‘It is a misconception to think that the technological alternatives that we can draw upon are without flaws and that only a combination of conspiracy and inertia prevents us from sailing into calm waters’ (Robert Socolow, The Ethics of Global Climate Change, 2011).
The buildout of renewable energy across key electricity markets has been nothing short of staggering. Globally, solar and wind generation capacity has grown from 0.1GW and 25GW respectively in 2001, to 170GW and 400GW in 2014 [13, 15-16]. The Australian National Electricity Market (“NEM”) is not to be outdone. Australia has the highest penetration of rooftop solar in the world and on 10 May 2015, South Australia recently supplied 89% of its generation on by wind alone . However, this rapid energy transition brings with it its own challenges and issues – key among which is the effective integration of variable renewables into a grid that has been historically catered towards dispatchable centralized generation. A series of articles over the next few weeks will examine several critical elements of grid integration including grid stability, frequency control and voltage regulation.
This first article will focus on generation variability and the challenges of system balancing under the new grid. First, we will explain the well-known Californian electricity ‘duck curve’ – which illustrates the intraday load patterns and grid strain under increased renewable penetration. Second we will assess the intraday load under the NEM in order to understand our own ‘duck curve’. Finally, we will observe that patterns vary significantly across the various NEM regions and conclude with some implications for regulation and grid operation.
CAISO Net Load – The Californian Duck Curve
In 2015, California sourced 18% of generation from non-hydro renewable generation. Renewable penetration is expected to grow driven by policy measures including a 50% Renewable Portfolio Standard (“RPS”) by 2030, an emissions trading scheme and local and national environmental regulations. This has resulted in major shifts to intraday load and generation patterns set out in the famous ‘duck curve’ produced by the California Independent System Operator (“CAISO”). The curve illustrates the system ‘net load’ – that is demand minus (non-dispatchable) wind and solar generation i.e. effectively the generation that needs to be supplied by conventional (predominantly thermal) generation. It shows that with increased solar and wind penetration, the net load is increasingly low during the day and then ramps up in the evening, as solar generation tapers off – creating a ‘duck-like’ pattern.
Figure 1: CAISO ‘Duck Curve’
Source: CAISO (2016)
NEM Intraday Net Load
How does net load in the NEM compare? On a macro level, the electricity markets of California and Australia are of similar scale with total generating capacity of 80GW and 55GW (including 5GW of rooftop solar) and annual demand around 195TWh and 194TWh (2015), respectively [4,9].
Figure 2: NEM Intraday Net Load over time (LHS) and with confidence envelope (RHS)
Figure 2 sets out NEM net load calculated using dispatch data subtracting for scheduled and non-scheduled wind and solar generation sourced from AEMO MMS datasite. The left panel illustrates yearly data, and the right panel shows 2015 data but with a 95% confidence interval. September was chosen as an observation month given higher renewable penetration, but lower overall demand indicative of times when the NEM is likely to be under most strain. Data for other months were also observed and assessed.
Key points of note include:
- Increasing intra-day variability and ramp requirements over time consistent with the ‘duck pattern’ observed in California (morning shoulder period with evening peak).
- The scale of the evening ramp in the NEM is less steep with ~4GW required across 3 hours relative to 9-10GW in CAISO.
- Off-peak (late night / early morning) load in CAISO remains significant and is higher than afternoon load. By contrast, load in the NEM tends to reduce in the off-peak hours, mitigating overnight generation requirements but adding a significant morning rampup of around 5GW. In addition, there is a late-morning rampdown of 2GW currently, which will likely increase with greater renewable penetration.
Intraday Net Load by NEM Region
Figure 3 below illustrates intraday load by NEM regions. The key observations are:
- Significant net-load variability in SA, and at times negative net demand (potentially leading to export via the Heywood interconnector)
- Significant ramp requirements in SA to almost 2x afternoon demand (750MW to 1250MW).
- TAS load has a ‘camel-like’ load pattern with two peak periods.
- Increasing ‘duck-life’ load patterns over time in VIC and NSW.
- Notable upwards demand shifts in QLD in 2015 (potentially indicative of the startup of LNG facilities), with high shoulder-period variability.
Figure 3: NEM Region Net Load over time (LHS) and with confidence envelope (RHS)
The observations above highlight key challenges in managing intermittency and intra-day load balancing. While individual ramp periods in the NEM are not as severe as in CAISO, there is an additional and notable morning ramp-up period followed by an rampdown (the latter of which is likely to increase with more renewables in the grid).
Flexible resource availability is a key area of focus of regulators. For regions such as South Australia and Tasmania, this issue is a critical one given the risk of islanding. Given increased wind development, AEMO has already highlighted the emerging need for frequency control ancilliary services (“FCAS”) in South Australia notwithstanding an ongoing upgrade to the capacity of the Heywood interconnector . As a broad comparison, CAISO currently has 11.6GW of OCGT capacity, relative to 6.5GW within the NEM – the bulk of which are on the eastern regions of the NEM (with only 900MW in SA) [4,9].
CAISO has also introduced market measures to incent flexible capacity . March 2015 CAISO incorporated a constraint into their 15 minute market dispatch optimization process to specifically cater for flexible capacity in local reliability areas. Such measures could be considered for local markets however a key point of difference relates to the electricity market design between the two regions. CAISO operates under a locational marginal pricing (“LMP”) or nodal model with over 3000 individual electric nodes with prices determined at each node. This provides price signals and transparency into local transmission congestion – allowing planners to develop capacity based on regional constraints. Nevertheless, given potential inter-regional congestion in the NEM and regional islanding risk such alternatives may still be relevant. This may also open up markets for energy storage technologies such as batteries or pumped hydro energy storage (“PHES”). Furthermore, the California Public Utilities Commission has also adopted an electricity storage procurement framework for major load-serving entities to source 1.3GW of storage by 2020.
With increasing solar and wind generation capacity, the issue of system balancing will become an ever-challenging issue particularly given the reliance in the NEM on interconnectors, and the disparity in generation sources between states. While the NEM does not currently have ramping requirements and the balancing needs of the scale seen in California, it is likely that with increasing renewable penetration, intraday load variances will exacerbate over time leading to an Australian ‘duck’ curve. In order to ensure reliable and efficient operation of the NEM in the future, regulators need to be looking to mitigate the issue today. The Californian experience provides a useful case study of an increasingly renewable grid and provides lessons (both positive and negative) for the future operation of the NEM.
Notes on Methodology:
Over 20,000 observations of NEM net load data and 800,000 observations of (scheduled and non-scheduled) generation data was sourced from AEMO MMS datasite (http://www.nemweb.com.au). The data and outputs were assessed and analysed using R (v3.2.2). Source code is available from this website.
Generator identifiers and locations are sourced from the latest NEM Registration and Exemption List (using data for Scheduled, Semi-Scheduled and Non-Scheduled generators). While Rooftop PV is not specifically specified but is incorporated into AEMO data (which calculates demand as net of self-supplied generation).
In order to create a load envelope (based on a 95% or 2 standard deviation confidence limit, the underlying assumption of a Gaussian (normal) distribution for net-load is assumed. This assumption for NEM load has not been tested for this piece of work. See  for further analysis of demand patterns and trends.
A daily average across the month is used to account for intra-month weather variations. Mean high and low temperatures for September for NEM region capital cities are broadly consistent across 2013-2015 (varying by 2-3°C). We note however that many factors may affect net load including generation availability, forced and unforced outages and climatic factors.
This article represents my own views and any errors and omissions remain mine alone.
 AEMO, Electricity Statement of Opportunities, (2015)
 AEMO, National Electricity Forecasting Report (2016)
 AEMO/Electranet, Update to Renewable Energy Integration in South Australia (2016)
 AEMO website (www.aemo.com.au)
 Australian Energy Regulator, State of the Energy Market (2015)
 Bureau of Meteorology website (www.bom.gov.au)
 CAISO, Annual Report on Market Issues & Performance (2015)
 CAISO, Flexible Resources Help Renewables – Fast Facts (2016).
 California Energy Almanac (http://energyalmanac.ca.gov/)
 Dannecker L, Energy Time Series Forecasting: Efficient and Accurate Forecasting of Evolving Time Series from the Energy Domain (2015)
 International Energy Agency, Harnessing Variable Renewables: A Guide to the Balancing Challenge (2011)
 Fowlie M, The Duck has Landed (2016). Accessed from https://energyathaas.wordpress.com/2016/05/02/the-duck-has-landed/
 IEA, Key World Energy Statistics (2015)
 Denholm P, O’Connell M, Brinkman G, Jorgenson J, National Renewable Energy Laboratory, Overgeneration from Solar Energy in California: A Field Guide to the Duck Chart (2015)
 IEA PVPS, 2015 Snapshot of Global Photovoltaic Markets (2016)
 Global Wind Energy Council, Global Wind Report 2015 – Annual Market Update (2016)