROL SYMPOSIUM (IREP 2022),
JULY 25-30, 2022, BANFF, CANADA 6
Fig. 5. Simulated network model used for initial PHIL testing.
Fig. 6. Trades executed between households over the simulation period.
Figure 7c shows that the total exports far exceeded imports
over the course of the simulation. This signifies that a surplus
of power generation was present during most periods, par-
ticularly during the daytime due to abundant PV generation.
This caused a large number of trades to occur from prosumer
households (Households 1, 2, 3 and 5) selling excess gener-
ation to the only available energy consumer (Household 4).
Household 3 became available as a buyer outside of daylight
hours due its the lack of on-site energy storage.
Figure 7a shows that all trades were executed at prices under
10c/kWh. This was despite a FiT of 6.1c/kWh and a peak ToU
price of 49.24c/kWh. The low execution prices were attributed
to the large surplus of energy supply driving lower prices in the
competitive market environment. The majority of value was
therefore unlocked by buyers of electricity, as demonstrated
in Table III.
Figure 7b reveals that potential buyers often chose to pur-
chase power at the retail ToU rate instead of the best available
ask price. This always resulted in a higher price being paid for
electricity. The irrational decision-making shown here high-
lights a key shortcoming of the chosen ZIP trader algorithm:
buyers do not account for the inelastic nature of electricity
demand. Most markets (such as stock markets) allow traders
the freedom to forego transactions without incurring a direct
loss at the end of a trade period. In contrast, P2P electricity
markets require each participant to secure enough supply to
meet demand during all trading intervals. This leads to traders
being forced to purchase shortages of electricity at a higher
retail price if not acquired elsewhere.
B. PHIL Testing Behaviour
The hardware prosumer model was allocated to Household
1 in the simulated scenario. The power interface was initially
designed to transfer only the voltage at the grid connection
point from the simulated network model to the hardware
environment. The current measured at the amplifier output was
then communicated back to the network model to complete
the feedback loop, creating an arrangement similar to that
shown in Figure 3. The communication of the signals was
done through a basic analogue connection.
The simulation was found to be unstable under this ar-
rangement. This was largely attributed to the high harmonic
distortion present in the outputs of the commercial PV and
battery inverters (Figure 8). It was concluded that reductions
to the total time delay of the PHIL feedback loop would be
needed to properly capture this behavior. The use of shorter
simulation time steps and fiber communication over analogue
connections are being explored as potential solutions.
Stability was obtained in the meantime by processing the
measured hardware signals in RSCAD. To do this, voltage and
current were measured synchronously from the power ampli-
fier and communicated to the simulation environment. These
signals were then converted artificially into pure active and
reactive power values using the in-built PQ meter block. The
effect of this transformation on removing harmonic distortion
can be observed in Figure 8a.
RSCAD uses equations (1) and (2) to do this operation:
P=1
TZT
t−T
V(ωt)I(ωt)dt (1)
Q=1
TZT
t−T
V(ωt)I(ωt −π
2)dt (2)

ACCEPTED FOR PRESENTATION IN 11TH BULK POWER SYSTEMS DYNAMICS AND CONTROL SYMPOSIUM (IREP 2022),
JULY 25-30, 2022, BANFF, CANADA 7
(a)
(b)
(c)
Fig. 7. 7a) best bid and ask prices during each trading period, 7b) best
unmatched ask and bid following period, and 7c) power imported by each
participant.
This method enables waveform distortion to be factored into
the result by explicitly handling non-sinusoidal conditions, as
discussed in [40].
This is in contrast to calculating power values in (3) and (4),
as done on the human-machine interface (HMI) of the power
amplifier:
P=V I cos θ(3)
Q=V I sin θ(4)
These assume that no distortion is present by only con-
sidering the fundamental frequency of the waveform. It was
discovered that this yielded values considerably different to
real-time integration when applied to the measured signals.
(a)
(b)
Fig. 8. 8a) comparison of Sunny Boy PV inverter distortion with the corrected
waveform using the RSCAD PQ block, and 8b) current oscillations introduced
by the Sunny Island battery inverter.
Fig. 9. Measured vs expected active power injection from the hardware
prosumer model.
The non-sinusoidal conditions created by inverter devices
can therefore have a large impact on meter device readings
depending on the calculation method. This could have real-
world implications in networks with a high penetration of
DER.
The PHIL system maintained stability under a wide range
of conditions using the PQ block conversion. Active power
settings issued to the battery inverter in response to P2P trading
closely matched values measured in the network model, as
shown in Figure 9. However, discrepancies could still be
observed between the power traded in the P2P market and
the amount actually dispatched. There is a need to further
investigate methods of resolving these instances within the
market framework.

ACCEPTED FOR PRESENTATION IN 11TH BULK POWER SYSTEMS DYNAMICS AND CONTROL SYMPOSIUM (IREP 2022),
JULY 25-30, 2022, BANFF, CANADA 8
VI. CONCLUSION
P2P energy trading is a viable alternative for coordinating
DER. It incentivizes prosumers to preserve the local supply-
demand balance through the underlying market mechanism.
The competition created between buyers and sellers fosters
more efficient pricing and allocations of resources when com-
pared to traditional approaches.
Challenges are introduced through the underlying interac-
tions between physical hardware devices. These included the
high levels of harmonic distortion introduced by commercial
PV and battery inverters and mismatches between traded
energy and amounts actually dispatched. The non-sinusoidal
conditions created by DER also highlight the importance of
the method chosen for measuring real and reactive power. All
of these issues are particular to the hardware used and are not
easily replicable in pure software. This highlights the need for
PHIL testing in the study of DER coordination strategies.
Capturing the harmonic distortion from inverter devices
presented a challenge for the PHIL test environment. It was
discovered that a highly-ideal power interface is necessary to
preserve system stability when capturing this behavior. Re-
ducing time delays through improved communication methods
and smaller simulation time steps are being explored towards
this end. The level of distortion observed could also have
real-world ramifications for distribution networks and warrants
additional investigation.
The effects of wireless communication methods on the scal-
ability of P2P energy trading are yet to be investigated. Future
studies will employ the RPi devices to explore the impact of
data loss and communication delays. This will provide insight
into the feasibility of large-scale P2P applications.
Further stability analysis is required to establish the exact
operating limits of the PHIL test environment. Quantifying
stability margins is critical for gauging which power system
behaviors can be studied. The information can also be used to
make informed decisions on future PHIL designs.
Finally, the inelastic nature of electricity demand was found
to be problematic when modeling prosumer bidding strategies.
ZIP traders were discovered to display irrational behavior
when used in an energy market setting. Opportunities to pur-
chase electricity in the CDA were routinely declined in favor
of more expensive retail tariffs, leading to lower economic
benefits for all parties. Future work is needed to refine the
simulation of prosumers in P2P markets.
REFERENCES
[1] Australian Electricity Market Operator (AEMO), “2020 integrated sys-
tem plan,” AEMO, Tech. Rep., 2020.
[2] Newport Consortium, “Coordination of distributed energy resources;
international system architecture insights for future market design,”
Newport Consortium, Tech. Rep., 2018.
[3] J. Guerrero, D. Gebbran, S. Mhanna, A. C. Chapman, and G. Verbiˇ
c,
“Towards a transactive energy system for integration of distributed
energy resources: Home energy management, distributed optimal power
flow, and peer-to-peer energy trading,” Renewable & sustainable energy
reviews, vol. 132, pp. 110000–, 2020.
[4] M. M. Haque and P. Wolfs, “A review of high PV penetrations in lv
distribution networks: Present status, impacts and mitigation measures,”
Renewable & sustainable energy reviews, vol. 62, pp. 1195–1208, 2016.
[5] M. Khorasany, Y. Mishra, and G. Ledwich, “A decentralized bilateral
energy trading system for peer-to-peer electricity markets,” IEEE trans-
actions on industrial electronics (1982), vol. 67, no. 6, pp. 4646–4657,
2020.
[6] S. Thakur, B. P. Hayes, and J. G. Breslin, “Distributed double auction for
peer to peer energy trade using blockchains,” in 2018 5th International
Symposium on Environment-Friendly Energies and Applications (EFEA).
IEEE, 2018, pp. 1–8.
[7] J. Guerrero, A. C. Chapman, and G. Verbiˇ
c, “Decentralized p2p energy
trading under network constraints in a low-voltage network,” IEEE
transactions on smart grid, vol. 10, no. 5, pp. 5163–5173, 2019.
[8] E. de Jong, R. Graaff, P. Vaessen, P. Crolla, A. Roscoe, F. Lehfuss,
G. Lauss, P. Kotsampopoulos, and F. Gafaro, European White Book on
Real-Time Power Hardware-in-the-loop testing, 2012.
[9] M. Khorasany, Y. Mishra, and G. Ledwich, “Market framework for local
energy trading: a review of potential designs and market clearing ap-
proaches,” IET generation, transmission & distribution, vol. 12, no. 22,
pp. 5899–5908, 2018.
[10] Y. Liu, L. Wu, and J. Li, “Peer-to-peer (p2p) electricity trading in
distribution systems of the future,” The Electricity journal, vol. 32, no. 4,
pp. 2–6, 2019.
[11] W. Tushar, C. Yuen, H. Mohsenian-Rad, T. Saha, H. V. Poor, and K. L.
Wood, “Transforming energy networks via peer-to-peer energy trading:
The potential of game-theoretic approaches,” IEEE signal processing
magazine, vol. 35, no. 4, pp. 90–111, 2018.
[12] J. Guerrero, A. Chapman, and G. Verbiˇ
c, “A study of energy trading in
a low-voltage network: Centralized and distributed approaches,” in 2017
Australasian Universities Power Engineering Conference (AUPEC), vol.
2017-. IEEE, 2017, pp. 1–6.
[13] Blockchain technology innovations in business processes, ser. Smart
innovation, systems, and technologies ; Volume 219. Singapore:
Springer, 2021.
[14] N. Rahbari-Asr and M.-Y. Chow, “Cooperative distributed demand man-
agement for community charging of phev/pevs based on kkt conditions
and consensus networks,” vol. 10, no. 3, pp. 1907–1916, 2014.
[15] M. R. Alam, M. St-Hilaire, and T. Kunz, “Peer-to-peer energy trading
among smart homes,” Applied energy, vol. 238, pp. 1434–1443, 2019.
[16] D. K. Gode and S. Sunder, “Allocative efficiency of markets with
zero-intelligence traders: Market as a partial substitute for individual
rationality,” The Journal of political economy, vol. 101, no. 1, pp. 119–
137, 1993.
[17] D. Friedman, The double auction market : institutions, theories, and
evidence. Boca Raton, FL: Routledge, an imprint of Taylor and Francis,
2018.
[18] J. D. Farmer, P. Patelli, and I. I. Zovko, “The predictive power of zero
intelligence in financial markets,” Proceedings of the National Academy
of Sciences - PNAS, vol. 102, no. 6, pp. 2254–2259, 2005.
[19] A. Kulmala, A. Angioni, S. Repo, D. D. Giustina, A. Barbato, and
F. Ponci, “Experiences of laboratory and field demonstrations of distri-
bution network congestion management,” in IECON 2018 - 44th Annual
Conference of the IEEE Industrial Electronics Society, 2018, pp. 3543–
3549.
[20] J. Montoya, R. Brandl, K. Vishwanath, J. Johnson, R. Darbali-Zamora,
A. Summers, J. Hashimoto, H. Kikusato, T. S. Ustun, N. Ninad,
E. Apablaza-Arancibia, J.-P. B´
erard, M. Rivard, S. Q. Ali, A. Obushevs,
K. Heussen, R. Stanev, E. Guillo-Sansano, M. H. Syed, G. Burt, C. Cho,
H.-J. Yoo, C. P. Awasthi, K. Wadhwa, and R. Br¨
undlinger, “Advanced
laboratory testing methods using real-time simulation and hardware-in-
the-loop techniques: A survey of smart grid international research facility
network activities,” Energies (Basel), vol. 13, no. 12, pp. 3267–, 2020.
[21] C. Molitor, A. Benigni, A. Helmedag, K. Chen, D. Cali, P. Jahangiri,
D. Muller, and A. Monti, “Multiphysics test bed for renewable energy
systems in smart homes,” IEEE transactions on industrial electronics
(1982), vol. 60, no. 3, pp. 1235–1248, 2013.
[22] J. Wang, Y. Song, W. Li, J. Guo, and A. Monti, “Development of a
universal platform for hardware in-the-loop testing of microgrids,” IEEE
transactions on industrial informatics, vol. 10, no. 4, pp. 2154–2165,
2014.
[23] M. Barrag´
an-Villarejo, F. de Paula Garc´
ıa-L´
opez, A. Marano-Marcolini,
and J. M. Maza-Ortega, “Power system hardware in the loop (pshil): A
holistic testing approach for smart grid technologies,” Energies (Basel),
vol. 13, no. 15, pp. 3858–, 2020.

ACCEPTED FOR PRESENTATION IN 11TH BULK POWER SYSTEMS DYNAMICS AND CONTROL SYMPOSIUM (IREP 2022),
JULY 25-30, 2022, BANFF, CANADA 9
[24] G. Lauss, B. Bletterie, C. Mayr, F. Lehfuß, and R. Brundlinger, “Power-
hardware-in-the loop simulations for electrical generators in lv grids,”
AIT Austrian Institute of Technology and DERLab, Tech. Rep., 2013.
[25] F. L. Alexander Viehweider and G. F. Lauss, “Power hardware-in-
the-loop simulations for distributed generation,” in 21st International
Conference on Electricity Distribution, no. 0437, 2011.
[26] W. Ren, M. Steurer, and T. Baldwin, “An effective method for eval-
uating the accuracy of power hardware-in-the-loop simulations,” IEEE
transactions on industry applications, vol. 45, no. 4, pp. 1484–1490,
2009.
[27] M. Pokharel and C. N. M. Ho, “Stability study of power hardware in
the loop (phil)simulations with a real solar inverter,” in IECON 2017
- 43rd Annual Conference of the IEEE Industrial Electronics Society.
IEEE, 2017, pp. 2701–2706.
[28] S. Vogel, H. Thi Nguyen, M. Stevic, T. V. Jensen, K. Heussen,
V. Subramaniam Rajkumar, and A. Monti, “Distributed power hardware-
in-the-loop testing using a grid-forming converter as power interface,”
Energies (Basel), vol. 13, no. 15, pp. 3770–, 2020.
[29] K. Sidwall and P. Forsyth, “Advancements in real-time simulation for
the validation of grid modernization technologies,” Energies (Basel),
vol. 13, no. 15, pp. 4036–, 2020.
[30] F. Lehfuss, G. Lauss, P. Kotsampopoulos, N. Hatziargyriou, P. Crolla,
and A. Roscoe, “Comparison of multiple power amplification types
for power hardware-in-the-loop applications,” in 2012 Complexity in
Engineering (COMPENG). Proceedings. IEEE, 2012, pp. 1–6.
[31] M. Panwar, B. Lundstrom, J. Langston, S. Suryanarayanan, and
S. Chakraborty, “An overview of real time hardware-in-the-loop ca-
pabilities in digital simulation for electric microgrids,” in 2013 North
American Power Symposium (NAPS). IEEE, 2013, pp. 1–6.
[32] J. Ihrens, S. M¨
ows, L. Wilkening, T. A. Kern, and C. Becker, “The
impact of time delays for power hardware-in-the-loop investigations,”
Energies (Basel), vol. 14, no. 11, pp. 3154–, 2021.
[33] E. Guillo-Sansano, M. H. Syed, A. J. Roscoe, G. M. Burt, and F. Coffele,
“Characterization of time delay in power hardware in the loop setups,”
IEEE transactions on industrial electronics (1982), vol. 68, no. 3, pp.
2703–2713, 2021.
[34] A. Viehweider, G. Lauss, and L. Felix, “Stabilization of power hardware-
in-the-loop simulations of electric energy systems,” Simulation mod-
elling practice and theory, vol. 19, no. 7, pp. 1699–1708, 2011.
[35] Spitzenberger & Spies, “Instruction manual - voltage source type aps
5000,” Tech. Rep., 2018.
[36] Australian Energy Regulator (AER), “Residential standing offer,
agl14980sre9,” 2021.
[37] Ausgrid, “Ns220 overhead design manual, nw000-s0092,” Ausgrid,
Tech. Rep., 2021.
[38] NSW Energy, “Service and installation rules of new south wales,”
Climate Change and Sustainability Division and NSW Department of
Planning, Industry and Environment, Tech. Rep., 2019.
[39] Ausgrid, “Solar home electricity data,” https://www.ausgrid.com.
au/Industry/Our-Research/Data-to-share/Solar-home- electricity-data,
2013.
[40] S. Svensson, “Power measurement techniques for non-sinusoidal condi-
tions,” Ph.D. dissertation, Chalmers University of Technology, G¨
oteberg,
Sweden, 1999.