security of supply – international review of standards and ... · market facilitation. energy...
TRANSCRIPT
Security of Supply – International Review of Standards and Implementation
IEEE LOLE Working Group Webinar
24th March 2017 Daniel Burke, BE, PhD, M-IEEEEMR Modelling, National Grid UK
Overview Background – National Grid System Operator
International Review of Security of Supply Standards and Implementation
Survey Participants and Contributors
Summary of International Review Findings Metrics for adequacy comparison – the capacity reserve margin
Reliability standards
Value of lost load (VoLL)
Modelling Frameworks and Tools
Consideration of emergency actions and contingency reserve
Capacity remuneration mechanisms
Load demand and variable generation modelling
Interconnection assistance contribution to reliability
Planning timeframe uncertainties and capacity procurement decision making tools
Topical Adequacy Issues in GB at present
Acknowledgements2
3
National Grid Electricity Transmission (NGET) – Who are we?
Scot
Pow
erN
atio
nal
Grid
TO
National G
rid SOSc
ot
Hyd
ro
One System OperatorMultiple GB Transmission Owners
System Design
Project Management
Engineering and Maintenance
System Planning
System Operation
Market Facilitation
Energy Trading
Electricity Market Reform (EMR) Delivery Body
Review Background As EMR Delivery Body, we are required to produce the annual GB “Electricity Capacity Report”, that
recommends the capacity that should be procured in the annual 4-year and 1-year ahead auctions
Great Britain (GB) has a reliability standard of 3 hours LOLE/year- this is set by the UK Government
In Q3 2016, motivated by reducing GB capacity margins and significant generation portfolio change,NGET SO carried out an international review of security of supply standards and implementation
This review took the form of written surveys sent to modelling experts, complemented in many casesby telephone interviews, further supported with available public-domain literature sources
Security of supply investigation in 20 different countries/systems around the world was carried out
A literature review report on the factual learnings from this project was prepared – this slide packsummarises some of the key findings
4
International Review Report
Many of the survey respondents are actively involved with the IEEE LOLE Working group
A copy of the literature review is thus soon to be available on the IEEE LOLE WG public website pending final edits
Information in the report is valid up to late 2016 only –Notable adequacy events e.g. in Australia since then are not reflected
It is grouped in to overall themes across all systems, aswell as individual sections on each system/country
Exhaustive detail is available in the publication, this is simply a summary overview slidepack
We have done our best to understand and summarise the information compiled and provided to us though some errors may (probably) remain - please flag afterwards!
5
Disclaimer to Note!
6
International Review - Systems Included
Consulted - EUROPE Ireland
Portugal
France
Germany
Netherlands
Belgium
Finland
Denmark
Norway
Spain
7
Consulted - USA PJM
North-West PCC
ERCOT ISO Texas
New York ISO
Midcontinent ISO
Southern Power Pool (SPP)
Californian ISO
ISO New England
Consulted – Other Countries Australia
Japan
Figure 2 – The different USA regions surveyed (a modified FERC diagram)Figure 1 – The different European regions surveyed (a modified ENTSO-e diagram)
Consulted – Other Organisations Astrape Consulting
GE Consulting
EPRI USA
University of Edinburgh
Heriot-Watt University
IEEE LOLE WG
Adequacy Comparisons - Capacity Margin Metrics Part of the goal of the international review was to better understand the reliability state of the GB system
when compared to international peers
Capacity reserve/planning margin metric numbers are often reported in different systems, which at first glanceappear to provide a means to compare and contrast reliability states. The GB metric is:
Survey indicates challenge to numerically compare capacity margins reported in external literature directly As the formulaic definitions for margins vary internationally, as do the techniques/assumptions used to populate them
Margin is also based on average or “50/50” peak conditions whereas LOLE is driven by extremes – thus margin comparisonbetween countries of different thermo-sensitivity of peak load (e.g. France vs GB) may not be meaningful
Margins may thus be a useful proxy for risk on a given system with experience over time, but comparisonsacross different systems may be inaccurate – this report stops short of comparing systems
Aside from a few exceptions, most of the systems surveyed seemed to report relatively comfortable reliabilitystates at present, with few medium-term concerns noted In many systems, the reliability standards are but a lower worst case bound, rather than a ‘target’ 8
International Reliability Standards
Many different reliability standard metrics are in place – though some systems have no standard at all
Belgium and Portugal have a reliability standard that is based on two different, but separately bindingmetrics. Belgium seeks to limit the weakness of an “expectation” based standard
LOLE expressed in hours/year is the most common European metric in place – values differ
The most common North American reliability standard (LOLE of 0.1 days/year, i.e. “one day in 10years” with outage events) is a much more binding type of standard than e.g. LOLE of 3 hours/year
The hydro-dominated system NWPCC has an “annual LOLP” standard to mitigate dry year risks9
• LOLE – loss of load expectation• EENS – expected energy not served• LOLP – loss of load probability
Figure 3 – Categorisation of the international reliability standard metrics employed
Value of Lost Load (VoLL) A customer survey derived VoLL parameter estimate (£17,000/MWh) is presently used to directly
set the reliability standard in GB, by virtue of the relationship LOLE in hrs/yr = CONE/VoLL
The “true” value of VoLL is dependent on many things (extent, frequency and duration ofshortages), and will vary from customer to customer
It was notable from the international survey that most systems presently do not have reliabilitystandards explicitly derived from such a direct capacity/unreliability economic cost trade-off
Estimates for VoLL were found to exist in some of the systems surveyed, though they are not usedto directly set the reliability standards - sometimes have been used for network planning
In the US in particular, there appears to be preference for a standard that prioritises reliability as a“good thing” in itself - some skepticism on the acuracy/concept of a VoLL estimate was encountered
Formal procedures for “rotational” load shedding were reported in a few countries, whereby outagesare circulated in rolling manner around the system during a “brown-out”, so that no individualcustomer is impacted for too long 10
Capacity Remuneration Mechanisms (CRM) A variety of capacity remuneration mechanisms exist around the world, however there are none at
all in a significant minority of systems, and strategic reserve only in many others
France and California appear to have de-centralised CRMs with bilateral contracts
Energy-only markets such as Australia and ERCOT regularly re-assess market price caps andscarcity pricing conditions to ensure marginal plants can be remunerated appropriately
California includes special consideration of flexibility/ramping issues in it’s supplier obligations
Ireland is moving from a capacity pot based approach (where all existing capacity receives pro-ratapayment) to an auction approach for a specific capacity requirement, similar to GB
11
CM CentralAuction
CM LSEObligation
CapacityPayments
StrategicReserve
No CRM
Ireland ^ France Portugal* Portugal* Denmark
PJM California Spain Germany NWPCC
NYISO Ireland ^ Netherlands ERCOT
MISO Belgium Australia
ISO-NE Finland Norway
Great Britain Japan SPP (?)
Table 1 – Summary of International Capacity Remuneration Mechanisms
• LSE – Load Serving Entity, supplier obligation
• ^ Ireland is moving from a capacity pot to an auction based system
• * Portugal has a number of capacity remuneration schemes
Load Demand Modelling
A number of different approaches were evident in the survey responses to develop load demand data that was input to the reliability simulation tools
Generally, there seems to be correction for average or “50/50” peak demand level variations in the applied load demand data before projection to future horizon year economic conditions
It seems that the historical recorded system demand data is used directly in a number of regions
In other regions, a common trend was to use a small number of recent annual demand time series shapes to grow out a larger number of synthetic years based on a longer time series of temperature data, and a benchmarked load temperature dependency model
The temperature data was observed to be from historical recorded meteorology databases or else hind cast from the NASA MERRA database or the ENTSO-e Pan European Climate Database
A further variation on this was to use temperature data completely artificially generated from a meso-scale weather/climate model.
12
Variable Wind Generation Modelling
From the survey responses, there can be a distinction made between the source of the wind data used, and how it is thereafter manipulated in the modelling frameworks
In terms of how the wind data is sourced:
o Historical recorded meter data is used directly in the majority of systems, in particular those with a reasonable history of wind farms in existence on their system
o Others again use either meteorological re-analysis approaches to attempt to estimate the original historical wind speed at given geographical locations, or meso-scale numerical models of completely synthetic wind speeds
In terms of how the data is used in the wind modelling framework:
o In a few systems it appears the wind effective load carrying capability was calculated a-priori using offline analysis, and a generator of fixed output equivalent to this value was used in all hours of the subsequent reliability simulation.
o In others the contribution of wind in the model was similarly fixed for each simulation hour using average observations of historical output at time of peak demand
o The majority of the rest used the time series of wind availability endogenously in the models to derive the reliability contribution jointly with all other input parameters of the simulation study.
Where historical recorded data or synthetic/re-analysis data are used, then due care to preserve any natural dependencies between wind power availability and other meteorology driven variables such as load demand was generally observed.
13
Emergency Actions Modelling
Emergency actions (distribution voltage reduction, conventional generation maximisation,emergency interconnection assistance) exist as operator options to mitigate shortage in real-time
The long term reliability modelling approach in GB is to exclude these resources as options tomeet the reliability standard – they are considered non-firm in nature
The vast majority of other systems were found to have a similar philosophy
Only NY-ISO and ISO-NE, and to a lesser extent Denmark, include these resources in theirreliability studies as options to meet the standard before any capacity procurement
14
Emergency Actions
Actions Excluded(GB)Actions Included
Undetermined
Figure 4 – International systems approach to inclusion of emergency actions in reliability modelling
Treatment of Interconnection
Significant diversity was also observed in the survey responses for the treatment of capacity adequacy assistance from neighbouring regions that accrues by means of market coupling and load diversity
In the Netherlands, the reliability standard is defined with respect to supply and demand balance in that country alone, and in the base case analyses no account of interconnection import/export is taken
Some systems appear to only consider those generators physically located externally to their system which have firm contracts to support them, and any additional non-firm assistance appears to be considered insignificant
A rather distinct approach is being used in GB, whereby off-line market modelling is used to develop a probability distribution of interconnection flows at different percentiles of GB load demand, distributions that are subsequently convolved in to the overall supply demand balance thereafter
.
In a few other locations e.g. Ireland and Portugal, interconnection is generally modelled as a two state unit similar to conventional plant, with a maximum capacity chosen by judicious experience, and a forced outage rate based on the reliability of the interconnection asset itself
In the other extreme, there are systems that have fully detailed models of supply and demand for the neighbouring systems directly in their reliability models. France and Belgium are clear examples of this in Europe, where full models of up to 19 nearby countries are included directly in ANTARES model.
15
Modelling Frameworks and Tools
Time sequential modelling seems to predominate amongst the tools in use
Sequential models are necessary for energy limited hydro and demand response in some systems, though many of the reliability standard metrics themselves seem to be time collapsed in nature
Only GB and PJM exclusively still use time collapsed convolution methods, though there are nuanced differences between their approaches in itself
16
0
2
4
6
8
10
12
14
Sequential MonteCarlo
Time Collapsed(similar to GB)
Mixture Undefined
Modelling Framework
Figure 5 – Summary of the reliability modelling frameworks surveyed
Consideration of Planning Uncertainty In the GB Electricity Capacity Report, we have a number of self-contained scenarios and we also
include multiple uncertainty sensitivities to a Base Case High/Low demand growth
Warm/Cold winter demand years
High/Low wind availability at peak demand
Conventional plant forced outage rate parameter variations
Non-Delivery of conventional plant in the CM auction delivery year, in multiple granular steps
Most other international systems perform these kinds of individual sensitivity analyses, as relevant
Some systems in the US (PJM and ISO-NE) perform their capacity procurement assessments on asingle base case only, with sensitivity analyses mainly presented for “information purposes only”
Aside from a few US systems where annual demand growth forecast uncertainty is treated as aprobability distribution, there is no system where specific weightings or priorities are applied to anyof the uncertainty sensitivities
It was important to note that a few systems (Portugal, Australia, Southern Power Pool) appeared toapply reliability studies where multiple individual sensitivities were combined together
Others (Netherlands in particular) carried out an assessment for the reliability of their system only,assuming it was isolated from interconnection imports/exports.
17
Capacity Procurement Decision Making
In the GB Electricity Capacity Report, a formal decision making procedure known as the “LeastWorst Regret” (LWR) is used to account for uncertainty in the optimal capacity to procure
This is essentially a mini-max regret optimisation approach which attempts to build in robustness
No other system appears to be using a formally specified decision making tool of this nature
Ireland is considering introduction of LWR in the new CM there
In Europe (Belgium, Netherlands, Finland, Portugal), different scenarios appear to be presentedseparately to Governments/Regulators, who then subjectively judge on strategic reserverequirements
In some US areas (NWPCC, NY-ISO) regional reliability panels (comprised of policy makers andenergy market stakeholders) iteratively vote to accept which scenario to cover
In other US areas, where only one central base case used in the studies (PJM, ISO-NE) the regionalreliability panel (of similar constitution) votes simply to accept or reject the ISO proposed installedreserve margin level
18
Contingency Reserve Modelling In GB we require additional capacity headroom set aside to cover contingency reserve for the
loss of largest in-feed and thus maintain control of frequency stability
The international system operator approach to this issue was more mixed – about half therespondents have the same approach as GB, others ignore the reserve requirement in studies Of those that ignore it, half of those again are presently re-considering their approach
This issue is non-negotiable in GB as we are a single stand-alone synchronous area wherefrequency control is critical Other systems in Europe/USA are part of wider AC interconnections with significant inertia and thus may arguably
allocate less importance to the issue
19
0
1
2
3
4
5
6
7
8
9
Reserve SetAside in Models(same as GB)
ReserveForegone
Undefined
Reserve Modelling
Considering Reserve SetAside
Figure 6– International systems approach to inclusion of contingency reserve requirement in reliability assessment
Summary Points
Lots of diversity in the reliability standards in the various different systems around the world
Subsequent interpretation and modeling assumptions can also be quite different – the end resultadequacy state of a system is a function of both
There may be reasons for these differences, as no two power systems are exactly the same
Looking at the reliability standards alone, then as a (broad and approximate) generalization, the USbased systems tend to focus priority on physical reliability above cost when compared to theEuropean ones
Comparing rather distinct systems using reported capacity margin definitions that are deriveddifferently, is not a sensible step
Some kind of translation to a common metric using identical study frameworks and model assumptions wouldbe ideal, but such studies on the international scale appear to be rare
There are benefits to surveying and sharing international best-practice amongst peer organisations –keep in touch!
20
Topical Adequacy Modelling Issues in GB
2017/18 review of the GB Reliability Standard by Government
Impact of tighter planning margin on system operational, balancing, reserve, and scarcity pricing costs that may not be accounted for in VoLL and (net-)CONE alone
Better means to account for uncertainty in the planning timeframe – probability scenarios?
Accounting for risk of CM contract default via Non-Delivery cases
Efforts to get better data on distributed generation, and their forced outage rate characteristics
Accounting for energy limited battery storage with an equivalent firm capacity (EFC)
Etc!
21
Acknowledgements
The list of international survey respondents who helped to gather the data
UK Department of Business Energy and Industrial Strategy (BEIS)
Ofgem the GB Electricity Market Regulator
Chris Dent, Amy Wilson, Stan Zachary at Uni Edinburgh and Heriot Watt respectively
IEEE LOLE Working Group members
See you in Chicago!
22