Exploring the Death Spiral:
A system dynamics model of the electricity network in Western
Australia
William Grace
Australian Urban Design Research Centre
University of Westem Australia
P PO Box 2729 Cloisters Square, Perth WA 6850
T: +61(8) 403 581 122
E: bill. grace@uwa.eduau
Abstract
Power networks worldwide are facing challenges from their own consumer base in the form
of private, grid-connected solar photovoltaic systems, and emerging growth in accompanying
energy storage. This paper reports the findings froma system dynamics model of the
electricity system of Western Australia, used to explore plausible scenarios resulting fromthe
impact of private solar and storage for the period 2015-2035.
The study finds the falling costs of solar PV systems will drive exponential growth that could
result in a tenfold increase in private solar capacity by 2025 - a much higher capacity than.
that currently predicted by the Independent Market Operator who operates the system.
Eventually the daytime export of excess solar energy to the network will be so great that
base-load generation will be affected, the network disrupted and tariffs will rise in a so-
called electricity death spiral. Despite this, economy-wide emissions and total energy costs
will be lower, which are positive outcomes for society and should be embraced rather than
resisted.
A coherent long termenergy strategy is required to address the major implications for the
network arising from the inevitable growth of private solar and storage, and for renewable
energy at the network scale.
Keywords
Electricity, Solar energy, Energy storage, System dynamics, Death spiral, Greenhouse gas
emissions, System costs.
Introduction
The electricity industry worldwide is talking about the so-called death spiral. Under this
scenario conventional electricity networks are undemined by customers reducing their
energy demand through energy efficiency measures and / or private generation, mainly
rooftop solar photovoltaic panels (PV). Both processes reduce the quantum of electricity
purchased from the network, thereby reducing revenue to the network. As many of the
network costs are fixed, this necessarily implies increasing unit costs and therefore increasing
tariff charges for electricity. The tariff increases merely exacerbate the problem - hence the
In this study these issues are considered in the context of Westem Australia's south-west
interconnected system (SWIS). The SWIS serves the south-west portion of the state, some
900,000 dwellings and 100,000 businesses. The effect of increasing rooftop solar PV and the
potential in the future for private electrical storage are modelled using the system dynamics
technique.
System dynamics has been used previously to simulate energy systems in general and electric
power systems in particular, notably by Professor Andrew Ford of Washington State
University (Ford 1997, 2007, 2008). The approach has also been used power industry policy
and strategy (Dyner and Larsen 2001, Bunn 1993).
The SWS
The system currently has a generation capacity of nearly 6,000 MW. In 2006, the previously
integrated system was split into three components: generation; transmission and distribution;
and retail. Synergy (a state owned business enterprise) owns around half of the generation
capacity and has a retail monopoly on accounts of less than 50 megawatt hours (MWh) per
year (essentially covering residential dwellings). Private generators provide the balance of the
energy source while private retailers compete with Synergy for accounts of more than 50
MWh per year. Westem Power (another state owned enterprise) operates the transmission
and distribution elements of the system, operating as a regulated monopoly.
An independent market operator (IMOWA)! operates the Wholesale Electricity Market?
(WEM). A key component of the WEM is the Reserve Capacity Mechanism (RCM). The
RCM was designed to incentivise investment to ensure that there is adequate generation and
Demand Side Management (DSM) capacity available each year to meet peak system
requirements. Retail electricity prices are set by govemment for Synergy customers.
Despite a growing population and economy, both peak and average demand on the SWIS
have plateaued in recent years (Figure 1 and Figure 2). Part of this change can be attributed to
general reductions in energy intensity in the economy (i.e. energy consumed per $ Gross
State Product).
300,000 10.0
90
250,000 80 5
Go
200,000 + aa
E on
% =
‘a. 150,000 50 €
%
6 «og
100,000 30 5
$
<
‘50,000 20
10
T T oo
2007-08 2008-09 2009-10 2010-11 2011-12 2012-13
Figure 1 Gross State Product and Energy Intensity
* The explanation of the IMO role is taken directly from their website http://www.imowa.comau
2 A review of the WEM is underway at the time of writing.
4,500 20,000
4 x 18,000
a 16,000 ina
3,500 —— Z ”
3,000 14,000 Min
2,500 12,000, Ave
z 2,000 -—- 1000 &
, | — ee * 3000 Energy (RHS)
7500 = _2—_5—__» = 6,000
47000 4,000
500 2,000
$ oS + Y
3 ra s go
a i A
Figure 2 Recent Energy Demand
Rooftop Solar PV
Rooftop solar penetration has increased significantly in recent years on houses within the area
served by the SWIS, and in Australia more generally (Table 1 and Figure 3).
Table 1 Solar PV in Australia 2013 *
Capacity|Proportion of
State; || Aaystems (mw) ‘ ath Solar Power
ACT 14,000) 38) 10%
NSW 252,000) 633 10%
NT 3,000) 11) 4%)
QLD 360,000) 986 22%)
SA 160,000) 450 25%
TAS 18,000) 55) 9%
vic 201,000) 532 10%
WA 149,000) 334 18%|
National |1,157,000) 3,039) 14%|
2003-10 2010-11 2011-12 2012-13
Figure 3 Rooftop solar PV on the SWIS*
3 Source: http://www.sunwiz.com au/index.phr ian-solar-pv-market- forecast. him
Households and businesses that export energy to the SWIS (at times when solar generation
exceeds electrical demand) are paid in accordance with the Renewable Energy Buyback
Scheme (REBS) which has just been reduced from 8.85 to 7.13 c/kWh, about one-third of the
household tariff. The goverment briefly introduced an additional feed-in tariff of 40c/kWh
but this was withdrawn (due to high subscription) in 2011.
The increase in take-up of private solar PV installations coincides with a significant drop in
unit price in recent years.°
2008 2009 2010 2011 2012
®Typical module price ~®=Typical small grid system price BOS price
Figure 4 Cost of solar PV in Australia ($/W)
Electrical Energy Storage
The other innovation that is likely to follow the rapid take-up of household solar energy is
energy storage. In the short term this will probably be battery storage and lithium-ion
batteries appear most likely to lead the market. Development of lithium-ion batteries is being
driven by both domestic scale renewable energy and the electric vehicle industry (Figure 5)°.
* Source: IMOWA
5 Source: cleantechnica.com
° There several factors associated with the large range in costs for existing lithium-ion batteries, including
battery size, required power demand / duration and geographical variations in manufacturing unit costs.
LI-ION BATTERY PACK COST AND PRODUCTION, 2010-
Total pack cost ($/kWh)
4,200 180
TODAY'S RANGE: 160
4,000 $400-1500/KWH ,
140
800 320
109
600
20
400 60
40
200
0
2010 2012 2014 2016 2018 2020 2022 2024 2028 2028 2030
Source. Bloomberg New Energy Finance
Bloombers (1 1]
Figure 5 Projected Cost and Production of Li-ion storage”
The Model
Model Structure
The purpose of the model is to explore the influence of growing solar penetration on the
SWIS network. The essential structure of the model is depicted in the causal loop diagram
below.
Population’ - Gross State
+ Product
Houses, Businesses
Nt s
Average household’ ey * at
jetwork deman
demand Energy intensity of
economy
Gg:
4 ‘Network energy
GHG emissions _,Benerated
Peak power
Unit price of
demand
network energy
Private energy
storage
capacity
Unit price of
solar PV
+
Private solar
PV capacity
Unit price of
energy storage
Figure 6 Causal loop diagram of model
” Bloomberg New Energy Finance Summit 2012
(http://about. bnef.com/summit/content/uploads/sites/3/2013/11/BNEF 2012 _03_19 University Battery Innov
ation.pdf)
5
The dynamic hypothesis is that the reducing price of solar PV systems (an exogenous
variable in the model) will increase private solar capacity which in tum will reduce the
quantum of energy generated on the network, pushing up unit prices and further incentivizing
take-up of private solar PV® (Loop R1). This reinforcing loop is the essence of the so-called
death spiral.
However, this is only part of the story. An increase in private solar capacity also reduces
network generation costs by reducing both the total amount of energy generated and peak
demand (Loops B1 and B2). Reducing peak demand directly reduces the cost of generation
(peaking plants have higher operational costs), and also reduces the necessary network
capacity, thus reducing capacity costs (Loop B3). Both factors will tend to offset otherwise
increasing unit prices (balancing loops).
The addition of private energy storage does not change the quantity of energy generated, but
does reduce peak network demand, thus creating a further balancing loop (Loop B4). This
suggests that increasing private storage will somewhat offset the impacts of private solar
generation on the network.
The behavior of the real system over time is determined by the relative strengths of the
reinforcing and balancing loops. The results of most interest are:
e network costs;
e network unit prices;
e the overall economic costs of supplying electricity to the community; and
e greenhouse emissions
The model uses Version 6.3 of the Vensim? Professional software. A fuller explanation of the
model structure, together with its documentation, is included in the supplemental information.
Electricity Demand
The model determines the electricity demand arising from:
e residential houses; and
¢ commetcial and industrial facilities.
The existing residential demand and commercial demand profiles have been determined from
historical half hourly reports of total network load, and presentations of the IMO WA on
residential and commercial loads. The annual residential demand has been derived from the
reported network loads, modified to include the demand met by private solar).
® Of course financial benefit is not the only driver behind the take-up of solar PV, including individual concem
about global warming. Although these are difficult to characterise and have been neglected in this model, they
will clearly only add to the momentum for growth.
° http://vensim.con/
GWh
Reported 2012-13 residential load 5,035
Energy produced by 336 kW private solar 540
Estimated total residential demand 5,575
Reported 2012-14 commercial load 12,914
Estimated total residential and commercial demand 18,489
Hourly demands for a typical day of each month of the year are inputs to the model. The
profiles for a typical January and Jume day are depicted in Figure 7 below. As these are
typical days they do not reflect the absolute annual peak demand (around 3,700 MW in
2012/13).
The recent demand history for residential electricity indicates that the average usage per
residence has not changed over the last five years if the growth in solar energy is taken into
account’®. Accordingly the electricity demand per dwelling is a constant in the model.
Average January day Average June day
3,500 m Commercial demand. 3,500 commercial demand
3000 MResdentieldemand 3,000 Residential demand
2,500 2,500
= 2,000 = 2,000
2 1500 = 1,500
1,000 1,000
500 500
1 4 7 10 2B 16 19 22 1 4 7 10 2B 16 19 22
Hour of the day Hour of the day
Figure 7 Electricity demand
The IMO base case forecast assumes population to grow at 2.1% per annum. The model
assumes that the number of houses will grow at this rate as a default but can be readily varied
through a slider.
The commercial and industrial demand is calculated in the model as the product of the state’s
Gross State Product (GSP) and the energy intensity (i.e. MWh per year per dollar of GSP).
The recent data shows that energy intensity has been dropping by approximately 1% per
annum and this is the default assumption in the model (again this can be readily varied). GSP
is forecast by the Treasury to grow at 3% per annum and this is the default figure used in the
model. It is assumed that the number of businesses grow by this amount. Again, these values
can be readily varied through sliders.
Residential solar
The model calculates the contribution of household scale solar energy generation. It
incorporates the following elements:
*° There is no research shedding light on the reasons why the recent regulatory changes requiring improvements
in themmal efficiency of housing has not reduced demand. The most credible theory is that benefits arising from
these measures have been offset by larger houses per person and demand from additional consumer appliances.
7
e Solar energy without storage; and
e Solar energy with storage.
Solar without storage
The existing number of residences with solar systems (15.5%) and the average size of their
solar array (2.4 kW) have been taken from the IMO data. The impact of a 2.4 kW amay on
network loads is shown in the figures below for typical January and June days.
15 15
Janua June
i am 1-
0S 0.5
0 7 0
= z
-0.5 -0.5
4 Nao! 1 — Daily residential demand by hour
15 Daly resid ential demand by hour “15 — Daily network load from a solar only house
: — Daily network load from a solar only house i
2 2
1 4 27 6 1 i i DD 1 4 7 0 2B 6% 19 22
Hour of the day Hour of the day
Figure 8 Impact of solar PV on network loads
It can be seen that solar has only a minor impact on reducing residential peak loads as these
occur in the evenings. While significant energy is exported to the network in summer, exports
are limited in the winter. Larger solar arrays maintain the shapes identified above, but the
troughs in network load become larger.
The model calculates the payback period for a household arising from:
e avoided electricity imports from the network at the residential tariff; plus
e electricity exports to the network at the residential feed-in tariff (renewable buyback
scheme); and.
e the installed cost of solar energy.
Both the size of the solar array and the fraction of houses with solar energy grow towards a
maximum fraction as a function of the payback period (Figure 9). The default value for the
maximum fraction of houses with solar is 60%, and maximum solar array is 7.2kW (3 times
the present average). At the commencement of the simulation period the payback period for
residential solar is around 7 years and the model assumes that this will incentivise some 10%
of remaining homes to purchase solar (spread over an adjustment time of 6.5 years). This
percentage increases as the payback period decreases until it reaches 1 year, at which point all
remaining houses would be incentivised to install solar PV.
% uptaeke
g
|
Payback in years
Figure 9 Uptake of solar PV
Avoided electricity imports are calculated in the model on an hourly basis in accordance with
the profiles exemplified in Figure 8. Savings are determined as the product of the total
avoided energy and the applicable network tariff. The existing residential time of use tariff
(SM1) is used by the model to calculate normal hourly, monthly and annual charges. The
model assumes that tariffs change over time pro-rata to changing unit costs of network
energy.
Electricity exports are merely the difference between the annual household solar generation
and the avoided imports. The value of exports is derived from the existing solar feed in tariff
($0.0713 / kWh).
The unit cost of solar PV is modelled as a stock with an initial value reflecting present unit
costs ($2,200 /kW). This is the approximate installed cost of systems in Australia presently,
excluding the benefit of the small scale technology certificates (STCs) which are presently
worth approximately $690 / KW in the SWIS area. As there is currently uncertainty about the
continuation of the STC scheme, the model neglects that benefit. The model assumes that the
unit cost transitions to a final unit cost ($1000 /kW) in accordance with Figure 10. The curve
is based on a review of published forecast costs including extrapolation of Figure 4 and the
U.S. Sunshot Target of $0.06 / kWh". Both the final unit cost and adjustment time can be
easily varied in the model.
™ http: //energy.gov/eere/sunshot/sunshot-initiative
2,000
1,500
1,000
$/kW installed
500
° -
2010 2015 2020 2025 2030 2035 2040
Figure 10 Solar PV cost curve
The quantity of residential solar electricity generated (in MW) is determined by the
residential solar capacity (total number of houses x fraction of solar houses x the average
residential array in kW) and the amount of solar energy generated annually per unit of
Solar with storage
The optimum storage capacity is 0 for solar arrays which are not sufficiently large to
completely offset demand during the hours of generation (summer conditions determine this
threshold which is about 1.1 kW for residential systems). For solar arrays above this capacity,
there is increasing benefit in storing more energy to avoid network imports. However, there is
an upper limit (about 10 kWh of storage for residential systems), above which there is
excessive storage capacity for the amount of generation. This upper limit has been
determined by selecting a storage capacity which discharges to (approximately) 0 on summer
days'*, Based on the hourly model of solar PV and household demand in Perth, the
approximate optimum storage capacity for a given solar capacity was determined for both
household and business systems’ (Figure 11). The difference between households and
businesses in respect of the minimum value is due to the difference in demand pattems.
” Taken as an average of 4.4 KWh/ KW in Perth (Clean Energy Council 2011).
* This is an approximation - most battery storage systems are recommended to only discharge regularly to
around 30% of capacity.
“The approach to selecting optimum storage is described in the Model Explanation in the Supplemental
Information.
10
Storage capacity /
‘Average hourly
demand
Solar capacity/
‘Average hourly demand
Households 15 5
Businesses 175
Figure 11 Optimum storage capacity
By adopting this approach to setting storage capacity, the model avoids multiple
combinations of solar capacity and storage capacity. Solar capacity in the model is
nonmalised as a fraction of average hourly demand.
The model calculates the additional benefit to the householder from adding storage to their
solar array. It is assumed that storage operates simply on the basis that:
e solar generation in excess of demand is stored (up to the limit of the storage capacity);
e the storage discharges to meet demand that cannot be met by solar generation; and
e remaining demand is met by the SWIS network.
The model determines the fraction of energy imported on an hourly, daily and annual basis
with the selected combination of solar array and storage.
Larger scale solar arrays will generate more electricity than can be optimally stored so there
is also a component of generation that is exported (allowing for storage losses).
The additional benefit of storage is therefore determined by:
¢ savings from the additional avoided network imports (at the normal tariff); plus
e the benefit of exports (at the feed in tariff).
The incentive to add storage is again determined by a payback period calculated from the
benefits noted above and the unit cost of solar storage. The latter has been determined from
the technical press and assumes the present storage costs of approximately $1,000 / kWh will
drop to around $200 / kWh (Figure 12). These values may be considered conservative given
the curve set out in Figure 5.
11
$/kwh
8
2015 2020 2025 2030 2035
Figure 12 Storage cost curve
The fraction of houses with storage trends towards the number of houses with solar energy in
accordance with the same algorithm assumed for solar arrays. The fraction is assumed to be
zero at the commencement of the simulation.
At any given time therefore the model predicts:
e the average residential solar array;
e the number of houses with solar arrays; and
e the number of houses with an optimal storage system
The combination of these impacts on the network at each time step is used to calculate the
network load on an hourly basis for each month of the year from:
e A house without solar
e A house with solar but no storage; and.
e A house with solar and optimum storage.
The net impact is merely the product of these values and the number of houses in each
category.
Conmercial Solar
The model structure for the commercial solar is identical to the residential model in all
respects, except for the initial conditions. The model assumes that at the outset there is no
commercial solar or storage.
There is no available information on the potential for uptake of solar PV in the commercial
sector. As this sector incorporates a spectrum of activities spanning small service businesses
in rented and shared accommodation and large industrial facilities with large roof space, the
opportunities will vary widely. In this study the maximum fraction of solar businesses
defaults to 50% in the model but this is variable. Because the electrical demands of the
business sector are much greater than the residential sector and the available space (on
average) for solar PV is larger, the model sets a much larger array size for the maximum
average array - 150kW. However, unit costs for both solar and storage are assumed to be the
same for residential and commercial systems.
12
It is important to note that paybacks for business PVs are marginally shorter than for
residential customers because the commercial peak is earlier in the day and is therefore
served more effectively by solar generation, meaning that more of the generated energy goes
to import avoidance (at the normal tariff) rather than export (at the lower REBS tariff).
The functions controlling take-up of both solar and storage are the same as for the residential
sector, as is the payback function. Payback again is determined by the savings and benefits
arising from avoided electricity imports and exports. In this case, the business tariff R1 is
read into the model and is used in this calculation. The model simplistically assumes that
tariffs change over time pro-rata to changing unit costs of network energy.
Utility Network
The network generation capacity is used to calculate the recurrent generation costs and spot
price arising from the Short Term Energy Market (STEM). The initial capacities of each type
of generation are based on the existing capacity credits allocated by the IMO, and other
information.
Coal 1,777 MW
Gas combined cycle (Gas CC) 715 MW
Gas combustion turbines (Gas CT) 2,609 MW
Diesel 210 MW
Wind 169 MW
Total 5,480 MW
Information on the capital and operating (including fuel) costs and other characteristics of
each type of generation has been derived from the Australian Energy Technology Assessment
(AETA) by the Australian govemment’s Bureau of Resources and Energy Economics (BREE
2012).
The model assumes there are no additions to the existing generation capacity (including
wind). As the network is presently over capacity the model allows for retirements to each
type of thermal generation. The default condition is for coal retirements at 20 MW / year
throughout the simulation period, commencing in 2016.
The network loads arising from residential and business premises are aggregated and, these
figures together with the generation capacity, are used to determine the hourly spot price on
the network. Under the present arrangements only a small proportion of energy is purchased
via the STEM. The vast majority is traded bilaterally by generators and retailers. As
information on these trades is not publicly available the model assumes all generation is
dispatched via a spot price mechanism. Accordingly, the spot prices in the model are not
directly comparable to the actual STEM prices on the SWIS.
The spot price mechanism applies only to the thermal network, assuming that both private
solar and network wind generation “must-run”.
Average bid prices for each type of thermal generation are assumed to approximate the fuel
and variable operating costs prices associated with each type of generation (set out in the
AETA report), together with a 10% markup. This yields the following average bid prices:
13
Coal $35.10 / MWh
Gas CC $98.22 / MWh
Gas CT $143.15 / MWh
The model assumes that each generation type will bid these figures + 20% depending on
demand. This model structure derives an hourly spot price on the assumption that the lowest
cost available generation is deployed. A ceiling price of $200 is assumed.
Wind generation in the model is derived from random function that assumes a capacity factor
of 0.38 on average is achieved (a figure normally assumed for the SWIS).
The model also calculates the hourly and monthly costs of generation, assuming that each
type of generation costs its average hid price per hour to operate. These calculations derive a
unit cost ($/MWh) for the generation component of system costs.
System Costs
The model calculates the total SWIS costs. The monthly fixed cost of generation is added to
the generation component described above. The fixed costs are again derived from the AETA
report. Guidance for the calculation has been derived from the IMO report outlining the
calculation for the reserve capacity credit cost per MW (which assumes that gas combustion
turbines will determine this figure).
The annual fixed cost of generation has been taken to be the annualised cost of capital and
fixed operations and maintenance costs amortised over 30 years using the weighted average
cost of capital of 7.01%. The capital costs per MW" are derived from the AETA report for
each type of generation.
The monthly costs of transmission, distribution and retail (TDR) are added to the generation
costs to give a total system cost figure per month. The costs for these items have been
calculated on the basis of the existing reported costs for the SWIS"®. The following includes
the cost of govemment subsidy to calculate an approximate overall unit cost.
Capacity 5,685 = MW
Energy 18,133,609 MWh
$/MWh $mpa $/MW
Generation costs 127.69 2,315
Transmission & distribution 105.67 1,916 337,064
Other costs 61.64 1118 196,621
Assume total. unit cost 295.00 5,349
The model assumes the cost of transmission and distribution is $337,000 per MW pa if the
capacity of the system remains the same or grows. If capacity falls below the initial value the
5 4 markup of 50% has been applied to the capital costs set out in the AETA to allow for legal, financing and
other costs associated with purchase of the asset.
18 http://www. westempower.com.au/aboutus/save_electricity/The_price_of_power_.himl
14
costs are assumed to remain at their current figure on the basis that this infrastructure will
continue to be in place and cost approximately the same to operate.
Other system costs are assumed to vary in accordance with capacity, i.e. at the rate of
$197,000 / MW pa.
The total system unit cost is simply the addition of the generation and transmission,
distribution costs and other costs (including retail) divided by the quantity of network
electricity generated. The average system unit cost is simply the average of the previous
year’s monthly values. The model assumes that changes to tariffs are pro-rata to unit cost
increases and applied as a multiplier to the existing household and business hourly tariff
regime (one year in arrears).
The model includes the cost of the REBS as a system cost, although the REBS tariff is
assumed to remain constant at the current rate.
The present network is heavily dominated by fossil fuel generation. At the time of writing, a
carbon price has been removed in Australia and hence no cost has been accrued in this study
to account for greenhouse gas emissions. Given the global impetus for pricing carbon, it is
unlikely that this situation will persist for the duration of the simulation period.
All costs in the model are un-escalated and therefore quoted in 2014$.
Greenhouse Gas Emissions
The greenhouse emissions arising from the thermal network are also calculated. The emission
intensities of each type of generator have been taken from the AETA report and the Account
Factors from the National Greenhouse and Energy Reporting scheme’’,
These figures combined with the monthly generation regime determine the emission intensity
of the network as a whole, which is presently 0.76 TCO,-e/ MWh.
Solar Costs
The model calculates the incremental solar and storage investment costs from the inflows to
the stocks of houses and businesses and multiplies those flows by the fraction of premises
that have solar or storage at that time step, and the unit cost of solar and storage. This
identifies the total private investment in solar and storage at each time step.
Public and Private Expenditures
The solar and storage investments are added to the recurrent network generation costs to give
an indication of the total private and network expenditures throughout the simulation period.
The calculation neglects the operations and maintenance costs associated with private solar
and storage systems.
This calculation also neglects the investment in both private solar and the existing thermal
network generation prior to the simulation period.
”” http://www. environment gov.au/climate-change/greenhouse-gas-measurement/publications/national-
greenhouse-accounts-factors-july-2014
15
Model Results
Scenarios Considered
The model has been used to investigate a number of scenarios related to the possible growth
of private solar PV and storage and the impacts of this growth on the SWIS network. For
comparison purposes a Base Case is included which assumes:
. Economic growth of 3% pa;
. Population growth (represented by housing growth) of 2.1% pa;
° The penetration of private residential solar remains at the present value of 15.5%;
. The average size of arrays remain at 2.4MW;
. There is no residential storage; and
. There is no business solar PV or storage.
° The recent reductions in energy intensity (approximately 1% pa) continue; and
. The network thermal capacity is reduced by 2OMW pa from 2016.
The model calculates demand and generation hourly (assuming a typical hourly pattem for
each day of each month) and runs for 20 years starting in 2015.
The purpose of the model is to explore the underlying dynamics of the SWIS over the
medium term. It does not seek to precisely simulate the short term dynamics such as the
backlog of works like pole replacements and undergrounding works. It assumes that the
essential nature of the existing system (including transmission and distribution costs) is
sufficiently representative to explore systemic changes over the 20 year time horizon of the
model.
Results - Solar Growth without Storage
This case examines the implications of residential and business solar growth without storage,
Le:
e growth in residential solar PV penetration continues; and
e growth in business solar PV penetration commences in 2015.
Under the model assumptions, payback periods continue to fall as system sizes grow and unit
costs reduce. This results in rapid growth in both the residential and business sectors. By
2035 around 50% of houses and 40% of businesses have solar PV (Figure 13).
16
™ Houses with solar
@ Houses without solar
9000 +-— m™ Business solar capacity
8000 --— m Residential solar capacity
1600
1400 +-
1200
00 +}
800
600
400
200
°
10000
increasing to around 4.5 kW for residential systems and 90 kW for business systems. The
Figure 13 Growth in solar PV penetration
total capacity growth is depicted in Figure 14.
203!
Reducing payback periods also lead to larger installed systems with average array sizes
spuesnoy)
Wy
occurs. By 2025, average hourly loads on the network are significantly reduced from the Base
There are significant implications for the SWIS if this scale of growth in private solar PV
Case (Figure 15) and this divergence grows thereafter.
Figure 14 Growth in solar PV capacity
3000
Se ae
1500 =
——Solar growth case
MW
500 -— ——Base Case
o+—r— — ——
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 15 Average hourly network loads (2025)
However, maximum loads are only marginally reduced by daytime solar generation.
Accordingly, the maximum hourly loads by 2025 show only a minor reduction in comparison
to the Base Case (Figure 16).
3500
3000
=
2000
=
Solar growth case
1500 +
—Base Case
1000
500
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 16 Maximum hourly network loads (2025)
This means that the capacity of the network is required to be maintained at the levels of the
Base Case, with consequent capacity based costs.
However, the most significant implication of this scale of solar generation is on minimum
network loads, i.e. when solar generation is at its maximum during the middle of the day in
general and in the summer in particular. By March 2030 available solar exports exceed the
total electricity demand in the middle of the day and accordingly, network loads fall to zero
(Figure 17).
18
Hourly demand
Solar exports
Hour of the day
Figure 17 Solar exports (March 2030)
The lead up to this circumstance is shown in the following figure which depicts the so-called
“duck curve” (Figure 18). This encapsulates the problem for networks of accommodating
highly varying daytime solar generation.
Typical March day
——2015
—2018
—2021
——2023
——2026
Hour of the day
Figure 18 Duck curve
This suggests that there could be an intermittent over-generation problem by 2020 when
minimum network loads fall below the normal operating capacity of baseload coal
generation, which is intended to run consistently and cannot be readily cycled down and up in
a matter of hours. Steep ramping of generation is required in the hours of declining solar
aaa’) A similar situation has been identified by the Califomia Independent System
Operator”.
18 ttp://www.caiso.comydocur ibleresourvest 3 fastfacts.pdf
19
Results - Solar Growth with Storage
This case examines the implications of residential and business solar growth with storage,
ie:
e growth in residential solar PV penetration continues;
e growth in business solar PV penetration commences in 2015; and
e both areaccompanied by growth in private storage.
Storage payback periods in the model are dependent on storage costs, savings and REBS
income which are all a function of solar amay size. Accordingly, it takes some time for
paybacks to drop to the level where take-up would be financially attractive. However, after
around 2020 paybacks have dropped to the 10-15 year range.
Penetration thereafter increases steadily in both residential and business facilities (Figure 19).
By 2035, there is some 13,000 MWh of storage capacity and 405,000 houses and 45,000
businesses possess storage (Figure 20). However, this is still only about 1.5 hours of storage
at the nameplate solar capacity.
1600 200
Thousands
a2 8
Businesses with solar / no storage
lM Businesses without solar
2035 20m 25 2m 25, 2015 2020 2025 2030 2035
Figure 19 Energy Storage Growth
14000
12000 |. ™ Business storage capacity
™ Residential storage capacity
(i
i i!
¥ i
2015 2020 2025 2030 2035
Figure 20 Storage Capacity
20
By observation (Figure 20) the storage case has little impact on the network for the first half
of the simulation period. The duck curve up to 2025 is similar to that for the solar only case.
It is only thereafter that the influences of storage are felt. By 2034 storage has reduced
maximum network loads significantly from the solar only case (Figure 21).
2034 Storage growth case
Solar growth case
Mw
w
8
\ fT
AER
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 21 Maximum hourly network loads
The over generation problem is somewhat ameliorated in the storage case but minimum.
network
4500
loads will still fall to 0 under certain circumstances (Figure 22).
——Solar plus storage case oe
|. ==—Solar only case
A. Y om
==
=i FS
1 4 7 10 13 16 19 22
Figure 22 January day network loads (2034)
The Implications of the Model Results
Solar growth
In a recent report, the IMO forecast the growth of private solar to grow linearly for the
coming decade to reach approximately 1,000 MW of nameplate capacity by 2024.
pak
1,200
1,000
2009-10 2011-12 2013-14 2015-18 2017-18 2019-20 2021-22 2023-24
Historic SHighcase “Expected case += ——=Low case
Source: IMOMMIEIR
Figure 23 IMO Solar penetration forecast
Tn contrast the model, which is based on an increasing take up as payback periods reduce,
suggests that this figure could be as high as 3,000 MW by 2024 and 3 times that figure by
2035.
The IMO forecast is based on linear growth. Exponential growth is a more common
phenomenon, and is usually observed in the time histories of innovative technologies,
including solar PV take-up in Australia’.
Although growth in the commercial sector has been slower than residential to date, as
awareness grows and unit prices decline, this is potentially the largest contributor to growth
by far. Payback periods are lower for commercial systems because there is a better match
between solar generation and demand, and time-of-use tariffs are higher during peak periods.
Network loads
The most dramatic effect on the network of growth in solar penetration of this scale is seen in
the impact on minimum loads on the network (Figure 24). This illustrates the impact of
growing solar exports to the network during daytime periods in general and in summer in
particular.
18 http://apvi.org.au/wp-content/uploads/2014/07/PV-in-Australia- Report-2013.pdf
22
— Solar only
Solar + storage
2000
1500 +
2
2 \ |
1000 |
500 |
, L
2015 2020 2025 2030 2035
Figure 24 Minimum hourly network loads
The following figures (Figure 25 and Figure 26) show the percentage of hours in each month
that network loads may reach over-generation and zero generation points.
2025
50%
as% + ——Network load < 1250 MW
40% + Network load = 0
35% —
30% —S —,
25%
Pa \
7 BY
10% if 7
5%
om \*
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 25 Over-generation risk by 2025
2034
50% ———__—— —
aese —Network load < 1250 MW
40% 4 —Network load = 0
| ee Nee ee
15% ie, _“ Z
ne bn Va
0%
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 26 Over-generation risk by 2034
Over the longer term the “hollowing-out” of network load would likely require a different
configuration of generation type on the network.
23
Private energy storage ameliorates the impacts of solar generation on the network as it
reduces both the peak and average energy demands, thus lowering system costs.
With projected growth in storage lagging the growth in solar PV, storage only partially
offsets the over-generation problem. By the end of the simulation period there is only around
1.5 hours of storage (at nameplate). This means that the amount of over-generation from solar
will likely continue to increase, albeit to a lesser extent than without storage (Figure 27).
2030
—— Solar + storage
— Solar only
al \ 5 VA
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 27 Typical day over-generation
Network Energy and System Costs
All of the cases considered result in lower annual energy required from the network than the
Base Case. In the most likely case that business premises also adopt private solar and storage
there would be a dramatic reduction in required energy, both in relative (45%) and absolute
terms (15%) as identified in Figure 28.
——Base case
—Solar growth case
——Storage growth case
10000 +
2015 2020 2025 2030 2035
Figure 28 Annual network energy
A reduced energy demand translates to reduced system costs. Under the storage scenario
(residential and business solar and storage), system costs would rise only slowly, and then
potentially fall to near current costs (Figure 29). This is due partly to a reduction in the
capacity of the network (assumed mainly related to coal generation), which induces savings
24
in the cost of capacity credits. However, this neglects any ongoing debt obligations associated
with this generation capacity which, while not a direct cost to the SWIS, is a cost to publicly
owned entities and hence the taxpayer. The costs of writing off this debt are not included in
this study.
——Base Case
Solar growth case
——Storege growth case
5600
5400 ae
5200 r
2015 2020 2025 2030 2035
$m per annum
w
3
8
Figure 29 Annual system costs
If tariffs presently delivered $295 / MWh, the SWIS would recover all costs. The model
suggests that unit costs will fall from this figure under the Base Case conditions. However,
under the other cases it is likely that, although overall system costs will decrease, unit costs
and therefore tariffs will increase substantially (Figure 30).
380
360 --— "Base Case
340 -|_— Solar growth case
Storage growth case
$/MWh
200
2015 2020 2025 2030 2035
Figure 30 System unit costs
Although reduced system costs indicate that increased solar penetration will lead to an overall
economic benefit, a rise in unit costs is still problematical. Individual consumers, who for one
reason or other cannot reduce their network energy consumption sufficiently to offset tariff
increases, will pay more for energy under this scenario. As this group will include those who
are least able to absorb the additional cost, equity will become an important element of the
policy response.
25
A Broader Perspective on Costs
Of course, the SWIS costs do not represent all the costs incurred as they do not take into
account the private investments in solar PV and battery storage. The model tracks these costs
and accrues them over time to enable a more complete cost comparison of the options (Figure
31).
14,000 |~ storage growth case
12,000 +— ——-solar growth case ft
10,000 Pl
$m
g 8
\
2015 2020 2025 2030 2035
Figure 31 Private investment in solar/ storage
A range between $11bn and $15bn is the most likely case for private investment in solar and
storage. As the net annual savings for the SWIS accrue to around $7.6 bn, the 20 year
accumulation of private and SWIS costs are only marginally higher than the Base Case
(Figure 32).
©
2 Storage
ix mSolar
@swis
Base Case Solargrowth case Storage growth
case
Figure 32 20 year public and private expenditures on energy
The Influence of Tariff Increases
Part of the death spiral metaphor is that increasing solar penetration will lead to higher tariffs
which will only further improve the cost-benefit equation for consumers. By default, the
model increases existing household and business tariffs pro-rata to system unit cost increases.
However, the influence of increasing tariffs is easily tested by “switching off” tariff increases
in the model. Doing so demonstrates that the tariff increases are not the main driver of
26
reduced paybacks. The reducing unit cost of solar PV has much more influence than
increasing tariffs (Figure 33).
9
8 —Case 28 _—
e7 ——No tariff increases
&
Se
Bs)
a4
x
&3
=
a2
1
o :
2015 2020 2025 2030 2035
Figure 33 Influence of tariffs on payback periods
Greenhouse Gas Emissions
The Base Case indicates that greenhouse gas emissions could rise by around 20% by 2035.
The most likely scenario, i.e. growth in private residential and business solar and storage
would reverse this and deliver reductions of around 25% (Figure 34). However, even this
contribution would be insufficient to represent a credible emission reduction target for south
west Westem Australia.
|
——Base Case
—Solar growth case
Mt CO2-e per annum
boo wo
——Storege growth case
oO
2015 2020 2025 2030 2035
Figure 34 Greenhouse gas emissions
Policy Response
Transition Strategy
The current review of the Wholesale Electricity Market (WEM) does not address the
dependence of the SWIS on fossil fuel generation, a situation that cannot continue
imrespective of the current adverse political environment. Even if supply is not dampened
through action on global warming, prices will inevitably and steadily rise forever after peak
production in gas and coal later this century (Maggio and Cacciola 2012).
27
Electricity industry investments are made for decades and so it is essential that the future
energy generation mix and network strategy is established now to ensure that new
investments minimise the risk of stranded assets.
This study identifies that the growth of renewables in the form of private solar is inevitable
and will have major implications for the network irrespective of any changes likely to arise
from the WEM review. The energy system will change, and therefore the implications of this
study should be considered in the context of this broader transition, for which a coherent long
term energy strategy is required. The inevitable increase in the take up of private solar PV
systems in WA homes and businesses will merely hasten a transformation of the electricity
network during the coming decade that is needed anyway.
An Integrated Energy Policy
The growth of private solar PV is being driven by global forces which are leading to lower
unit costs for solar PV panels and economies of scale are driving down balance of system
costs. Although it may be affected in the short term by various factors, Westem Australia
cannot escape the reality of this momentous technological shift.
The days of the electricity industry being the sole provider of energy services to consumers
are over. They are now competing with their customers and their response to this challenge
will determine where the balance between network and private assets eventually lies. Policy
must drive the most efficient economic outcome, not seek to “protect” the existing industry
players. Lower emissions and lower total energy costs are positive outcomes for society and
should be embraced, not resisted. While the fact that Synergy and Westem Power are state
owned enterprises obviously must influence policy, it should not cloud the fact that major
change is inevitable. The future energy system must effectively and efficiently integrate
private and network generation.
Network Storage
This study identifies that excess solar generation in daytime hours will create an over
generation problem on the network in the coming decade, initially in the mid-seasons and
then in the summer. This will require either network baseload generation to be intermittently
reduced and / or private solar generation to be “floated”. The former is problematical from
the operational perspective and the latter wastes energy that has no marginal cost. Although
private storage will ameliorate this situation somewhat, the capacity of private storage will
not be sufficient to eliminate it
The only way to avoid the steep ramping evident in the “duck curve” is to introduce network
storage into the SWIS. This could potentially occur at existing substation sites which dispatch
and receive electricity from the private systems in homes and businesses”. The introduction
of storage, if commenced soon enough and with appropriate policy settings (e.g. in respect of
the REBS rate), could potentially “head off” the growth in private storage modelled in this
study. Economies of scale would mean lower costs per MWh for network scale storage, and
?9 Although not covered in this study, storage at this scale could also facilitate decentralised micro-grids
operating at the precinct / suburb scale.
28
if this was of sufficient scale to store all excess private solar generation, this could lead to the
encouragement of private solar while dis-incentivising private storage.
Storage at this “downstream” scale would logically be complemented by larger scale storage
“upstream” aimed at smoothing supply and demand from network generation. This would be
part of a strategy to transition generation from fossil fuels to renewables, many of which are
intermittent in nature, e.g. wind. Ontario is one jurisdiction that is planning for network
storage for this purpose. The Independent Electricity System Operator (formerly the Ontario
Power Authority) has procured approximately 35 MW of storage and is presently in a tender
process to procure a further 15 MW. The United States is expected to add some 220 MW of
energy storage to networks in 20157.
System dynamics has previously proven useful in simulating the role of centralized energy
storage. Examples include the proposed strategic fuels (i.e, gasoline) reserve in Califomia
(Ford 2005) and the use of pumped hydro storage for wind integration in the Pacific
Northwest (Llewellyn 2011).
Professor Andrew Ford (personal commumication 2014) has developed an innovative
approach to the system dynamics modelling of network systems with storage that combines a
long-term model (30-year interval, monthly time step) with an operations model that
simulates a typical week (hourly time step). Inputs for the operations model are selected to
match the corresponding results for a particular month and year in the long-term model. The
operational simulations are studied to obtain aggregate measures of performance of the
storage facility, with results transferred to the long-term model.
Tt is proposed to extend the model described in this study by incorporating network scale
storage in the SWIS in Westem Australia.
** http: //energystorage.org/news/esa-news/us-eneryy-storage-market-grow-250-2015-0
29
References
Bunn 1993: Derek Bunn, Erik Larsen and Kiriakos Vlahos, Complementary Modeling
Approaches for Analysing Several Effects of Privatization on Electricity Investment, Joumal
of the Operational Research Society, 44, 10.
BREE (Bureau of Resources and Energy Economics) 2012, AETA Report and Model
Version 1_0, 2012, Australia. http://www. bree.gov.al
Clean Energy Council 2011: Consumer guide to buying household solar panels.
http://www.solaraccreditation.com.au/consumers/purchasing-your-solar-pv-systenysolar-pv-
Quide-for households. html
Dyner and Larsen 2001: Isaac Dyner and Enk Larsen, From Planning to Strategy in the
Electricity Industry, Energy Policy, Vol 29, p. 1145-1154.
Ford 1997: Andrew Ford, System Dynamics and the Electric Power Industry, System
Dynamics Review, Vol 13 (1).
Ford 2005: Andrew Ford, Simulating the Impacts of a Strategic Fuel Reserve in Califomia,
Energy Policy, Vol 33
Ford 2008: Andrew Ford, Simulation Scenarios for Rapid Reduction in Carbon Dioxide
Emissions in the Westem Electricity System, Energy Policy, Vol 36.
Maggio and Cacciola 2012: G. Maggio, G. Cacciola, When will oil, natural gas, and coal
peak? Fuel 98 (2012) 111-123
30
