This study into the costs and economics of concentrating solar power in Australia formed the major part of a one-year investigation towards completion of Warwick Johnston’s Master in Science (Renewable Energy). This dissertation analyses the costs of thermal trough CSP and its revenue from connection to Australia’s National Electricity Market (NEM). As the wholesale price of electricity typically peaks well after midday, the investigation assesses the optimum amount of energy storage in order to maximise Internal Rate of Return (IRR).
The following document is a synopsis of findings – scroll down to read the full disseration:
The following document is the full text:
Economics of Storage in a NEMConnected
Concentrating Solar Power
Author: Warwick Johnston – Manager, SunWiz – BEng (Hons)/BSc
Extract of Dissertation for Masters of Science (Renewable Energy)
Please contact: email@example.com or 0413361534 for further information
This is a summary of a dissertation undertaken by Warwick Johnston into economically-optimum
levels of energy storage in a Concentrating Solar Power Station (CSPS) connected to the National
Electricity Market (NEM).
It answers the questions:
In the context of operation within the Australian wholesale electricity market, is there value
in using energy storage in a solar power station? Does this vary by site, dependent on solar
radiation characteristics and wholesale price fluctuation?
What amount of energy storage generates the greatest revenue from a solar power station?
What is the most cost-effective investment in energy storage?
Low NEM prices mean IRRs from CSP are unlikely to be sufficiently high to attract
investment, even with 50% government funding of upfront costs.
Incorporating storage leads to higher LCOE but improves IRR, even when price-insensitive
energy dispatch methodologies are employed.
Use of price-sensitive energy dispatch methodology can lead to 10-20% greater revenue
Generation during peak power prices can increase annual revenue by 25%.
Use of measured solar radiation data is critical, as differences between measured and
Typical Mean Year solar data can lead to revenue miscalculations
Higher NSW NEM prices can lead to more favourable investments than those in sunnier
Queensland, but NEM price variability can easily cause this situation to reverse.
The NEM electricity price tends to peak in the late afternoon, long after the peak in solar radiation,
as shown in the following graph. A CSPS that was able to delay its output to coincide with this price
peak should be able to generate more revenue.
However, loss of energy occurs whenever storage is used; and although thermal storage may be less
expensive than battery storage, its costs must still be covered by increased revenue. Thus there may
be an optimal level of storage, in which revenue per unit of storage is maximised; an optimal level
that is dependent upon the cost of storage and the time-value of electricity.
The graph above suggests that storage can reduce Levelised Cost of Energy (LCOE). LCOE is an
appropriate measure by which to financially compare CSPSs if revenue is fixed by a Feed-in Tariff, as
producing lower cost energy should maximise return if revenue is only based upon output energy.
However, when the electricity price varies with time, as occurs in the NEM, optimal configurations
are those that maximise revenue per invested dollar (return on investment).
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0
NSW Ave NEM price & Solar Radiation vs
Time of Day
Direct Solar Radiation Global Solar Radiation
This leads to the questions:
What amount of energy storage maximises Internal Rate of Return (IRR)?
Does this amount vary by location?
How do NEM price and solar radiation characteristics influence this outcome?
The challenge for CSP developers is that the NEM price varies by state, year, month, day, hour, and
minute. The annual variation between states is shown in the graph below; the subsequent graph
shows monthly variation for NSW in 2004.
These graphs show that a tendency toward afternoon price peaks exists, but the price profile
changes significantly as the profile is only an average of a half-hourly varying price. As the annual
revenue is the sum of the instantaneous product of NEM price and CSPS output, dispatching energy
2004 2005 2006 2007
Average NEM Price 2004-07
1 3 5 7 9 11 13 15 17 19 21 23
Hour of Day
Average Electricity Price vs hour of day –
2004 NSW January
during varying times of peak energy price is critical for revenue maximisation. CSPS design should
therefore incorporate a sufficient amount of storage to cost-effectively capture higher electricity
Solar Analysis Model (SAM) CSP simulation software produced by the National Renewable Energy
Laboratory (NREL) was used to model the input and operational costs of various CSP configurations.
A financial spreadsheet that incorporates NEM electricity pricing data was built around the SAM
simulator output. This allows calculation of LCOE and IRR for a number of input variables across a
variety of locations and NEM electricity price datasets.
The modelling took a selected year’s hourly NEM price and assumed it applied in the following thirty
years of operation. To provide a sensitivity analysis to NEM price, each simulation was run four
times, once for each of the four chosen years of hourly NEM price data (2004-2007). This also allows
investigation of the relationship between financially optimum levels of storage and the NEM price
An investigation of the variance of IRR with respect to inputs such as solar multiple (SM; the solar
field capacity divided by the generator output capacity), hours of thermal energy storage (hTES),
NEM price, and location was performed. To model the CSPS initially proposed under the solar
flagships program, a 250 MW CSPS was chosen. LCOE was investigated for the purpose of
comparison to international literature, and was not subsidised in order to enable direct comparison.
A currency exchange rate of 1USD=0.8AUD was used.
It was found that a 33% government contribution to initial capital cost (as proposed by the
announced Solar Flagships Program) did not guarantee good outcomes, and negative IRRs were
found in certain years. Instead a 50% government contribution was modelled for IRR calculations, so
that comparisons between from multiple datasets with positive IRR could be drawn. The sunniest
NEM-connected location in Australia (Longreach, QLD) was chosen as the base case site of the power
station, with a sensitivity analysis performed upon location.
The Importance of Weather Data
The study found publically-available half-hourly DNI solar radiation data was only available at six
NEM connected sites, as shown in the table below. In contrast, Typical-Mean Year half-hourly data is
available at a large number of sites.
Frequency Data Source Locations of interest to study
Bureau of Meteorology
SA: Adelaide, Mt Gambier
NSW: Wagga Wagga
Vic: Melbourne, Mildura
Typical Mean Year Australian Greenhouse
Office, via EnergyPlus
NSW: Armidale, Coffs Harbour, Dubbo,
Mascot, Moree, Nowra, Orange, Richmond,
Sydney, Thredbo, Wagga Wagga,
QLD: Gladstone, Longreach, Mackay, Mt Isa,
Oakey, Rockhampton, Townsville
SA: Adelaide, Ceduna, Mt Gambier, Mt Lofty,
Vic: Ballarat, Cape Otway, East Sale,
Melbourne, Mildura, Moorabbin,
Even though a profile for solar radiation and NEM price shows some inter-relationship on average,
there is no strong correlation between the two at the half-hourly and daily level (seen below). In
order to investigate multiple sites at the sunniest locations, TMY data was used for this study, with a
sensitivity analysis performed that compares use of measured and TMY data.
Correlation between 2005 Daily Radiation at Rockhampton and QLD NEM Price – Half hourly measured data used.
Base Case: Longreach, Queensland, 2007 NEM Dataset, $350/m2 solar field,
$40/kWhth storage, $50/REC
The graph below shows that the LCOE from a 250 MW CSPS in Australia is greater than A$0.20/kWh.
It has a minimum for a Solar Multiple of 1.66 with 2 hours of thermal storage, although comparable
LCOEs can be achieved with a SM of 2 and 4 or 6 hTES.
Although the LCOE may be least for a SM of 2 and 6 hTES, the IRR for the 2007 QLD NEM price data
set is greatest for a SM of 2.4 and 6 hTES. Even the significantly higher LCOE that results from
increasing the SM to 3 (with 6 hTES) still results in a comparably favourable IRR.
Sensitivity Analysis: NEM Dataset
The following section investigates the IRR that can be achieved with varying NEM-price datasets. The
figure below shows the average hourly profile for each of the annual datasets, and gives insights into
why particular configurations may be more favourable in some years than others.
Average Queensland NEM Price Hourly Profile for 2004-2007
In summary, the configuration that produces greatest IRR varies depending on the NEM price
dataset used. Extremely poor IRRs are obtained for 2005 and 2006 NEM price datasets. 4-6 hours of
storage generally produces best results.
IRR for a 250 MW CSPS in Longreach, 1USD=0.8AUD, and 50% government contribution. 2006 QLD NEM dataset
1 3 5 7 9 11 13 15 17 19 21 23
Hour of Day
Average Qld NEM Price Hourly Profile
IRR for a 250 MW CSPS in Longreach, 1USD=0.8AUD, and 50% government contribution. 2005 QLD NEM dataset
IRR for a 250 MW CSPS in Longreach, 1USD=0.8AUD, and 50% government contribution. 2004 QLD NEM dataset
Sensitivity to Location
As shown below, Longreach has the highest average DNI solar radiation resource in NEM connected
locations with available TMY data. Moree has 12% less DNI than Longreach, Woomera 6%, and
Mildura 16%. However, the subsequent graph demonstrates that the electricity power price in other
states was significantly higher than that in Queensland in some years – for example, in 2004 NSW
had an average price that was 35% higher than Queensland’s; SA’s average NEM price was 22%
higher than Queensland. Thus, a sensitivity analysis was performed across the sunniest locations in
Annual Direct Normal Incidence Solar Radiation
2004 2005 2006 2007
Average NEM Price 2004-07
The graph below presents the LCOE and the IRR for the locations of Moree (NSW), Woomera (SA),
Mildura (Vic) and Longreach (Qld), using the base case inputs and a 2007 NEM price dataset and the
best plant configuration identified at Longreach. The following graph repeats these values for a 2004
NEM dataset. Note that just as optimal plant configuration in Longreach varied depending on which
year’s NEM price dataset was used, the investigated plant configuration may be not be optimal for
each of these locations.
LCOE and IRR for best locations in each state, 250MW CSPS, $50 REC, SM=2.4, 6hTES, 10% discount rate. 2007 NEM data
LCOE and IRR for best locations in each state, 250MW CSPS, $50 REC, SM=2.4, 6hTES, 10% discount rate. 2004 NEM data
Naturally, the LCOE remains independent of the price of the delivered power. However, Longreach
obtains the best IRR under 2007 NEM price scenarios, though Moree would have eclipsed Longreach
in 2004. Although Woomera receives only 6% less DNI than Longreach but had a 22% higher average
price in 2004, Longreach’s 2004 IRR was better than Woomera’s. This may highlight that it is not the
average power price that is important, but the alignment of prices with the ability to deliver stored
Variation with Solar Radiation Data
As no ground-based hourly solar radiation data measurements are available at the sunniest NEMconnected
location (Longreach), Typical Mean Year (TMY) solar radiation data was used. Although
there is no strong correlation between measured solar radiation data and NEM price, a sensitivity
analysis was performed in order to determine the scale of impact of using TMY rather than
measured solar radiation data. To do so, a location was selected for which both data was available,
that of Wagga Wagga (NSW). Rather than re-run the entire simulation, the sensitivity analysis
mimicked output from a storage-less CSP by summing the instantaneous multiplication of solar
radiation data with NEM price.
The graphs below depicts a situation on February 15, 2004. On this day the NEM price spiked (red
line of upper graph) whilst the true measured solar radiation (orange line of upper graph) was far
higher than the TMY (blue line of upper graph). This would have resulted in greater revenue from a
true CSP (brown line of lower graph) than the simulation predicted (yellow line of lower graph).
Solar Radiation, NEM price, and their product for February 14, 2004
Overall, 2004 measured solar radiation at Wagga Wagga were 2.6% higher than TMY radiation for
that year. However, use of measured data would have resulted in 24% less revenue for a storageless
CSP than the TMY based simulation, chiefly due to missed opportunities (low radiation at time of
peak power prices) – as shown in the following graph. In 2005 the situation changed so that a 3.8%
greater measured DNI occurred than TMY, resulting in a 9.6% greater revenue than would have been
Influence of Peak Power Price Capitalisation on Revenue
The impact of peak NEM prices upon revenue is shown in the figure below, switching back to
Longreach with base case configuration. The slopes of the 2004, 2005, and 2006 cumulative
revenues are almost the same, but peak price spikes increase revenue from $30m to $50m, or 66%.
The slope of the 2007 cumulative revenue is higher, indicating a generally higher electricity price,
but the revenue is still strongly influenced by peak power prices. Given the infrequent occasion on
which these price spikes occur, and the uncertainty with which they coincide with solar radiation, it
could be worthwhile to remove such peak power prices from the analysis and consider peak power
price events as windfall profit.
Figure 1: Cumulative Revenue from SM2.4 6hTES Longreach CSPS with TMY Solar Radiation for Various NEM Years
Whilst this analysis does not examine the benefits that storage can bring by delaying solar output, it
does show that variability between the NEM price, the solar radiation resource, and their product
can have significant impact upon revenue. Storage would not have changed the peak power price
situation that occurred in 2004 (depicted below): there was little radiation in Wagga Wagga on that
day to store when the power price exceeded $9000/MWh. Ultimately, this clearly demonstrates the
need for reliable (“bankable”) hourly direct measurements of DNI in order to convince investors so
that the project may proceed, but that an unavoidable risk of missing peak revenue remains.
December 2, 2004 Peak Power Price
Variation with energy dispatch methodology
Storing energy brings with it the ability to dispatch the energy at a chosen later time, and the
opportunity to maximise revenue by scheduling power delivery based upon NEM price. The SAM
modelling software dispatches energy based upon stored and incident solar radiation, with only
limited ability to dispatch based upon time of day, week, and month. In order to maximise profits, a
CSPS could base its energy dispatch methodology upon the forecast power price and the forecast
weather. Such a dispatch methodology may offer significantly increased revenues when compared
to operating the power plant in the fixed manner which produced the aforementioned results.
The benefit of storage to revenue is shown in the figure below, which demonstrate the energy
received by the solar collector field, the part of which is sent to TES, drawn from TES, and sent to the
power block (both directly and from the TES), with the corresponding net output of energy for a
particular 24 hours period. On the right axis, the figures also show the NEM price and the amount of
revenue earned by that power delivery at that time.
The graph shows a day with a limited amount of sunshine. On this day the power price was higher in
the early morning than during the middle of the day, and it peaked in the early evening. However,
SAM’s scheduled power delivery algorithm stored the energy during the morning price peak, and
delivered it when the price was lower, also missing out on the evening’s price peak.
Energy Flows and Revenue – Aug 16 2007
In order to investigate the amount of additional revenue that may be gained by a more priceresponsive
scheduling algorithm, the output power from a storage-less CSP has been dot multiplied
with the NEM price delayed by a nominated amount. The bars in the graph below indicate the
amount of revenue that would be created from delaying the entire year’s production by zero to six
hours. The seventh bar mimics a price-sensitive dispatch mechanism by summing each day’s
maximum amount of revenue from zero to six hours of delay. The line graph depicts the number of
days in 2004 for which zero, one… to six hours of delay produced the greatest revenue.
Revenue Maximising through Strategic Dispatch of Stored Solar Energy – 2004 QLD Longreach
The graph above demonstrates that on most days of the year, six hours delay produced greatest
revenue. However, simply bypassing storage on 90 days would have produced the greatest possible
revenue for each of those days. Naturally, this is a simplistic model that does not account for storage
losses or the financial costs of storage. However, it does show that the resultant revenue could be
up to 15% greater than was calculated by SAM’s price-insensitive dispatch mechanism,
demonstrating that incorporating storage could provide additional benefits to those stated.
This dissertation has demonstrated that the IRR achievable from a NEM-connected CSPS is too low
to lead to significant development of CSPS around Australia without far greater government support
than has been offered. Contributions of 50% of the initial capital cost are needed to obtain
reasonable IRR in the sunniest location for half of the years studied; the Solar Flagships program’s
proposed 33% government contribution is insufficient in the light of these results. The $1.1b-$3.1b
required to build a 250 MW CSPS would quickly exhaust the government’s earmarked funding of
$1.5b. Therefore, greater support is needed if the government is to achieve its policy objectives.
Although the input costs are based upon sound methodology (described in Appendix 2), improved
financial outcomes should come through expected 40% learning-curve LCOE price reductions.
Characteristics of the NEM price also contribute strongly to poor financial outcomes. The peak
power prices of $10,000/MWh occur too infrequently to benefit CSPS. The variability of the NEM
price from year to year creates strong risk when choosing CSPS location, and the variability in the
NEM price daily profile inhibits the optimisation of energy storage value.
Indeed, there seems to be little value in incorporating energy storage into NEM-connected CSPS,
unless storage costs drop significantly. The LCOE is least for storage-less configurations at the
assumed storage cost of $40/kWhth. Although storage can lead to a 1% increase in IRR, such benefits
may be outweighed by an increase in perceived risk. Countering this is the ability to dispatch power
based upon NEM price, which leads to flexible revenue maximisation opportunities now and in
The optimum location for a NEM-connected CSPS is not necessarily the location with greatest directnormal
solar radiation. The variation between states in NEM power price and daily profile can have
dramatic impact upon achievable IRRs.
It is clearly apparent that accurate, “bankable” solar radiation data is required in order to accurately
assess likely project returns. As only a few solar radiation measurement sites exist that are NEM
connected (of which only a couple are very sunny), government investment in solar resource
measurement may greatly facilitate the deployment of CSP in Australia.
Sensitivity analyses to costs of storage and collectors, REC price, government funding contribution,
and electricity price increase were also performed, and can be read in the complete dissertation.
Appendix 1: Input Data
The following inputs and configuration was used for SAM:
Item Setting Parameters
System Degradation 0%
Heat Transfer Fluid Hitec XL Other Parameters Default
Solar Collector SolarGenix (as used in Nevada Solar
Power Block Rated Turbine
Power Block Design Turbine
Power Cycle As per Library: SEGS 80MWe
Default, Wet-bulb Temperature
Two-tank Only available Option
Thermal Storage Fluid Type Hitec XL
Thermal Storage Dispatch
SCE (A Californian utility pricing
structure included in SAM)
Thermal Energy Storage
Linked to hTES as per SAM user
guide Table 241
Parasitics SEGS VIII Reference Default
The heat transfer fluid and the energy storage fluid were chosen to be Hitec XL, in order to remove
the cost and inefficiency of heat exchangers.
SAM Financial inputs
Item Setting Remarks
Analysis Period 30 years
Inflation Rate 2.5%
Real Discount Rate 10%
Federal Tax (Business Tax
State Tax 0% Land taxes etc covered elsewhere
Sales Tax 0%
Insurance 0.5% Default value
Depreciation 6.66% Custom depreciation values used to reflect
standard Australian depreciation curve
Tax Credit Incentives None 33%/50% government contribution only factors in
Payment Incentives A$0.05/kWh Reflective of REC price of $50, taxable income
Site Improvements US$20/m2 Default value
Solar Field US$350/m2 Default value
HTF System US$50/m2 Default value
Storage US$40/kWhth Non default value reflective of literature
Power Plant US$880/kWe Default value
Electricity Price above 0% Sensitivity analysis performed
Indirect Costs: EPC 15% Default value
Indirect Costs: Product, Land,
3.5% Default value
O&M: Fixed Annual Costs $0/year Default value
O&M: Fixed Cost by Capacity $80/kW/year Default value
O&M: Variable Cost by
$3/MWh Default value
See the appendices for a more detailed description of the default costs used in SAM, which are
based on quotations and a study commissioned by NREL and undertaken by expert consultants2.
Other Financial Parameters
The economic model used the following assumptions:
The upfront cost is placed entirely in year zero. Any government contribution also occurs in
year zero and is not taxed.
Annual Revenue is the sum of electricity generation revenue – the constant annual net
electricity output multiplied by reference year NEM price dataset multiplied by (inflation
plus electricity increase) – plus RECs Revenue – constant REC price multiplied by net
electricity output. Inflation is not applied to the REC price in order to reflect expected REC
price decrease over life of system.
The annual expenses comprise of insurance, plus Operation and Maintenance.
The annual after-tax cash-flow is Tax Savings (accounting for depreciation, and tax on RECs
and expenses) plus REC creation minus expenses plus taxed electricity generation
The LCOE is the sum of the discounted future after-tax costs divided by the sum of the
discounted future electricity generation
The IRR is the discount rate that sets the sum of the discounted after tax cash flow (NPV) to
No cost of upgrading or extending the electricity transmission infrastructure is assumed.
Appendix 2: SAM economics inputs
The following is an extract of a paper3 which details the input cost assumptions that are covered in
“The optimum design must consider the capital cost, operations and maintenance cost, annual generation,
financial requirements, and time-of-use value of the power generated
NREL has developed a detailed cost model for parabolic trough solar power plants. The model is a based
largely on input from FSI, which supplied the mirrors for all of the Luz plants, and has been actively working to
promote parabolic trough plants since Luz’s bankruptcy in 1991 . FSI has developed a detailed cost model
based initially on the cost data from the Luz SEGS X project and later updated with more recent vendor quotes
. FSI provided cost data to NREL as part of its participation in the 1998 Parabolic Trough Road-Mapping
Workshop  and updated the solar field costs under contract
The FSI cost model is very detailed and uses reference quotes for each cost element. Land: A parabolic trough
field uses approximately one hectare per 3,000 m2 of collector area, or a coverage of factor of about 0.3 m2 of
collector for every 1.0 m2 of land area.
Site Works and Infrastructure: The site works and infrastructure includes general land preparation, roads,
fences, and site infrastructures, such as firewater system, warehouse, and control building. The cost model
assumptions are based on the FSI input. This category scales based on the size of the solar field.
Solar Field: The solar-field cost estimates are based on an updated cost assessment produced by FSI . The
cost estimate is based on the LS-3 collector design. Several adjustments are made to the collector cost to
account for a specific collector design used:
The number of receiver tubes, flex hoses, drives, sensors, and local controllers are adjusted per unit area
The drive costs are adjusted to account for the collector size.
The mirror, steel structure, pylons, header piping, and civil work costs are assumed to be the same on a
per-square-meter basis for different collectors. Heat Transfer Fluid (HTF) System: The HTF system includes
the HTF pumps, solar heat exchangers, HTF expansion vessel, piping, valves, and instrumentation. HTF
system costs scale based on the power-plant size, except for the HTF pumps, which scale based on solarfield
size. The HTF costs are based on the FSI roadmap data. The later data was only appropriate for an
Thermal Energy Storage (TES): The thermal storage costs are based on the detailed design study performed by
Nexant for a two-tank, molten-salt storage system . Thermal storage tanks and costs are based on detailed
data from Solar Two and Solar Tres. The heat exchanger costs are based on manufacturer quotes. Storage
costs were broken into mechanical equipment (pumps and heat exchangers), tanks, nitrate salt, piping,
instrumentation and electrical, and civil and structural. The mechanical equipment and piping,
instrumentation, and electrical costs were scaled by power-plant size. The tank, salt, and civil costs were scaled
by storage volume. All storage costs assume a scaling factor of 1.0, so a storage system twice as big costs twice
as much. Thermal storage tank and salt costs are consistent between the trough and tower designs. The
trough thermal storage system must be approximately three times as big as the tower storage system (both in
tank size and volume of salt required) to store as much energy because of the much lower temperature
difference between the fluid in the hot and cold tanks in the trough plant.
Power Cycle: The power cycle includes the steam turbine and generator and all condensate and steam cycle
equipment including pumps, heat exchangers, piping, valves, instrumentation, and controls. The FSI studies 
have the most recent Rankine steam-cycle cost data for the systems used in trough designs.
Balance of Plant: The BOP includes other power plant systems, such as cooling towers, water treatment and
storage, electrical, and control systems.
Contingencies: Contingencies of 10% are included for all costs, except the solar field (5%), structures and
improvements (20%), and thermal storage. The cost of the solar field is very well understood at this point. The
larger contingency for structures and improvements is included to account for potential differences in site
preparation. Nexant included cost contingencies separately in the thermal storage.
Indirect Costs: Indirect costs include services, project costs, and management reserve. The indirect cost
assumptions were based on input from Nexant. Service costs include project management, project
engineering, and construction management services. Project costs include permits and licenses, utility
connections, and telecommunication links. No interest during construction is included; this is accounted for in
the financial model.
The primary advantage of the NREL trough simulation model is that it integrates the capital cost, O&M cost,
performance and financial constraints into a single model. This allows detailed design or project optimizations
to be carried out where all interactions between cost and performance can be accounted.“
1 Solar Advisor Model User Guide for SAM 2009.10.12.
2 Solar Advisor Model – Parabolic Trough Default Cost Values. October 5, 2009. Refers to a study undertaken by
WorleyParsons commissioned by NREL under contract KAXL‐9‐99205‐00
3 Price, H. A Parabolic Trough Solar Power Plant Simulation Model. ISES 2003. March 2003