Fast Demand Response as an Enabling Technology
for High Renewable Energy Penetration in Isolated
Power Systems
D. Nikolic M. Negnevitsky M. de Groot S. Gamble
Hydro Tasmania University of CSIRO Hydro Tasmania
Hobart, Australia Tasmania Sydney, Australia Hobart, Australia Hobart, Australia Hobart, Australia
dusan.nikolic@ Hobart, Australia
entura.com.au
AbstractIsolated power systems (IPSs) up to 20MW capacity
are often based on diesel generators. Due to the high cost of
diesel fuel supply, as well as environmental concerns, IPSs
should ideally include a large proportion of renewable energy
generation. We report on the implementation of a simple, fine-
grained and fast demand response (DR) technology to enable
high wind penetration in an IPS. This solution is based on
centralized two-way communication and control of residential
and commercial loads. DR can be dispatched and confirmed
within 1 second. The technology has been installed and
successfully tested in an IPS on an island in Australia.
Index TermsIsolated Power Systems, Smart Grids, Demand
Response.
I. INTRODUCTION
Electricity consumers living in remote areas or on islands
often cannot be supplied from conventional interconnected
power systems. These consumers are usually serviced by a
local electricity generation and distribution system which we
will refer to as an isolated power system (IPS). Electricity in
IPSs is traditionally generated using diesel fuel. Due to
remoteness and consequent high cost of diesel fuel supply, the
cost of electric energy in IPSs is high compared to
conventional interconnected systems. In some locations the
price might exceed US $1/kWh which is an obvious incentive
for introducing renewable energy (RE) generation.
Unfortunately, RE from the two most abundant energy sources
wind and solar incurs significant stability and reliability
issues due to the intermittency of those sources. This issue has
been widely discussed in the research and industry literature
[1].
Many existing conventional and isolated power systems
have already integrated RE. Because a small investment can
have a relatively large impact, the RE penetration in some
IPSs has reached high levels (i.e. greater than 30%).
Consequently, IPSs are at the forefront of high RE power
systems research. One of key lessons this research has
provided is that RE sources need complementary, or enabling,
technologies to deal the attendant issues of highly variable
power output and low inertia. Examples of complementary
technologies include: energy storage (e.g. battery[2],
hydrogen, flywheels[3]), new types of conventional generators
(e.g. bio diesel, low load diesel engines) and, demand control
J. Forbes M. Ross
CSIRO Hydro Tasmania
(e.g. smart grid services). Most researchers see battery or
hydrogen storage as the appropriate complementary
technology for high renewable penetration in IPSs. However,
these technologies are complex and costly. They also have
integration challenges and low inertial response. Some
biodiesel fuels still rely on mineral diesel to some extent (e.g.
5%, 20% biodiesel blends), and while they often reduce the
ecological footprint, they do not necessarily reduce operating
costs. Low load diesel engines have proven that they can
support high levels of RE in some applications, however this
technology is still young and all impacts on operational and
maintenance costs have not yet been fully assessed [4].
This paper presents fast (i.e. sub-second) demand response
(DR) as an RE enabling technology in IPSs. The discussion
has the following structure. The advantages of DR in IPSs are
presented in Section II. Our concept and specification of a fast
DR system is developed in Section III. A case study of the
first implementation of the fast DR system, along with some
preliminary results, forms Section IV. Section V concludes the
paper.
II. DEMAND RESPONSE IN ISOLATED POWER SYSTEMS
All power systems need to constantly balance electricity
demand and supply. Demand management (DM) in addition
to just managing supply - is beneficial for existing power
systems through:
Providing support for frequency and/or voltage
regulation[5], [6].
Reducing operational costs and emissions by increasing
utilization of RE sources[7]. Note that reducing operating
costs in turn leads to a greater return on investment
(ROI) which would incentivize expansion of the RE
industry.
Reducing operational costs and emissions caused by
traditional generators installed to provide spinning
reserve for RE [8].
Relieving stress from transmission and distribution
infrastructure by coordinating loads close to RE
sources[8].
Reducing utility operating costs through advanced
metering infrastructure (AMI) installed to enable DM
[9].
 Most DR research focuses exclusively on large power
systems which have a steady and predictable load profile
defined by morning and evening peaks. The time and size of
these peak periods can be fairly accurately estimated using
historical load data and weather forecasts. A typical daily load
profile for a large power system is presented in Figure 1.
Figure 1. Profile of typical system daily load in larger system. Load as seen
by conventional generators is slowly changing during the day.
A less volatile and predictable demand curve means that it
is much easier to schedule conventional generation (e.g. coal
or nuclear). While there is still some cost and operational
benefit to be gained by peak shaving in large power systems,
demand peaks only occasionally threaten supply security.
Furthermore, conventional power systems can often
implement financial DM measures to reduce peaks such as
critical peak pricing and time-of-day tariffs[10].
In contrast to conventional grids, IPSs not only supply less
power (usually MWs, not GWs) they are also geographically
much smaller. Being smaller in capacity means that demand is
less predictable. Being smaller in area means that supply from
RE sources is more variable, as a larger percentage of the RE
generators are likely to be affected by the same weather events
(e.g. a lull in the wind or passing clouds). Due to their reduced
demand predictability and increased variability of RE supply,
conventional generation scheduling in IPSs with RE is more
challenging. From a generation scheduling perspective - where
RE generation is usually treated as a load offset the daily
load curve becomes extremely volatile. An example IPS with
RE load curve is giving in Fig 2. Note that it does not even
have predetermined daily peaks.
High load variation makes scheduling of diesel generation
more difficult and less efficient, as diesel engines will rarely
operate at their peak efficiency, and more generator start-ups
are required. This is where fast DR can help. It can smooth the
variability of required diesel generation by quickly adjusting
system load. To be able to do that, fast DR cannot rely on
typical load patterns - it must be based on the current, or near-
real time, system load status. If DR is to complement RE in an
IPS, it has to be as fast as the speed at which RE generators
change their power output.
Figure 2. Daily load diagram in IPS as seen by conventional diesel
generators is highly variable, and there are no predetermined peaks or dips.
Data for this example is taken from our case study system, described in
Section IV.
III. DEMAND RESPONSE AS A VIRTUAL POWER PLANT
The idea of aggregating and controlling small loads to
create a large block of variable demand has been discussed in
the power engineering literature where it is often referred to
as a virtual power plant[11]. Since isolated power systems
generally have only one power station, the idea of adding
another (virtual) power plant to this system might be a bit
misleading. However, high RE IPSs might have several
generating sources (e.g. diesel, wind, etc). Therefore,
aggregated DR in IPS could be more usefully treated as a
virtual generator. Although such a DR Generator (DRG) is
not generating real power, the Power System Controller
(PSC) perceives it as one due to the DRGs ability to decrease
the amount of energy needed from other generating sources.
IPSs are usually automatically controlled power stations
with one central PSC. The PSC is typically implemented with
a programmable logic controller (PLC). The role of the PSC
is to schedule available generation in accordance with the
current power system constraints and to maintain system
stability. In addition, a PSC may be programmed to maximize
the amount of RE generation and consequently, minimize
running costs. The PSC effectively controls the entire IPS by
collecting data on the current system status and by issuing
commands to various generation sources (Figure 3).
When the PSC has a goal of maximizing RE, it will
dispatch as much renewable generation as the power system
can handle while simultaneously maintaining an appropriate
level of spinning reserve to ensure system stability. If the
amount of RE drops, and the system suddenly does not have
enough spinning reserve, the PSC will start a diesel generator.
The role of the DRG is now clear it has to support high RE
penetration by providing additional spinning reserve. If
sufficient spinning reserve is provided by the DRG, diesel C. Slave controllers
generator startup is prevented. Slave controllers are located in each DR capacity
providing site. They consist of a pair of gateways to provide
WiMAX-Ethernet-ZigBee signal translation between the
backhaul network and the individual load control devices. The
load control devices perform both metering and load
switching. They provide a range of power metrics and also
support setpoints.
A DRG providing spinning reserve must be extremely
responsive and reliable. The master controller must be able to
monitor and dispatch slave controllers at all times. This
critical requirement becomes obvious in the two most
common scenarios:
If the PSC requests DR for extended periods of time,
Figure 3. Smart Grid generator in IPS control system
some slave controllers may override the dispatch as they
A DRG consists of three main components, which are exceed their maximum dispatch duration. In this situation
shown in the Fig. 4: the master controller must quickly identify and dispatch
DRG Master controller, another device (or devices) with an equivalent load.
Communications network, and If a slave controller or communication link is unreliable,
Slave controllers. the DRG may be forced to always dispatch more DR
Residential Ethernet TCP/IP than requested to ensure a suitable margin of error in
WiMAX either load switching or reporting. This is not an efficient
Communications ZigBee
use of capacity and may reduce the overall effectiveness
network
Power System of the DRG.
Commercial Controller (PSC)
IV. CASE STUDY
The fast DR technology discussed above was
Residential implemented in an IPS as part of the Hydro Tasmanias King
Master
Controller
Island Renewable Energy Integration Project (KIREIP) [12].
Slave King Island (KI) lies in the Bass Strait between Tasmania and
Demand Response
Controller
Generator
the Australian mainland. It has a population of approximately
2000 people, and an economy based on agriculture and food
Figure 4. Demand Response Generator architecture
processing.

A. DRG Master Controller A. The King Island power system

The master controller collects information on available Customer load on KI ranges between 1 MW and 3 MW,
DR and aggregates it into predefined virtual loads (e.g. with an average of around 1.5 MW. The KI power system is
geographical regions). It constantly communicates available shown in Figure 5.
DR capacity to the PSC while checking for DR dispatch
requests from the PSC. DR requests identify a target virtual
load and the amount of demand to curtail. The master
controller selects the individual loads to curtail from the
virtual load, and immediately sends a switch-off signal to
each.
B. Communications network
A multi-protocol bidirectional communications network
delivers information between all elements of the DRG.
Ethernet is used within the control system. A dedicated
WiMAX network provides the backhaul capability. Within
individual customer sites, a WiMAX gateway is connected by Figure 5. Schematic representation of the King Island power system
Ethernet to a ZigBee gateway for the final link to the load
metering and switching devices. This communications There is one power station on the island with four
configuration is configured to ensure a sub-second round trip distribution feeders delivering electricity to customers. The
for DR requests from the PSC out to the load control devices power station houses four diesel generators with a total
generation capacity of 5.8MW. Three fixed speed Nordex
and back again.
N29 (250kW each) wind turbines are installed on a nearby
hill, together with two Vestas V52 turbines (850kW each)
with doubly fed induction generators. Two 800kW diesel
engines with flywheels are also connected to the system. In
these generators the flywheels are separated by a clutch from
a diesel engine and provide system with additional inertia.
This system is currently capably of operation with 100%
sustained RE penetration under suitable wind conditions.
B. King Island Smart Grid project
The ongoing King Island Smart Grid (KISG) project is
part of the larger KIREIP mentioned above. It has the goal of
supporting higher levels of wind energy integration on KI by
providing:
1. Spinning reserve by implementing the DRG concept
described in Section III, and
2. Immediate fine-grained under frequency load shedding
(based on the slave controller level).
Various major components of the KI DRG architecture
outlined in Fig.4 can be seen in the installation photos in
Figures 6-8.
Figure 6 (left). King Island DRG Master Controller is installed on a server
computer located in the power station control room and directly connected to
power station controller.
Figure 7 (right). Standard WiMAX network antennas were used for the
communications backbone. Photo shows equipment installed on a standard
15m distribution pole.
Figure 8. WiMAX modem is installed at the customer premises (box on the
right hand side), together with a slave controller which sends signals to smart
switches inside switch box on the left.
A range of loads were integrated into KISG including:
water and space heaters, air conditioning units, commercial
freezers and electric car chargers. The master controller
constantly monitors the available connected DRG load and
passes this information to the PSC - which is a fully
automated PLC based control system. At the same time, the
PSC constantly monitors the current level of power system
spinning reserve. If the spinning reserve falls under a
predefined threshold, the PSC instructs the DRG to curtail
some load and effectively raise the spinning reserve. This
behavior can be observed in Figure 9, where due to a sudden
drop in wind generation, the spinning reserve suddenly drops.
The reduction in spinning reserve caused the PSC to ask for
all available DR as shown in the lower graph.
Figure 9. Early King Island DRG operation results
The testing results presented in Figure 9 show that the
implemented DRG was able to accurately respond to given
setpoints. The figure also shows that DR capacity is being
dispatched reliably in 1 second. In fact, not only is the load
being switched, but confirmation of the switching is being
reported back to the PSC with sub-second response.
C. Preliminary results
Prior to roll-out on KI, the DRG was performance tested
with 10,000 simulated customer loads with minimal
performance degradation. At the time of writing this paper,
the KI DRG had only 50 sites under management and had
been fully commissioned into the power system. When
complete, the DRG will be extended to include more than 150
households and several larger commercial loads. This implies
that the full DRG will supply more than 100 kW of sub-
second DR capacity, which we believe would be sufficient to
have a material impact on generation scheduling.
The effectiveness of the KI DRG depends largely on its
integration with the PSC. Fig. 10 shows how the PSC calls
the DRG frequently and effectively uses it as a tool for
precise regulation of demand. During the period of over 2
hours, the KI power system was running in zero-diesel
operation and the DRG was called for support during small
and short dips of wind power generation. In this mode of
operation, the PSC prioritizes DRG dispatch and potentially
postpones the startup of a diesel generator.
Figure 10. Automated operation of King Island DRG
D. Engaging with community
Throughout the KISG project, Hydro Tasmania consulted
the local community on how the smart grid could benefit both
electricity supply in general and customers needs
specifically. Customers were keen to participate in a Smart
Grid project which would promote King Island as a
renewable energy community. Customers volunteered to
participate in the project and were offered small incentives
such as a single monetary payment, and energy audits of their
premises. DRG metering data for each individual customer is
available through a smart phone application which allows
customers to monitor their entire load in near real-time.
The key lessons from the community engagement process
were:
Customers are generally happy to participate in projects
that clearly contribute to community wellbeing,
Small and diverse incentives help attract a larger number
of customers, and
As long as agreed operating constraints are respected,
customers will follow the set and forget principle for
their smart grid participation.
V. CONCLUSION
IPSs with high levels of RE need enabling technologies to
achieve significant operational cost savings without
sacrificing power system stability. Fast DR could be an
appropriate enabling technology.
In this paper we presented a DRG which, on the basis of
preliminary results, can be used to reliably provide spinning
reserve in IPSs. We have also supplied some system
measurements showing that the requisite sub-second response
time can be achieved using relatively cheap components.
The implementation results clearly demonstrate that the
fast DR approach is an enabling technology for high RE
penetration in IPSs.
ACKNOWLEDGMENT
The authors would like to thank Hydro Tasmania for
support, technical guidance and provision of data. They also
acknowledge the support of the CSIRO Energy Transformed
Flagship.
REFERENCES
[1] M. Piekutowski, S. Gamble, and R. Willems, "A Road towards
Autonomous Renewable Energy Supply, RAPS case," in CIGRE 2012,
Paris, France, 2012.
[2] D. Kottick, M. Blau, and D. Edelstein, "Battery energy storage for
frequency regulation in an island power system," Energy Conversion,
IEEE Transactions on, vol. 8, pp. 455-459, 1993.
[3] N. Hamsic, A. Schmelter, A. Mohd, E. Ortjohann, E. Schultze, A.
Tuckey, and J. Zimmermann, "Increasing Renewable Energy
Penetration in Isolated Grids Using a Flywheel Energy Storage
System," in Power Engineering, Energy and Electrical Drives, 2007.
POWERENG 2007. International Conference on, 2007, pp. 195-200.
[4] ABB-PowerCorp. (2012). Low-Load Diesel (LLD) product. Available:
http://www.pcorp.com.au
[5] J. A. Short, D. G. Infield, and L. L. Freris, "Stabilization of Grid
Frequency Through Dynamic Demand Control," Power Systems, IEEE
Transactions on, vol. 22, pp. 1284-1293, 2007.
[6] T. L. Vandoorn, B. Renders, L. Degroote, B. Meersman, and L.
Vandevelde, "Active Load Control in Islanded Microgrids Based on the
Grid Voltage," Smart Grid, IEEE Transactions on, vol. 2, pp. 139-151,
2011.
[7] M. de Groot, D. Nikolic, and J. Forbes, "Demand Response in Isolated
Power Systems," in Australasian Universities Power Engineering
Conference (AUPEC), Hobart, Australia, 2013.
[8] D. Westermann and A. John, "Demand Matching Wind Power
Generation With Wide-Area Measurement and Demand-Side
Management," Energy Conversion, IEEE Transactions on, vol. 22, pp.
145-149, 2007.
[9] N. Rajakovic, D. Nikolic, and J. Vujasinovic, "Cost benefit analysis for
implementation of a system for remote control and automatic meter
reading," in PowerTech, 2009 IEEE Bucharest, 2009, pp. 1-6.
[10] H. Saele and O. S. Grande, "Demand Response From Household
Customers: Experiences From a Pilot Study in Norway," Smart Grid,
IEEE Transactions on, vol. 2, pp. 102-109, 2011.
[11] J. Kumagai, "Virtual power plants, real power," Spectrum, IEEE, vol.
49, pp. 13-14, 2012.
[12] Hydro Tasmania. (2013). King Island Renewable Energy Integration
Project (KIREIP). Available: www.kireip.com.au