Datasets:

SouthernCrossAI
/

Arxiv

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 132k rows

Filename stringlengths 22 64	Paragraph stringlengths 8 5.57k
Processed_Unusual_shock-excited_OH_maser_emission_in_a_young.txt	known, although it may be slightly more evolved than other sources showing these masers. We detect Zeeman splitting in the 1720 MHz transition at two epochs, which suggests that the 1720 MHz OH masers are formed in a magnetized environment. These 1720 MHz OH masers in PN IRAS 16333−4807 may trace short-lived equatorial ejections around the formation time of the PN.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	A Case for Change In 1863, a single arc lamp on Observatory Hill in Sydney was lit to celebrate the marriage of Prince Albert of Wales and Princess Alexandra of Denmark. It was the first use of electricity anywhere in Australia.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	It took 25 years from this first light to the 1888 establishment of Australia’s first permanent 240 V electrical grid in the small country town of Tamworth, New South Wales. Its two 18 kW DC coal-fired generators were supplied by the plentiful Gunnedah black coal basin nearby. In the same year, on the other side of the continent, C.J. Otte supplied electricity to the Western Australian Government House with a small 15 kW dynamo. By 1899, a full three-phase 240 V AC grid had been built on the east coast, establishing the foundation of the future power system across the country.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Then, as today, synchronous coal generators provided the majority of system services to maintain security and reliability. Those services included the inertia to maintain stable frequency, system strength to maintain stable voltage waveforms, and energy reserves to maintain the balance of supply and demand even with changing demand and unexpected contingency events.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Under this arrangement, the supply of these services has been conveniently tied to the supply of electrical energy with synchronous generators providing support simply by being synchronized with the electric grid. For over a century, as the electricity infrastructure and trading systems grew, no separate mechanisms were developed to manage these “ancillary services” to the power systemidll. Instead, grid connection standards implicitly regulated an equitable division of costs among facilities in rough proportion to their size. Operators could recover these “costs of doing business” as part of their energy revenue.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	average of 600 W installed per person, growing at 250 W per person per year)–the highest rate of PV uptake in the world. At times, over 100% maximum instantaneous solar and wind penetration is achieved in some regions.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Solar and wind generators connect to the AC grid via power-electronics-based inverters, which do not provide traditional system services by default. This means that while inverter-based resources can replace the energy previously provided by synchronous coal and gas generation, the provision of system services is not replaced in proportion.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The remaining fleet of synchronous resources then faces a growing burden of providing system support services, such as frequency and voltage control and spinning reserves, while revenues simultaneously fall with electricity prices and reduced market share or energy generation. Left unchecked, this dynamic undermines the implicit stability that has historically supported the electricity system.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	In Australia’s National Electricity Market (NEM), this has manifested in a 10-fold increase over the past five years in the number of occasions the system operator has had to intervene outside of normal market operations to maintain security and reliability (Figure 1a). There has been a significant reduction in frequency control performance since 2007 (Figure 1b) due to reduced provision of primary frequency control. Uncertainty and variability in net demand from increasing renewable penetration are expected to triple in the NEM over the coming five years (Figure 1c) as solar and wind are projected to regularly meet 100% of demand (Figure 1d).
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	As the generation mix has changed, a handful of events have catalyzed political interest and action. After a September 2016 state-wide blackout in South Australia, the Australian government commissioned a “Review of the Future Security of the National Electricity Market” by Australia’s chief scientist Alan Finkel. This led to the establishment of an overarching Energy Security Board (ESB) to implement a “long-term, fit-for-purpose market framework” to deliver a “secure, reliable and lower emissions electricity system at least cost” in the NEM.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	A key workstream of this reform program is to establish new markets and mechanisms for providing system support services. These were traditionally called ‘ancillary’ services but are now increasingly being referred to as essential system services (ESS) in recognition of their changing value in grids with low levels of synchronous generation.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	There is a growing consensus that without market reform, the market operator’s remit to “keep the lights on” will likely be accompanied by increased curtailment of renewables and increased complexity of operation. This trend is already being observed. In 2019-20, renewables were curtailed on average 7% of the year in the NEM due to ancillary service requirements, and operator interventions were in place more than 10% of the year.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	This article presents the Australian approach to the challenge of providing ESS in grids with a very high penetration of renewables, outlining first the physical and regulatory contexts of two comparative systems and markets – the NEM (with five regions across Australia’s eastern and southern states), and Western Australia’s wholesale electricity market (WEM) – and their concurrent programs of reform.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The changing nature of ESS in both markets, principles of market design, the spectrum of opportunity for procurement of various new services, and the integration and congruency challenges of addressing them holistically are then discussed. Finally, we present the Australian pathway of reform and a vision for the future of ESS with the hope that it may prove helpful for other countries across the world on similar decarbonization pathways.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Context Australia’s electricity networks span vast distances across a continent roughly the area of the United States, but with less than a 10th of the population.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The energy industry, historically government-owned, was deregulated in the 1990s to disaggregate the vertically integrated state utilities and support competition. This allowed cross-border trading of electricity between states and territories.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The NEM was established in 1998. The isolated nature of the Western Australia and Northern Territory electricity systems was a significant barrier to continent-wide integration of both infrastructure and policy. It was only in 2006 that Western Australia’s WEM was established, covering the southwest region of the state and serviced by the South West Interconnected System, spanning an area roughly the size of the United Kingdom. The interconnected NEM power system is serviced by approximately 40,000 km of transmission network. The islanded South West system integrates approximately 7,800 km of transmission network.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The Australian Energy Markets Arising from a period of widescale deregulation, the NEM was established with a strong commitment to market efficiency with real-time five-minute dispatch intervals, no day-ahead capacity markets, and very high market-price caps (currently $1,000/MWh to $15,000/MWh). In 2021, the settlement time will reduce from 30 minutes to five minutes to align with dispatch, further sharpening market efficiency in the continuous matching of electricity supply and demand. Along with providing efficient incentives for participants, real-time price mechanisms also allow the possibility of contracts for difference and hedging to support long-term contracts and risk mitigation. This is extensively conducted in the NEM through hedging and swap contracts.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	NEM markets for system services were set up with a similar commitment to real-time pricing (for those services that were remunerated). The NEM has six frequency response markets (frequency control ancillary services): contingency frequency response raise and lower services for each of six seconds, 60 seconds, and five minutes response times, and an additional causer- pays primary frequency response service. There are additional non-market services for network support and control (such as transient oscillation control) and system restart.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Reflective of its smaller and more concentrated nature, the WEM balances market efficiency with greater structured procurement, including a capacity market (the Reserve Capacity Mechanism), a day-ahead energy market, the short-term energy market, and a real-time energy market with 30-minute dispatch (with lower market-price caps, currently $1,000/MWh to $382/MWh).
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	For system services, the WEM prioritizes structured procurement via a regulation market (load- following ancillary services) and other system services procured under contract, including frequency response (spinning reserve) and, like the NEM, similar non-market services for network control and system restart.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The Post-2025 Program In 2019, the Australian federal and state and territory governments asked the ESB to advise on a long-term, fit-for-purpose market design for the NEM that could apply from 2025 in response to the profound energy transformation occurring across the country.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Industry and customer stakeholders have been extensively involved and consulted, and there is broad recognition that the individual workstreams are intrinsically interrelated and must be considered together for a coherent whole design.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	There is a wide range of views about each of the workstreams, but responses indicate the reform of system service provision is the highest priority and most urgent. Such reform needs to occur before 2025 to address tighter frequency control, structured procurement for synchronous generation commitment (for system strength and inertia) potentially combined with an ahead mechanism to support scheduling, and the exploration of possible operating reserve and inertia spot markets.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	WA 2022 Program In 2019, the Western Australia government formed the energy transformation task force charged with making clear policy decisions with robust consultation to ensure coherent reform for a full overhaul of the market regulatory framework to go live in 2022. The task force has an explicit focus on the assessment and redevelopment of a new ESS framework.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	All power systems require a suite of system services, traditionally known as ancillary services but increasingly referred to as “essential system services (ESS),” necessary for the secure and reliable operation of the system. Services can often perform the same function but vary in their names, implementation, competitiveness, and remuneration mechanisms across jurisdictions. Below is a summary of the various services that exist in Australia with their WEM and NEM implementations.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Possible new Fast Frequency Response (<2s) from 2022. Slow Raise/Lower (60s) and Delayed Raise/Lower (5m).
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Possible scheduling of synchronous resources through a Unit Commitment for Security mechanism or Synchronous Services Market.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Possible scheduling of synchronous resources through a Unit Commitment for Security Mechanism or Synchronous Services Market. No explicit service except for Reliability and Emergency Reserve Trader function.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Possible new market for Operating Reserves or Ramping Availability in the NEM Table 1: A summary of various ESS in Australia and their implementations within the WEM and the NEM.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Principles for Procurement and Market Design For the impending challenge of redesigning procurement frameworks for ESS, it helps to first consider broad principles of market design alongside the intrinsic valuation of power system security. The objective of procurement frameworks should be to create efficient and effective economic mechanisms to deliver operational requirements.. The operational requirements of power system security must focus on the management of the underlying physics of an electrical network with sufficient redundancy and robustness in the face of uncertainty and risk.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Market Design for ESS A recent report compiled by FTI Consulting for the ESB highlighted seven principles important to the design of effective procurement frameworks for ESS (see Figure 2 Principles of market design for essential system services. Source: Adapted from FTI Consulting Report to the ESB, 2020.).
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Figure 2 Principles of market design for essential system services. Source: Adapted from FTI Consulting Report to the ESB, 2020.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	These principles are fundamental in framing the design problem from a regulatory and market perspective. Alongside these principles is recognition that any design process necessarily involves a compromise between elements to achieve an overall workable design.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	In particular, there is a natural tension between the idealized theoretical design of markets with assumptions of economically rational behaviour and the physical reality of operation which is complex, uncertain, non-linear and failure prone. There are additional asymmetric costs of market efficiency and market failure. While designers may prefer complex, multi-layered, and co-optimized markets, operators may prefer conservative, expensive, and un-optimized solutions. Striking the right balance to develop efficient and robust economic solutions to technical challenges requires the rigorous and combined efforts of power system engineers and economists.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Policy makers have a variety of regulatory and market instruments available to them. Options include technical standards or licenses, operational directions or interventions, regulatory delegations (such as to network monopolies or other central agencies), individual contracts with providers, ESS auctions, and tenders and short-term spot markets.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	compete technological innovation. Market solutions can also optimize against the technical specification of a service, creating a lack of resilience.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	A case in point is the design of contingency frequency response markets in the NEM, where technical specifications guided by normal operating frequency bounds enabled wide frequency governor dead bands. In the face of uncertainty, this led to poor frequency performance and system fragility, only recently corrected by re-implementation of stringent mandatory primary frequency response requirements. By contrast, the WEM complements a spot market for regulation services with an obligatory droop requirement, which has led to improvements in frequency management.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Trade-offs abound for investment considerations given commercial risk appetite. While spot markets, if appropriately designed, can provide efficient scarcity price signals, investment decisions on long-duration assets are typically made in the context of longer-term revenue and cash flow visibility.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Framework flexibility is needed in managing current principles of provision (such as from synchronous generators and synchronous condensers) while accommodating future innovation (inverters providing “synthetic inertia” and grid-forming capability).
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The complexity of co-optimization in the context of uncertainty. • The challenge of valuing ESS and the consequent difficulty of allowing procurement quantities beyond minimum levels to provide additional robustness or resilience. • The trade-offs of operational complexity and market sophistication: Complex markets create more points of failure.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	During this period of rapid change, adaptive governance and procurement approaches are helpful. For ESS, a flexible contractual framework would support the operator to mitigate fast- evolving system risks, potentially accompanied with an adaptive regulatory change process that supports participant decision-making.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Other International Approaches While Australia’s power system finds itself in uncharted territory in the penetration of variable renewable energy (VRE) and distributed solar, there are pioneering advances in market design for system services being explored concurrently across the world. This section reviews some key developments in comparable systems in the United Kingdom and United States.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	In the United Kingdom, electricity systems operation and procurement of system services are delegated to National Grid Electricity System Operator (NGESO), a subsidiary of the for-profit private National Grid UK, which also owns and operates the transmission network. This framework provides a comparatively high degree of flexibility in the approach to procurement with NGESO utilizing competitive tenders of varying duration and structure in procuring services.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Standardized system-wide frequency and reserve products have contributed to shorter-term frequent contract auctions, while more individually tailored and longer-term contracts were used to secure requisite investment for services with locational requirements and smaller provider pools. A recent initiative is the Stability Pathfinder tender, which procured a combination of services including fault-levels and inertia. Reactive power, traditionally an obligatory service, is also increasingly procured through competitive tender approaches.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	NGESO is subject to a unique financial incentive scheme with payments based on performance evaluated by the regulator Office of Gas and Electricity Markets (Ofgem) through an annual score card assessment. The discretion provided to Ofgem has been particularly useful in a rapid-change environment, providing flexibility to respond to evolving technical scarcities and to modify and adapt procurement on an ongoing basis. This has also left NGESO to deal with the issue of supporting investment by initially procuring newer services via longer-term contracts (to underpin investment), moving towards shorter-term auctions as business models become established.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	By contrast, regulatory regimes in the United States and Ontario, Canada, have delegated system services to independent system operators (ISOs), which are not-for-profit entities with relatively less discretion to make decisions on ESS procurement. Procurement approaches tend to be codified in regulations with changes subject to detailed review, stakeholder engagement, market participant vote, and, in some cases, approval of the Federal Energy Regulatory Commission.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Given the need for transparency, ESS have tended to be procured via either short-term spot markets (predominantly frequency and reserve products) or mandatory provision. Spot markets have provided transparency and visibility of pricing, allowing financial markets to develop around services underpinning investment. However, given regulatory structures, incentive mechanisms for U.S. ISOs have proven to be challenging due to narrow incentive thresholds and forecasted delivery targets. In practice, these challenges, combined with the regulatory processes, have limited the ability of ISOs to develop new products expediently.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	While many jurisdictions are adapting technical standards for inverters, there has been less emphasis to date in international jurisdictions on service procurement concerning system strength.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The meshed nature of North American grids means system strength and the provision of fault current is of less concern from a technical perspective, and as a result, it is not explicitly defined as a system service for many regions, including the New York ISO, Mid-Continent ISO, and Ontario Independent Electricity System Operator.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Australian market designs have strong parallels with security-constrained gross power pool models common across markets in North America, apart from procedures for centralized unit commitment and two-settlement market clearings which are not part of the NEM. However, given the extent of VRE penetration and the unique operational phenomena observed in Australian grids, the grids will likely have to forge novel approaches to procure these complex and multi-faceted technical services. These approaches will also have to work alongside the broader challenge of a 5-minute spot market framework without ahead or capacity markets.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Spectrum of Opportunity Procurement Frameworks Having identified a case for change and reviewed the principles of market design for ESS, the challenge then progresses to canvassing the “spectrum of opportunity” in resolving the missing services that arise as inverter-based resources (IBR) replace synchronous generators.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Although there is a preference for real-time signaling, not all system services are suited for market-based procurement. The assessment of market design for each service includes factors such as the measurability/fungibility of the product, competition and co-optimization scope, complexity and simplicity, and locationality.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	This section introduces various options for market design for each stream of ESS under consideration, namely operating reserves, frequency management, synchronous services, and inertia.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Figure 3 A spectrum of opportunity for ESS in the NEM and WEM, indicating the status of current ESS market mechanisms, with an implicit axis of “greater market efficiency” towards the right. Source: Adapted from FTI Consulting Report to the ESB, 2020.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Operating Reserves Energy markets must maintain supply and demand in instantaneous balance with prices set through a spot market. Market participants often have separate contracts across their portfolio to manage the risk around the energy spot price. The market operator, however, typically does not see these contracts. Instead, it must rely on faith in participants displaying economically rational behaviour and taking advantage of high prices at times of supply scarcity.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	This faith is increasingly being tested by the changing nature of generation, with the “invisibility” of behind-the-meter distributed PV generation, and the variability and uncertainty of large-scale wind and solar (Figure 4 a) The probability distribution of 30-minute forecast errors for South Australia, Summer 2019-20, 2-6 pm.; b) probability curve that the 30-minute forecast error is higher than any particular level of reserve, which may inform an efficient reserve demand curve; c) an example 30-minute ramping “availability” product to address unexpected ramps over a 30-minute time horizon, adapted from Brattle Consulting Report to AEMO, 2020.a). The likely result of this uncertainty is the system operator managing the system more conservatively, leading to greater VRE curtailment as risk becomes excessively high. The possible design of an operating reserve or ramping availability service under current consideration may help address this challenge in the NEM.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	There are several market options to procure operating reserves including a) procuring firm “availability” in the dispatch interval 30 minutes ahead (Figure 4 a) The probability distribution of 30-minute forecast errors for South Australia, Summer 2019-20, 2-6 pm.; b) probability curve that the 30-minute forecast error is higher than any particular level of reserve, which may inform an efficient reserve demand curve; c) an example 30-minute ramping “availability” product to address unexpected ramps over a 30-minute time horizon, adapted from Brattle Consulting Report to AEMO, 2020.c); b) holding a certain level of spinning “callable” reserve available to be triggered to dispatch as energy; or c) procuring operating reserve “headroom” in the coming 5 minutes to dispatch as energy. With each option, the use of a demand curve constructed from historical forecast errors may inform the efficient level of reserve or firm availability to procure (Figure 4 a) The probability distribution of 30-minute forecast errors for South Australia, Summer 2019-20, 2-6 pm.; b) probability curve that the 30-minute forecast error is higher than any particular level of reserve, which may inform an efficient reserve demand curve; c) an example 30-minute ramping “availability” product to address unexpected ramps over a 30-minute time horizon, adapted from Brattle Consulting Report to AEMO, 2020.b).
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Figure 4 a) The probability distribution of 30-minute forecast errors for South Australia, Summer 2019-20, 2-6 pm.; b) probability curve that the 30-minute forecast error is higher than any particular level of reserve, which may inform an efficient reserve demand curve; c) an example 30-minute ramping “availability” product to address unexpected ramps over a 30-minute time horizon, adapted from Brattle Consulting Report to AEMO, 2020.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	These options are being developed for possible NEM implementation in the next two years. Decisions on a final preferable new market will be based on trade-offs between operator confidence, market efficiencies, and potential adverse impacts on the energy spot market.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	this “global pool” of resources lends itself to procurement via a centralized, co-optimized spot market (energy dispatch can be considered a very slow class of frequency control). However, the desire for a universal and highly optimized market design must be carefully weighed against the complexities and irregularities this can create in a physical system.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Figure 5 illustrates this consideration through a system operations abstraction of frequency management for contingency response services. In this view, the physical response of the entire generation fleet is aggregated and considered according to different performance requirements for deployment and sustainability of power output into inertial, primary, secondary, and tertiary response. These distinctions are not fundamental, but reflect control structures formed around physical properties and useful trade-offs optimized in the allocation of power system resources.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Figure 5 Top: System frequency appropriately managed after a contingency event; Bottom: Frequency management mechanisms to support the restoration of system frequency following a contingency event.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Exact operating limits vary due to jurisdictional norms and reliability standards, however, these standards ultimately reflect the physical tolerances of an electrical plant. Inverter-based facilities, for example, generally have a higher tolerance to ROCOF than rotating machinery. The ideal procurement model also incorporates incentives to reward tolerances. It reduces overall service requirements in addition to procurement of suppliers. The NEM reform program is reviewing the feasibility of including additional fast frequency contingency response (with response times of less than 1 second) alongside mechanisms to support the efficient provision of (currently mandated) primary frequency response within the normal operating frequency band.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Inertia and System Strength The procurement of synchronous services, namely system strength and inertia, is particularly complex to transition away from their traditional provision as a by-product of generation from synchronous generators. Options to replace this include building additional synchronous resources by the network operator (for example, synchronous condensers with flywheels), or creating incentives for the provision of services through advanced power electronics [See Box: Australia’s Big Battery].
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	There is an opportunity to procure inertia as a separate service, an option being implemented as part of the market reform in Western Australia through the ROCOF Control Service [See Box: WA ROCOF Control Service].
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	In August 2019, the Western Australian energy transformation task force found that a real- time co-optimization of all frequency control services–including inertia–was most appropriate for the future WEM, driven by a mixture of physical, operational, and market considerations.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Historically, the WEM relied on an empirically derived rule of thumb: 70% of the largest generation contingency (in MW) was allocated as headroom across a set of designated facilities. Analysis and comparison of this approach identified that the combination of isolation and relatively small size resulted in the WEM being run close to its technical limits.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The transition to a greater penetration of renewables has necessitated a more sophisticated market design and led to a preference for a real-time spot market to optimize both system inertia and primary response speed. In this context, initial design options focused on the correct balance of service definition “segmentation,” for example, adding a 1-second, 2- second, and 3-second service to complement the co-optimized 6-second and 60-second markets as done in the NEM. With system requirements abstracted to fundamental quantities (i.e., generic MW specifications), the optimal delivery of these services would be by market dynamics, irrespective of the underlying technology.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	Each segment adds complexity and increases the degree of “fantasy” space in which the commercial abstraction diverges from physical reality. In practice, there is no clean, linear separation of MW into convenient buckets.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	machinery has a fundamentally instantaneous response while power electronics suffer from a delay in electronic detection on the same order (<1 s) of the critical ROCOF period.
Processed_Essential_System_Services_in_Grids_Dominated_by_Re.txt	The task force decided a single segment was most appropriate. While multiple segments allow for more service differentiation, in practice these gains were marginal while the downsides were guaranteed.
Processed_Stochastic_Contrastive_Learning.txt	While state-of-the-art contrastive Self-Supervised Learning (SSL) models pro- duce results competitive with their supervised counterparts, they lack the ability to infer latent variables. In contrast, prescribed latent variable (LV) models enable at- tributing uncertainty, inducing task speciﬁc compression, and in general allow for more interpretable representations. In this work, we introduce LV approximations to large scale contrastive SSL models. We demonstrate that this addition improves downstream performance (resulting in 96.42% and 77.49% test top-1 ﬁne-tuned performance on CIFAR10 and ImageNet respectively with a ResNet50) as well as producing highly compressed representations (588× reduction) that are useful for interpretability, classiﬁcation and regression downstream tasks.
Processed_Stochastic_Contrastive_Learning.txt	Learning meaningful representations without human domain knowledge has been a long-standing goal of machine learning. Recent work in large scale SSL (Chen et al., 2020a,b; Alayrac et al., 2020; Grill et al., 2020; Caron et al., 2020; Zbontar et al., 2021; Caron et al., 2021) has advanced this pursuit and narrowed the gap against fully supervised models, all the while relaxing the use of potentially biased human labels. And yet, the SSL methods and toolkits lack a method to add inter- pretable, prescribed distributions into the representation learning process. In this work, we address this shortcoming through the introduction of Bernoulli and Isotropic-Gaussian latent variables into the SimCLR (Chen et al., 2020a) contrastive learning framework.
Processed_Stochastic_Contrastive_Learning.txt	The use of Bernoulli latent variables enables extracting meaningful discrete representations of image data, providing a natural means of data dependent compression2 that is useful for downstream tasks such as classiﬁcation and regression. Interestingly, we ﬁnd that the use of discrete latent variables improves downstream performance when fully ﬁnetuning the representation learning backbone, out- performing SimCLR (Chen et al., 2020a) on CIFAR10 and ImageNet1000 (Deng et al., 2009).
Processed_Stochastic_Contrastive_Learning.txt	The similarity operator (sim) from Equation 1 typically is modeled with a cosine-similarity on the representation feature vectors {v𝑖, v𝑗 }, with a controllable temperature hyper-parameter 𝜏.
Processed_Stochastic_Contrastive_Learning.txt	Figure 1: StochCon without multi-crop (Caron et al., 2020) using plate notation. Following Chen et al. (2020a), two data augmentation operators 𝑡, 𝑡 (cid:48) are sampled from the augmentation family T, producing views ˜x, ˜x(cid:48). A backbone network 𝑓 is applied to each example, producing h, h(cid:48) ∈ R𝐷Backbone (we include spatial pooling in our deﬁnion of 𝑓 ). On the bottom branch, the h(cid:48) speciﬁes the natural parameters of a distribution 𝑞(z; 𝜋 (h)), via an encoder 𝜋 : R𝐷Backbone → R𝐷Latent. Variates z(cid:48) ∼ 𝑞 are decoded by 𝜌 : R𝐷Latent → R𝐷Backbone and compared using the standard LInfoNCE (Equation 1) using the head 𝑔. This work’s contribution is the stochastic mapping φ(cid:48) ↦→ z(cid:48) ↦→ h(cid:48)(cid:48). Gray circles denote random variables. Bottom branch negative variates are reparameterized in the same manner.
Processed_Stochastic_Contrastive_Learning.txt	We describe our model in Algorithm 1 and Figure 1. We modify SimCLR (Chen et al., 2020a) by forcing bottom branch variates through a pathwise differentiable (Mohamed et al., 2020) dis- tribution, z(cid:48) ∼ 𝑞θ (z\|x). Importantly, z(cid:48) can be optionally projected to a lower dimensional space, \|z(cid:48)\| << \|h\|, through linear projection layers, {𝜋θ, 𝜌θ}. Upon ablation, we observe minimal perfor- mance degradation when projecting one branch of the SimCLR model through a low dimensional distribution, with the advantage of having more interpretable features (Section 4.1).
Processed_Stochastic_Contrastive_Learning.txt	In this work, we explore the isotropic-Gaussian (Kingma & Welling, 2014) and Gumbel-Bernoulli (Jang et al., 2017; Maddison et al., 2017) distributions. We apply the differentiable distribution on the output of the backbone model 𝑓θ, given an optional bottleneck projection 𝜋θ.
Processed_Stochastic_Contrastive_Learning.txt	variants. For the Gumbel-Bernouilli distribution, the temperature is annealed from 1.0 → 0.1 using a single cycle cosine schedule during training. Finetuning procedure is described in Appendix A.3.
Processed_Stochastic_Contrastive_Learning.txt	Model performance for linear-probes on a non-updated (Frozen) and ﬁne-tuned (Fine-Tuned) back- bone is given in Table 1. We observe that StochCon (Fine-Tuned) outperforms an equally tuned SimCLR model, as well as a supervised model with the same ResNet50 and ResNet200 (He et al., 2016) architectures, while the Frozen probe is competitive. We validate that this performance differ- ence does not arise purely from the Gumbel-Bernoulli through ablations presented in Appendix A.2.
Processed_Stochastic_Contrastive_Learning.txt	To evaluate the beneﬁts of our StochCon framework, we propose a series of ablations. In Figure 2-Left, we train StochCon Bernoulli models with varying bottleneck φ(cid:48) dimensions and present the top-1 Frozen performance of each model. Results show that StochCon is robust to variadic sized latents. We believe this robustness is due to the model learning to compare a full h ∈ R2048 dimensional vector to an upsampled low dimensional latent, z(cid:48) ∈ R𝐷, 𝐷 << 2048. In Figure 2-Right, we analyze the mean 𝐹1 performance for a multi-class Random Forest evaluated by varying the number of feature units. Surprisingly, we ﬁnd that to accurately classify CIFAR10, the StochCon model with a 64 dimensional latent Gumbel-Bernoulli only requires 11 binary feature units. We also observe that performance decreases for the Isotropic-Gaussian as we increase the latent dimensionality, holding the number of Random Forest units constant. Note that this does not happen in the Gumbel-Bernoulli case, as this variable does form distributed representations in the way an Isotropic-Gaussian does.
Processed_Stochastic_Contrastive_Learning.txt	Figure 2: Left: Ablation over bottleneck latent dimensions for Bernoulli-StochCon. Right: Mean 𝐹1 performance across all classes on CIFAR10 for a multi-class Random Forest, varying the number of feature units. Units are identiﬁed on the training set using Random Forest feature importance under 5-fold stratiﬁed sampling. Performance is the mean over the held-out test sets.
Processed_Stochastic_Contrastive_Learning.txt	ially collapse to 0. To work around this and provide meaningful uncertainties, we force the network to learn to estimate variances of the opposing set of views, so that z(cid:48) ∼ N (cid:0)µ(𝜋 (h(cid:48))), σ2(𝜋 (h))(cid:1). We validate this below in Figure 3 and ﬁnd that as the bottleneck dimension reduces, the model learns to rely more on the available stochasticity.
Processed_Stochastic_Contrastive_Learning.txt	Figure 3: CIFAR10 test top-1 (Left) and test aggregate variance, Var𝑥 [𝑞θ (z\|x)], (Right) for varying isotropic-Gaussian bottleneck dimensions.
Processed_Stochastic_Contrastive_Learning.txt	Since StochCon works with discrete representation vectors, it enables analysis through countable metrics. We present the average count of representation bits for the R2048 dimensional StochCon Bernoulli model in Figure 4. We ablate four different variants: {hard bottom, hard top, soft bottom, soft top}. The difference between these variants is where the distribution is applied: the top-* models apply the reparameterization on the global image views (Caron et al., 2020), while the bottom-* models apply them on the local views (Caron et al., 2020). The hard-* models use a differentiable mechanism to always feed-forward discrete variates (see Appendix Section A.1), while the soft-* models use the standard variates extracted from the Gumbel-Bernoulli distribution.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	The observations and simulations have revealed that large-scale magnetic ﬁeld and outﬂows can exist in the inner regions of an advection-dominated accretion disc where the resistive diﬀusion may also be important. In the present paper, the roles of large-scale magnetic ﬁeld and outﬂows in the structure of resistive advection-dominated accretion discs are explored by assuming that the accretion ﬂow is radially self-similar. In the non-ideal magnetohydrodynamic (MHD) approximation, the results show that the angular velocity is always sub-Keplerian when both the outﬂow and the large-scale magnetic ﬁeld are taken into account. A stronger toroidal ﬁeld component leads to faster rotation, while the disc rotates with faster rate if the vertical ﬁeld component is weaker. The increase of magnetic diﬀusivity causes the infall velocity to be close to Keplerian velocity. Although the previous studies in the ideal MHD approximation have shown that the disc temperature decreases due to the vertical ﬁeld component, we ﬁnd that the eﬀect of vertical ﬁeld component on the temperature of a resistive disc depends on the magnetic diﬀusivity. When the magnetic diﬀusivity is high, the more eﬃcient mechanism for decreasing the disc temperature can be the outﬂows, and not the large-scale magnetic ﬁeld. In such a limit of the magnetic diﬀusivity, the components of the large-scale magnetic ﬁeld enhance the gas temperature. The increase of temperature can lead to heating and acceleration of the electrons and help us to explain the origin of phenomena such as the ﬂares in Sgr A*. On the other hand, the infall velocity in such a limit rises as the temperature increases, and therefore the surface density falls to too low values. Any change in the density proﬁle can alter the structure and the emitted spectrum of disc.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	low-luminosity AGNs (Park et al., 2019; Younes et al., 2019), narrow-line Seyfert 1 galaxies (Haba et al., 2008; Meyer-Hofmeister & Meyer, 2011), and Sgr A* (Yuan et al., 2002; Yuan & Narayan, 2014).
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	ical solution demonstrated that the radial-dependence of the accretion rate is as a power-law function. Their power law index has been assumed to be in the range from 0 to 1. The numerical simulations have also shown that the accretion rate in the presence of outﬂows de- creases as such a power-law function (e.g. Ohsuga et al., 2005; Yuan et al., 2012b,a; Bu et al., 2013; Yang et al., 2014; Bu & Gan, 2018). Although recent theoretical studies showed that the power index can lie between 0.4 to 1.0 (Yuan et al., 2012a; Bu et al., 2013; Yang et al., 2014), the observations of Sgr A* combined with the radiatively ineﬃcient accretion model have predicted a range 0.3 < s < 0.4 (Yuan et al., 2003).
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	Diﬀerent mechanisms have been suggested that drive the outﬂows (e.g. Chelouche & Netzer, 2005; Ohsuga & Mineshige, 2011; Cao, 2014; Yuan et al., 2015; Hashizume et al., 2015). One of them can be large-scale magnetic ﬁelds (Blandford & Payne, 1982; Igumenshchev et al., 2003; Cao, 2014; Li & Cao, 2019). The magnetic ﬁelds by transferring the angular mo- mentum of a magnetized disc due to the magnetic braking (Stone & Norman, 1994) or magnetorotational instability (MRI) (Balbus & Hawley, 1998) can also modify the disc structure. Several attempts to ex- amine the roles of outﬂows and large-scale magnetic ﬁelds in ADAFs have been made over the last sev- eral years. When large-scale magnetic ﬁelds with only toroidal component are considered, self-similar solu- tions indicate that this component of magnetic ﬁeld rises the gas temperature of an ADAF with outﬂow (e.g. Abbassi et al., 2010; Ghasemnezhad & Abbassi, 2016; Bu et al., 2019; Ghoreyshi & Shadmehri, 2020). Such high temperatures do not match values obtained from ﬁtting the observational data of Seyfert galaxies (see uanYUAN & Zdziarski, 2004). Thus, the eﬀect of other components of the ﬁeld on the structure of ADAFs must be investigated. The inﬂuence of poloidal ﬁeld component in an ADAF in the absence of outﬂows has been examined by Zhang & Dai (2008). Adopting the ideal magnetohydrodynamic (MHD) approximation, they concluded that the poloidal ﬁeld component can reduce the temperature (see also Ghasemnezhad, 2017). But Bu et al. (2009) showed that the role of poloidal ﬁeld component in the temperature reduction can be more signiﬁcant when the outﬂows are considered (see also Mosallanezhad et al., 2013).
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	Simon & Hawley, 2002a,b; Masada & Sano, 2009; Bai & Stone, 2013; Rodgers-Lee et al., 2016), and also the rate of angular momentum trans- port (Ziegler & Rüdiger, 2001; Fromang et al., 2007; systems. The Longaretti & Lesur, 2010) importance of resistivity in other systems like the discs around Kerr black holes (Kudoh & Kaburaki, 1996; Qian et al., 2017, 2018; Shaghaghian, 2020), and even Sgr A* at the Galactic Center (e.g. Melia & Kowalenko, 2001) has also been recognized. In hot accretion ﬂows modeled for Sgr A*, the Coulomb coupling between ions and electrons is not strong enough. Since these ﬂows, such as ADAFs, have the mean free path for Coulomb interactions larger than the typical size of the system, they are eﬀectively collisionless (Quataert, 1998) and the ideal MHD approximation may not be suﬃcient to study their properties. The recent studies have shown that the resistive dissipation in hot accretion ﬂows around black holes can modify the amount of magnetic ﬁeld and the role of magnetic ﬁeld in transferring the angular momentum (Zahra Zeraatgari et al., 2018). the (2018) Zahra Zeraatgari et al. magnetic dissipation heating in regions of disc where outﬂows may be important is comparable to the viscous heating. Hence, the magnetic dissipation heating for increasing the electron temperature of ADAFs may be an eﬀective agent (Bisnovatyi-Kogan & Lovelace, 1997; Ding et al., 2010). On the other hand, the astrophysical outﬂows depend on the strength of magnetic diﬀusivity and type of the ﬁeld component (Kuwabara et al., 2000; Fendt & Čemeljić, 2002; Igumenshchev et al., 2003; Čemeljić et al., 2014; Qian et al., 2018; Vourellis et al., 2019). Thus, one can expect that the structure of is modiﬁed through the presence accretion discs resistive diﬀusion. Faghei & Mollatayefeh (2012) of showed when the resistive accretion ﬂows with outﬂow is investigated, the disc has higher temperature due to magnetic diﬀusivity. Perhaps this result arises from considering a purely toroidal magnetic ﬁelds. Although Abbassi & Mosallanezhad (2012) studied the eﬀect of other ﬁeld components on the dynamics of resistive ADAFs, the role of outﬂows was neglected in their work. In the present paper, we shall extend the work of Faghei & Mollatayefeh (2012) by considering other ﬁeld components. Indeed, our main aim is to investigate the roles of outﬂows and large-scale magnetic ﬁelds in restive ADAFs.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	Here, M∗ is the mass of the central object. In this pa- per, we use deﬁnitions βφ,z = (1/2)(cφ,z/cs)2. For the hot accretion ﬂows, the typical values of βφ and βz may be between 0.01 and 1 (e.g. De Villiers et al., 2003; Beckwith et al., 2008). However, the numerical simula- tions indicate that the accretion ﬂows in the hot accre- tion disc change from a gas pressure-supported state to a magnetically supported state as soon as the thermal instability arises (Machida et al., 2006). Under these conditions, one would expect that βφ and βz exceed unity. Hence, we shall also consider the cases with val- ues greater than 1.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	Here, s represents the outﬂow strength in the disc. Note that s = 0 corresponds to a disc without winds.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	The last term of right hand side in equation (3) rep- resents the angular momentum extracted by the wind where ℓ is the length of the lever arm. In the present pa- per, ℓ is assumed to be equal to 1 describing the case in which outﬂowing material carries away the speciﬁc an- gular momentum it had at the point of ejection (Knigge, 1999). The cases with ℓ < 1 (or > 1) are expected to describe the outﬂows that carry away less (or more) an- gular momentum.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	where VK (= rΩK = GM∗/r) is the Keplerian speed and the eﬃciency factor ζ may be a function of ℓ. The eﬃciency factor for a case with ℓ = 1 is equal to 1 (Knigge, 1999).
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	Here, Pm is the magnetic Prandtl number. Although the magnetic Prandtl number is generally considered to be a constant, its value depends on the disc radius (Balbus & Henri, 2008). The studies show that the mag- netic Prandtl number may lie in the range from 10−3 to 103 in the discs around compact X-ray sources and AGNs (Balbus & Henri, 2008). Note that we use the α- prescription of Shakura & Sunyaev (1973) for the kine- matic viscosity coeﬃcient ν.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	When the large-scale magnetic ﬁeld has only the sim- toroidal component, equation (24) and the of (25)-(28) ilarity solutions Faghei & Mollatayefeh (2012). In the absence of mag- netic diﬀusivity, one can obtain the ﬁndings of Mosallanezhad et al. (2013) without the radial ﬁeld component. By setting βφ = βz = 0, our solutions tend to the solutions presented by Shadmehri (2008) when thermal conduction is ignored.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	In order to examine the eﬀects of outﬂows and large- scale magnetic ﬁeld on the properties of a resistive ADAF, we solve the equation (24) numerically. To in- vestigate the disc properties physically, only real roots 2 must be adopted. We il- corresponding to positive C2 lustrate the disc variables as a function of inverse mag- −1. In this paper, the mag- netic Prandtl number, i.e. Pm netic Prandtl number is assumed to be in the range from 0.1 to 10 (Lesur & Longaretti, 2007). Note that the magnetorotational instability in this range is ex- ited (Käpylä & Korpi, 2011). We also set ˙m = 0.1, f = ℓ = ζ = 1, γ = 4/3, α = 0.1, βφ = βz = 1.0, and s = 0.2 unless otherwise is stated.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	gas temperature. Consequently, the enhancement of the gas temperature can puﬀ up the accretion disc and the disc gets thicker. Noting that the C4−dependence of density in equation (25), we also ﬁnd that the surface density of the disc reaches a maximum value at the high- est adopted value of Pm. In the high-Pm limit, the sur- face density of a disc without outﬂow is more than that in a disc with moderate wind. Indeed, the outﬂows by extracting the disc material can lead to a reduction in the surface density. When the limit of small-Pm is con- sidered, however, the surface density of the disc in the absence of outﬂows becomes less. At Pm = 0.5, for ex- ample, the surface density for a case s = 0.3 is about twice the value in s = 0.0. But Faghei & Mollatayefeh (2012) showed that the net eﬀect of the outﬂow on the surface density is independent of the magnetic Prandtl number. We ﬁnd that such a diﬀerence in results is due to the presence of vertical component of magnetic ﬁeld. The reason is as follows. According to equation (22), the radial velocity depends sensitively on the vertical magnetic ﬁeld and the magnetic Prandtl number (see also the proﬁle of the infall velocity in Fig. 3). The de- pendence of the surface density on the infall velocity (equation (19)) leads to such a trend for the density proﬁle.
Processed_Self-similar_structure_of_resistive_ADAFs_with_out.txt	Figure 1. Proﬁles of the physical variables of the accretion disc versus for diﬀerent values of s, as labeled. It is assumed that and α = 0.1, βφ = βz = 1.0, ζ = ℓ = f = 1.0.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.