Reinventing the Utility for DERs: A Proposal for a DSO-Centric Retail Electricity Market
Rabab Haider, David D'Achiardi, Venkatesh Venkataramanan, Anurag Srivastava, Anjan Bose, Anuradha M. Annaswamy
RReinventing the Utility for DERs: A Proposal for a DSO-Centric RetailElectricity Market ⋆ Rabab Haider a , ∗ , David D’Achiardi a , Venkatesh Venkataramanan a , Anurag Srivastava b , Anjan Bose b and Anuradha M. Annaswamy a a Massachusetts Institute of Technology, Cambridge, MA, USA b Washington State University, Pullman, WA, USA
A R T I C L E I N F O
Keywords :Distribution Grid Retail MarketHierarchical Market DesignOptimization
A B S T R A C T
The increasing penetration of intermittent renewables, storage devices, and flexible loads is intro-ducing operational challenges in distribution grids. The proper coordination and scheduling of theseresources using a distributed approach is warranted, and can only be achieved through local retailmarkets employing transactive energy schemes. To this end, we propose a distribution-level retailmarket operated by a Distribution System Operator (DSO), which schedules DERs and determinesthe real-time distribution-level Locational Marginal Price (d-LPM). The retail market is built usinga distributed Proximal Atomic Coordination (PAC) algorithm, which solves the optimal power flowmodel while accounting for network physics, rendering locationally and temporally varying d-LMPs.A numerical study of the market structure is carried out via simulations of the IEEE-123 node networkusing data from ISO-NE and Eversource in Massachusetts, US. The market performance is comparedto existing retail practices, including demand response (DR) with no-export rules and net metering.The DSO-centric market increases DER utilization, permits continual market participation for DR,lowers electricity rates for customers, and eliminates the subsidies inherent to net metering programs.The resulting lower revenue stream for the DSO highlights the evolving business model of the modernutility, moving from commoditized markets towards performance-based ratemaking.
1. Introduction
The electricity deregulation movement of the 1990s di-vided the vertically-integrated value chain along the powersystem into generation, transmission, distribution, and elec-tricity markets. Through these regulatory changes, whichinclude the sale of generation assets to third parties or un-regulated subsidiaries, retail sales deregulation, and the re-vision of electricity tariffs, market competition was enabledacross the value stack of the electric system. This led tomore efficient pricing, free entry, free exit, and competitionamongst transmission-level assets that comprise the first twovalue chain buckets. More recently, a similar deregulationmovement facilitated competition within retail power sales,triggering the emergence and growth of competitive retailsuppliers and Community Choice Aggregations (CCAs).Within grid operations, transmission-level assets consist-ing of transmission lines and large-scale generators intercon-nected at high voltages have constituted the backbone of thepower system, with a largely centralized decision and controlarchitecture. However, this is rapidly changing with the in-creased penetration of distributed energy resources (DERs)into the distribution grid, which include demand response(DR), customer-sited and behind-the-meter generation suchas solar photovoltaic (PV), fuel cells, electric vehicles (EVs),storage, and Combined Heat and Power (CHP) generators. A ⋆ This work was supported in part by the Department of Energy underAward Number DE-IA0000025 for UI-ASSIST Project. ∗ Principal corresponding author [email protected] (R. Haider); [email protected] (D. D’Achiardi); [email protected] (V. Venkataramanan); [email protected] (A.Srivastava); [email protected] (A. Bose); [email protected] (A.M. Annaswamy)
ORCID (s): (R. Haider) centralized paradigm may no longer be adequate with suchan increased penetration. Rather, decentralized and distributedapproaches are called for, as the same goals of maintaininga safe, reliable, resilient, and affordable operation have tobe met by the emerging power grid. A distributed paradigmmust be invoked in the economic substrate of the power gridas well, which leads to new retail market mechanisms toefficiently operate assets and support investments within adistribution system. This paper proposes an architecture forsuch a retail market.Neither of the two deregulation exercises, of the vertically-integrated utility or in retail power sales, have fostered com-petition across distribution-level assets, and so we look to-wards retail markets. Discussions of DER-level markets hasbegun in high penetration states, such as New York, Hawaii,and California, but continues to remain in a nascent stage,with limited market-design innovation for the distributiongrid (NASEM, 2021). The design of such markets, the oper-ational changes, and the regulatory requirements are all openquestions, for which a growing body of literature is develop-ing (Nudell et al., 2019; Zinaman et al., 2015; IRENA, 2019;MITEI, 2016). These works look toward the establishmentof a Distribution System Operator (DSO), which is chargedwith not only operating the distribution grid, but with over-seeing the operations of a new retail market and interfacingwith the wholesale market. While DSOs exist in much ofEurope, they are asset-centric companies, charged primarilywith a managerial role within the distribution infrastructure,and in some cases, of telecommunication, gas, and water net-works. Moving forward there is an overwhelming push forDSOs to take on a service-oriented role, that includes over-sight of a retail market with participation from DERs (Vler-
R Haider et al.:
Preprint
Page 1 of 14 a r X i v : . [ ec on . GN ] F e b SO-Centric Retail Market ick Energy Centre, 2020).Distribution-level markets potentially displace agents onall revenue streams associated with the original vertically-integrated utility. In such a market, retail customers woulddemand and sell energy services from distributed genera-tors with time-varying pricing (TVP). For this reason, pre-cursors to distribution markets have taken a wide range offorms to compensate resources along the power system valuechain, including TVP, retail DR, and net metering. Theseexisting retail compensation schemes rely predominantly onstatic or averaged tariff rates which do not incorporate lo-cational or temporal price differentials. As a result, theseschemes fail to provide adequate compensation for DERs,whose grid services are inherently variable in location andtime. TVP-based tariff structures include time-of-use (TOU)rates and Critical Peak Pricing (CPP), and have the abil-ity to provide cost-reflective price signals and enable moreefficient operation of DERs; however, retail customer en-rollment in these programs has been limited. Other retailcompensation schemes typically overcompensate resourcesat the expense of other customers, and higher future retailprices. In the absence of retail markets, wholesale electric-ity market (WEM) participation has also been opened upto DERs, through aggregation companies and new marketmodels, which will continue to grow as per FERC Order2222 (FERC, 2020). The WEM, however, is not designedfor a large penetration of DERs (MITEI, 2016). FERC Or-der 2222 not withstanding, as enrollment is limited, whole-sale participation models alone are not sufficient for efficientand effective DER integration, especially as DER penetra-tion increases and a zero marginal cost system is desired.In addition, misalignment of different tariff structures be-tween wholesale and retail programs limits the profitabilityfor DER services (Tansy et al., 2018). Regulatory and tech-nical barriers continue to prohibit DER participation as well.Other obstacles include high costs for participation whereeconomies of scale do not apply for DERs, misalignmentin interconnection procedures between retail and wholesaleprograms, and rules for 24/7 participation not suited for behind-the-meter resources (Gundlach and Webb, 2018).In this paper we propose a retail market mechanism thataims to address these limitations, through a distribution-levelmarket which coordinates the flexibility of DERs, leverag-ing the concept of transactive energy . Defined by NIST as“a system of economic and control mechanisms that allowsthe dynamic balance of supply and demand across the entireelectrical infrastructure using value as a key operational pa-rameter” (NIST, 2017), transactive energy bridges the gapbetween physical power flow in the grid and market deriva-tives. Such a retail market has its foundation in an advanceddistributed optimization algorithm, which enables local andprivate bidding transactions, to achieve network-level objec-tives. In particular, we propose a DSO-centric retail mar-ket that determines the appropriate incentives for DERs toparticipate in the market (Haider et al., 2020). These mone-tary incentives take the form of distribution-level LocationalMarginal Prices (d-LMPs) to participants at the distribution primary feeders, similar to the notion of LMPs employedas pricing signals in the wholesale energy market by Inde-pendent System Operators (ISOs) at the transmission level(EIA, 2011). The d-LMPs are determined using a distributedoptimization algorithm, termed Proximal Atomic Coordina-tion (PAC) developed in (Romvary et al., 2020; Haider et al.,2020; Romvary, 2018), as a core component. All underlyinggrid physics and constraints in the distribution system are in-corporated in deriving the d-LMPs. As a result, they havethe potential to fully exploit the emerging flexibility of thedistribution system, and reduce operational costs across thepower supply chain. Technologies such as Advanced Meter-ing Infrastructure (AMI) umbrella, ubiquitous even now, andadopted by several utilities across the US and Europe, canall be leveraged to implement the proposed retail market. Inaddition, an advanced communication technology that sup-ports peer-to-peer message passing is assumed to be presentto support implementation.This paper is organized as follows. Section 2 providesan overview of the current US regulatory landscape, andthe future environment within which retail markets are de-veloped. Section 3 discusses the precursors to distributionmarkets, including existing retail and wholesale compensa-tion schemes for DERs, and outstanding limitations in mar-ket participation. Section 4 describes the design of an inte-grated energy and ancillary services market for distributionsystem, provides simulation results validating market per-formance, and discusses the technologies required to deploythe proposed retail market design. Finally, conclusions andpolicy discussion are provided in Section 5.
2. Regulatory Environment
Integral to any plans for DER integration into marketoperations is the understanding of the regulatory landscape,and forms the focus of this section. We restrict our discus-sions to the US grid and the electricity landscape and theassociated markets therein. The proposed retail market ismore broadly applicable across the globe. In what follows,we provide a brief overview of the current US regulatorystructure, and the future setting within which the proposedmarket could operate.In the US, legal jurisdiction over energy and electric-ity interconnection, markets, and operations is divided intofederal and state authority. The Federal Energy RegulatoryCommission (FERC) has authority over all wholesale marketoperations and participation, tariff structures, generation andtransmission planning, and interstate commerce, and relia-bility standards that are overseen by North American Elec-tric Reliability Corporation (NERC). At the regional level,ISOs and Regional Transmission Operators (RTOs) are third-party organizations which operate the grid, oversee the whole-sale market, ensure reliability and economic efficiency, andensure non-discriminatory market participation, under theoversight of FERC and NERC . There are seven ISOs/RTOs For the discussion in this paper, ISOs and RTOs are interchangeable FERC does not have regulatory oversight of the Electric Reliability
R Haider et al.:
Preprint
Page 2 of 14SO-Centric Retail Market currently approved in the US, covering two-thirds of the US,the rest of which is operated by investor-owned utilities, co-operatives, or Federal Power Marketing Administrations. Atthe state level, each state has governance over retail tariffs,in-state transmission regulations, and policies for distribu-tion system-level resources, including their interconnectionpolicies. This taxonomy across wholesale and retail gover-nance is particularly relevant in the discussion of DERs andtheir market participation.
The federal regulations governing resource interconnec-tion and market participation are laid out in the Federal PowerAct of 1935 which established FERC as an oversight body,the Public Utility Regulatory Policy Act (PURPA) of 1978and 2005 which opened market participation to non-utilitygenerators, and subsequent Orders. As is well known, FERChas issued a series of Orders to reduce the regulatory barri-ers to entry for variable renewable energy generators, de-mand response, and storage participation in wholesale mar-kets. FERC Orders 2006 and 792, issued in 2005 and 2013respectively, required interconnection procedures for smallgenerators ( < MW, including storage devices) and estab-lished a fast-track process to reduce the timeframe for ap-proval (Chernyakhovskiy et al., 2016). Orders 719 and 745,issued in 2008 and 2011 respectively, require all ISO/RTOsto accept bids from DR aggregators acting on behalf of retailcustomers, and requiring the full LMP to be paid to DR re-sources that are dispatched to balance supply and load (Gund-lach and Webb, 2018). In 2013, Order 784 required ancillarymarkets to include a pay-for-performance pricing scheme,increasing the opportunity for storage resources, and pavingthe way for service-based remuneration for resources (Chernyakhovskiyet al., 2016). More recently Order 841, issued in 2018 andheld up by the US Court of Appeals in 2020 (John), and Or-der 2222, issued in 2020, address barriers for storage devicesand DERs, requiring ISO/RTOs to enable participation inwholesale markets for these resources and DER aggregators,and compensate them for the services they provide. Of these,Orders 719, 745, and 841 have spurred the most regulatoryactivity, and Order 2222 is poised to apply additional pres-sure for ISOs/RTOs to revise wholesale market participationmodels.
The role of DSOs has been broadly discussed in the con-text of existing regulatory and market structures, where DSOstypically own and operate a part of the electrical distribu-tion network and are compensated through a rate-base modelregulated by a local regulator (e.g. state public utilities com-mission) (Bös, 2015; Faruqui, 2012; Hogan, 2010). In manycases they also act as an intermediary between end customersand wholesale power markets by procuring electric supply.Studies such as (Ruester et al., 2014; Gerard et al., 2018;
Council of Texas (ERCOT), Hawaii, or Alaska, as they do not have anyinter-state electricity flows, but these states must comply with NERC relia-bility standards (Chernyakhovskiy et al., 2016)
Bell and Gill, 2018) examine the evolution of DSOs in Eu-ropean countries, and report on a wide range of roles and re-sponsibilities for the DSO, as well as coordination schemesbetween DSOs and Transmission System Operators (TSOs).These studies highlights the urgency with which DSOs shouldreform policies about access, usage, and compensation ofDERs for the services they provide. Most notably, (Gerardet al., 2018; Bell and Gill, 2018) focus on emerging retailmarkets operated by DSOs, comparing different responsibil-ity coordination schemes between the DSO and the TSO toappropriately operate and compensate distribution-level as-sets. The centralized approach is a ‘business-as-usual’ casewherein flexibility resources are transacted within a TSO-operated market, with little to no knowledge of distribution-level constraints, and the distribution system largely contin-ues to operate under the traditional “Fit-and-Forget” model.In contrast, the decentralized approach constitutes local DSO-operated markets, enabling direct purchase of flexibility re-sources and the aggregation of DER operation into the TSOmarket. The DSO-centric market structure proposed in thispaper is similar to the latter approach.
3. Precursors to Distribution Markets
In the absence of distribution markets, policies and pro-grams have been developed to compensate distribution-levelresources for the services they provide to the broader grid,including direct incentives and feed-in tariffs. Many of theseprograms, which include DR and TVP can be viewed as pre-cursors to distribution markets. However, these policies fallshort of yielding efficient investment and operations of dis-tribution systems, as they do not coordinate resources throughbidding, dispatch, and settlement rules. In particular, thesepolicies do not price the fine-grain locational and temporalvariation in the services that DERs are capable of providing,and are therefore unable to meet network-level objectives orefficiently manage grid conditions, and struggle to maintainreliability under high DER penetration. In contrast, the pro-posed retail market does not have these advantages, and de-tailed in Section 4. This section will discuss existing retailand wholesale programs, and highlight the existing barriersto market participation for DERs.
Time variable pricing tariff structures attempt to providepricing signals for customers to shift their consumption be-haviour and more accurately recover the costs observed bythe power system from final customers that are otherwise in-sulated from the dynamics of the electricity system. Suchdynamic retail rates were motivated as early as the emer-gence of the first failures of newly deregulated transmission-markets, such as California’s blackouts and market powerscandals in the early 2000’s (Borenstein, 2002). In morerecent years, utilities look towards TVP to reduce electric-ity use throughout the day, especially during times whenthe grid is stressed, thus reducing costs from operating fast-responding peaker plants. This translates to reductions in
R Haider et al.:
Preprint
Page 3 of 14SO-Centric Retail Market electricity bills for end-use consumers, and a non-wires al-ternative (NWA) for utilities. These rates provide more cost-reflective price signals for retail customers, and can enablemore efficient use of DERs such as demand response andretail-customer-sited storage systems.TVP constructs include Time-of-Use (TOU) Rates, whichvary the energy supply charge based on time blocks in a dailyschedule, and vary seasonally; Real-Time Pricing (RTP) andDay-Ahead Hourly Pricing, which both introduce a variableelectricity rate based on the underlying hourly wholesale mar-ket prices; and Critical Peak Pricing (CPP), which charges asignificantly higher energy rate at times of congestion. Ratestructures which hybridize these core TVP structures havealso been developed, such as Variable Peak Pricing, whichhybridizes TOU and RTP to have a dynamic price for on-peak blocks based on utility and market conditions; and Block-and-Index Pricing (also called Block-and-Swing Pricing), whichhybridizes fixed pricing and RTP, so only a portion of thecustomer’s load floats with market pricing, allowing them tohedge risk from price variability.TVP programs are designed as both opt-in and opt-out;programs that are opt-in typically have lower participationrates, while opt-out programs result in greater reduction inelectricity consumption through collective buy-in power. Vol-untary TOU programs exist for residential and C&I customers,including Eversource in ISO-NE; PG&E, SDG&E, and SCEin CAISO, and Xcel in ERCOT. However, the majority ofthe mandatory TOU programs in the US are for commer-cial and industrial (C&I) customers, such as PSE&G in PJMand NYISO; Madison Gas & Electric in Wisconsin; and CityLights in Washington. (EERE) This limits the energy andcost savings which can be achieved. Other popular TVPmechanisms include CPP which is widely used by CAISO,and is the default rate structure for most small, medium, andlarge businesses, including agricultural customers servicedby SCE and PG&E (EERE), and Day-Ahead Hourly Pric-ing which is mandatory for customers serviced by ConEdin NYISO (of Energy Efficiency & Renewable Energy), andDelmarva in PJM (EERE).
Retail demand response programs compensate for reduc-tions in electrical demand during peak usage periods, com-promised grid reliability, or during high wholesale marketprices. The design of DR programs varies greatly, with someprograms allowing the participation of retail customers withlocal generators, where increased DER output reduces netload. Retail customers are remunerated through capacitypayments, which are based on the committed available loadreduction, and/or performance payments, which are basedon the delivered demand reductions or increased power out-put of DERs. Some programs also have penalties for fail-ure to reduce load (CPower, 2020; Energy, 2016). Capac-ity payments are typically used to incentivize customer en-rollment. The structure of these DR programs vary widely,from reductions with ahead notice ranging from 10 minutesto 24 hours; load shifting to adjust usage to times with lower network load; and automated DR or direct load control pro-grams, in which customers allow utilities to automaticallyreduce load during high usage periods. In the latter, con-trol actions include adjusting programmable thermostats, cy-cling AC units, turning off electric water heaters, or shuttingoff pumps for agricultural customers (EERE). In some pro-grams, the location of DR resources is restricted by voltagelevel, to below 69kV (Paso Electric Company, 2020).Despite their diverse structure, there are several prob-lems with these retail DR programs. First, performance pay-ments only incentivize customers to change their behaviorduring performance periods or call windows (e.g. 2-6pmnon-holiday weekdays June 1-Sept 30). Many programs alsocap the number of times a resource can be called (e.g. 10times in the summer months of May-August). As a result,they do not translate into continuous compensation mecha-nisms for DER operation across the year, with studies show-ing little difference in consumption outside of call windowsfor consumers participating in DR programs (Wolak, 2006).Second, baselining methodologies are used to determine theexpected consumption, and to calculate demand reductions.Unfortunately, these schemes frequently over- and under- com-pensate customers as they attempt to determine a counterfac-tual with no load reduction (Wolfram, 2017). This method,of predicting the baseline using historical data, is not robustto changing customer behaviour or environmental events,such as weather-specific behavior events in summers andwinters when thermally-dependent loads can be larger thanpreviously observed. Further, it is prone to customer gam-ing, as DR participants can increase consumption outsideof call windows, especially during baseline setting times, tobe compensated for reducing this higher load, as done bythe Baltimore Orioles baseball stadium in 2013 (Borenstein,2014; Pepper, 2013). Third, incentive payment rates [$/kWreduction] are highly variable between programs. Customersare typically remunerated at fixed prices set in contracts withthe host utility, or variable prices based on the tier of ca-pacity commitment and time duration. As such, they arelargely static rates which overcompensate resources (Wolak,2006). These payments reflect utility peak demand costs,equipment upgrades, and emergency conditions, and are typ-ically not reflective of the actual service provided by theDERs. Finally, and most importantly, retail DR does notprovide adequate incentives for desired behaviours to meetstate-wide RPS or energy goals. The overcompensation ofDR undermines long-run investments in energy efficiency(Borenstein, 2014), as it lowers baseline consumption andtotal kW curtailable load. These DR programs have not seenany clear positive growth, as apparent in Fig. 1 which plotsretail DR capacities by NERC region between 2013 and 2016.It is necessary to note, however, that the inefficiencies high-lighted above lie within the design of the retail programs,and not within the resource itself. The flexibility and dis-tributed nature of DR poises this resource as a key compo-nent of the future electricity grid, as displayed by the demandreduction and flexibility offered to the California grid dur-ing rolling blackout conditions from extreme weather events
R Haider et al.:
Preprint
Page 4 of 14SO-Centric Retail Market
Figure 1:
Retail Demand Response by NERC Region (2013-2016) FERC (2018) in 2020. These success came with a shift away from tradi-tional DR composed of large C&I, and towards aggregationsof devices and appliances such as thermostats, batteries, andhome car chargers, which are dispatched by the wholesalemarket (Trabish, 2020). This newer model clearly reflectsthe potential for DR resources in retail markets.
In the absence of retail markets in the distribution grid,several compensation schemes have been proposed for behind-the-meter resources including generators (both conventionaland renewable) and storage devices, based on passive meter-ing and load management by end-use consumers. In regionswhere compensation schemes are not present, local gener-ation resources simply reduce the customer load, and ex-cess generation is curtailed. Compensation schemes allowexporting to the grid, for which the customer is compen-sated, through either net energy metering (NEM) or net en-ergy billing (NEB) (Energy Solutions Center). In NEM, theexcess generation (which is exported to the distribution grid)is subtracted from the imported energy (delivered by the dis-tribution grid). This net energy is charged to the customerat the retail rate if net imports exceed exports, or purchasedfrom the customer if exports exceed imports at a purchaserate. In NEB, all imported energy is charged at the retailrate, and all exported energy is purchased by the utility atthe purchase rate. These amounts are netted and charged orcredited on the customer’s final bill.Many state regulators have implemented such NEM andNEB schemes to compensate DER exports. These policiesrange from mandates for load serving entities to establisha rate for every customer within the state, to no mandatoryrules, as can be seen in Figure 2. The purchase rate in NEMprograms are typically fixed retail rates, while NEB pro-grams may have varying prices based on the underlying whole-sale electricity price. In both cases, the compensation is ad-justed to estimate the utility’s avoided cost, based on the off-set by net energy exports from DERs onto the broader grid.In some states such as Texas, retail net metering is not al-lowed, so the purchase rate is set to be a lower “wholesale” rate, which compensates the electricity import without sub-sidizing its production costs (XcelEnergy).While net metering has gained considerable traction inUS markets, NEM policies overcompensate DGs, often pro-viding a premium of up to 2-3 times the energy value. Thepurchase rate is typically comparable to the retail rate, de-spite the fact that the energy generation component only makesup half to a third of the rate, and NEM customers rely on thegrid 24/7 for backup and do not contribute to grid mainte-nance (Puckett, 2020; Wood, 2016). The price for distribu-tion system management is then offloaded onto non-NEMcustomers, and supports rate increases by utilities, further-ing the social imbalance in electricity prices. The purchaserate can be even higher than the retail rate in states promotingsolar PV uptake. In this way, NEM policies subsidize privatesolar, at the expense of non-solar owners, resulting in a re-verse Robin Hood effect (Ritchie, 2016; Smith et al., 2018).These inequalities have spurred some regulatory reform ef-forts (Wood, 2016), including the use of NEB in some statessuch as Texas, charging NEM customers a fee for grid main-tenance, lowering the purchase rate, or pushing for compen-sation at the wholesale rate. Regardless of the purchase ratehowever, none of these compensation schemes fully incor-porate locational or time price differentials, virtue of aver-aged tariff rates. These stagnant rates with limited volatilitylimit the adoption of dispatchable resources, such as stor-age, across the distribution system. Further, treating thesebehind-the-meter resources as primarily load modificationresources limits the services they can provide to the grid,particularly in grid stability provisions (Hinson, 2019).
In addition to retail programs and compensation schemes,participation of DERs in wholesale markets has been at theforefront of regulatory activity, and is further bolstered byFERC Order 2222 which requires ISO/RTOs to create DERaggregators as a market participant class. In this section weaim to cover a few diverse participation models.
R Haider et al.:
Preprint
Page 5 of 14SO-Centric Retail Market
Net Metering
State-developed mandatory rules for certain utilities (40 states + DC+ 4 territories)Statewide distributed generation compensation rules other than net metering (5 states)
KEY
U.S. Territories:
40 States + DC,
AS, GU, PR, & USVI have mandatory Net Metering rules DC No statewide mandatory rules, but some utilities allow net metering (2 states)
GUAS PRVI
In transition to statewide distributed generation compensation rules other than net metering (6 states)
Figure 2:
Net Energy Metering Policy By State (April 2019)DSIRE (2019)
The participation of DR resources in wholesale marketsbegan in the early 2000s, with more ISO/RTOs creating suchprograms to comply with FERC Orders 719 and 745. Mostnotably, CAISO has had aggregation programs for DR since2001 (Gundlach and Webb, 2018), and have since expandedthe program to include more participation models, in theProxy Demand Resource (PDR). In the PDR model, resourcescan bid load curtailment into both the energy and ancillarymarkets, with a minimum 100kW and 500kW load curtail-ment required respectively. The PDR model also includesthe Reliability DR Resource (RDRR), in which curtailmentis triggered only under emergency conditions. As of 2019,the Load Shift Resource (LSR) was introduced to allow bidi-rectional dispatch and reward resources for increasing con-sumption during negative pricing (CAISO). Many behind-the-meter resources participate through DR aggregations. How-ever, DR are still classified as behind-the-meter load mod-ification, not generation supply solutions, thus imposing ano-export rule on generators and storage participating in DRaggregations (Ulmer et al., 2018).Participation of DR in the energy and reserves marketwas also introduced by ISO-NE in 2018, through the cre-ation of the ‘active’ DR class. Termed Price ResponsiveDemand, retail customers with DR capabilities are knownas DR assets (DRAs), and are classified as either passiveor active resources. Passive DRAs are nondispatchable, re-sources including energy efficiency resources and behind-the-meter solar PV, and cannot participate in energy or re-serve markets. Active DRAs are dispatchable resources in-cluding load reduction, on-site generators, and storage. Sim-ilar to the CAISO model, these resources bid load reductionsinto the energy and reserve markets, and are cleared as re-sources comparable to generators. Aggregation is also per-mitted for active DRAs smaller than 5MW. Both active andpassive resource classes can participate in the capacity mar-ket (ISO-NE, 2019b; Yoshimura, 2018; ISO-NE, 2019a).In PJM, DER participation in the wholesale market oc-curs through Curtailment Service Providers (CSP) which pro-vide both emergency and energy resources. Emergency DRparticipate largely in the capacity market under the Reliabil-ity Pricing Model (RPM), with remuneration based on ca- pacity commitments to be called on during emergency con-ditions, while energy DR can participate in both energy andancillary markets. Energy DR are called upon to displacegenerators in the energy market when the wholesale priceexceeds PJM net benefits price (PJM).Similar to the other participation models, MISO has al-lowed Aggregators of Retail Customers (ARCs) to bid intothe wholesale market as reduction in demand since 2012.Based on the resource class, these ARCs can take part inthe energy market, operating and planning reserves markets,and emergency response, and require minimum sizes of ei-ther 1MW for participation in energy and reserve markets,or 0.1MW for participation in emergency response. Behind-the-meter generation are subsumed within load modificationresources, and cannot participate in energy and reserve mar-kets, but can be used to meet resource adequacy require-ments (MISO, 2020a).
CAISO has been at the forefront of DER integration bycreating of the DER Provider (DERP) participation model.Established in 2015, this model allows aggregations to en-able small-scale DERs, each < MW in size, to collectivelymeet the minimum 0.5MW requirement to participate in theCAISO energy and ancillary markets. Aggregations are anew type of market resource, similar to a generating facility,and can bid into the market to be cleared as a single unit.Such aggregations can be composed of different resourcetypes, and do not have to be geographically co-located. Rather,aggregations can span multiple transmission node connec-tions, and therefore multiple pricing nodes, but must remainwithin electrically defined zones which have minimal pricedifference between the nodes, called sub-Load AggregationPoints (subLAP). Aggregations spanning multiple nodes can-not exceed 20MW. Each of the underlying resources are re-munerated through a weighted average LMP across the pric-ing nodes of the aggregated resource, to reflect congestionrelated benefits from each resource (Ulmer et al., 2018; CAISO,2016). The DERP model has greatly influenced FERC’s de-cision in issuing Order 2222.The participation model for ISO-NE does not use aggre-gators, but rather waives the minimum size requirement. TheSettlement Only Resources (SOR) class consists of genera-tors connected to the distribution system, and are less than5MW. These resources participate in the RTM as price takers- they do not bid supply offers into the DAM or RTM; ratherthey self-dispatch and are paid the RT LMP when they pro-duce energy. The SOR class can also participate in the ca-pacity market if they are a minimum of 100kW (ISO-NE,2019b; Yoshimura, 2018). Resources participating in thecapacity market are permitted to submit composite bids, inwhich resources with seasonal capacities can be aggregatedto meet the year-round availability requirement. This allowssummer-only distributed generation to couple with winter-only resources, widening the participation model for sea-sonal DERs (Nichols and Lehman, 2019). In 2019, ISO-NEbecame the first to permit hybrid resources to participate in
R Haider et al.:
Preprint
Page 6 of 14SO-Centric Retail Market capacity market auctions, with the 2022-2023 auction clear-ing a bid for an aggregated residential solar-plus-storage re-source, using a virtual power plant (VPP) model (Gheorghiu,2019). The clearing price for this 14th auction was the low-est in the auction’s history, at $2 compared to $7.03 in 2016and $3.80 in 2019 (ISO-NE, 2020c), continuing the down-ward trend in capacity prices, driven primarily by increasedparticipation of solar resources supported by batteries (Foleyet al., 2019; Eckhouse and Martin, 2019).
Despite the existence of these models, there is limitedparticipation of DERs in wholesale markets. In a reviewof the CAISO DERP aggregation model conducted in 2018,CAISO only had four participants registered in the DERPprogram, of which none had begun participating in eitherenergy or ancillary markets. Interviewed active and poten-tial participants indicated that participation in the wholesalemarkets would likely be limited to short- or medium-term,due to limited profitability. Further, DERs such as behind-the-meter storage are not well supported in the DERP model,which requires 24/7 settlement, prohibiting resources fromstepping out when electricity prices are too high and dis-couraging DERs that were acquired primarily to meet on-site energy needs (Gundlach and Webb, 2018). In the ISO-NE region, only 40% of solar PV resources were partici-pating in the wholesale market in 2019 (ISO-NE, 2019b),though retail compensation schemes including NEM poli-cies are thought to have contributed to the rapid growth ofdistributed solar (Gundlach and Webb, 2018). While PJMhas a large capacity of DER participation in wholesale mar-kets compared to other regions, an estimated 7GW of DERpotential still weren’t participating in 2019. The share ofDERs participating in PJM DR programs has also been de-creasing since 2017, and DER participation as a DR resourcein the energy markets has been decreasing since 2014. Fur-ther, of the locations within PJM with behind-the-meter re-sources like generation and batteries, most do not have ex-port access; less than 5% of these resources participate ineither retail or wholesale activities, of which less than 8%participate in wholesale markets (PJM, 2020). Similarly inMISO, 43% of unregistered DERs are solar PV (MISO, 2020b).
A summary of the inefficiencies and/or barriers to par-ticipation of the above programs follows.•
Static pricing:
Temporal and locational pricing is notavailable to realize the flexible and unique nature ofgrid services from DERs. The limited variability ofpricing signals in retail programs such as NEM andDR performance payments limits the adoption of re-sources which can respond quickly and dynamicallyto local conditions and provide grid-level support.•
Voluntary enrollment:
Effective demand responseprograms must incentivize behavioral changes through- out the day, not just during performance periods. Es-sential to this is the increased participation in TOUrates, which are primarily opt-in programs, whose suc-cess enrollment success depends on promotion by util-ity companies. Although enrollment into TVP pro-grams has been increasing since 2013, only a smallfraction of retail customers are enrolled. With an es-timated 200 GW of flexible load by 2030, widespreadadoption of TVP programs by retail customers, espe-cially EV owners, is necessary in realizing this poten-tial. (Foster et al., 2019)•
Competing retail and wholesale programs:
Currentmarket designs do not permit participation in both re-tail and wholesale programs. As a result, these pro-grams compete with one another for DER enrollment.For example, the DERP program inadvertently com-petes with the wholesale PDR and retail NEM, of whichthe latter two programs are less costly to participate in(Gundlach and Webb, 2018). However, both of thesestructures provide limited services to the grid, sub-ject to no-export rules, limited ancillary market par-ticipation, and introduce barriers to the entry of stor-age. The misalignment of the different tariff structureslimits the profitability for DER services (Tansy et al.,2018).•
Prohibitive technical requirements:
The technicalrequirements from DERs participating in both retailand wholesale programs are limiting in two ways: mis-alignment with grid services, and economic barriers.For example, Rule 21 interconnection standards re-quire residential DERs to have only hourly or day-ahead functionalities, thus limiting their usefulness,particularly in their ability to provide stability provi-sions. Further, the metering and telemetry require-ment for aggregated resources is the same as for tra-ditional generators, despite their different capacitiesand capabilities. Resource aggregations also typicallydo not benefit from economies of scale. In CAISO’sDERP program, each DER which is part of the aggre-gation must install its own revenue meter, which intro-duces prohibitive costs for small operators (Gundlachand Webb, 2018).•
Prohibitive regulatory requirements:
Interconnec-tion rules and procedures vary between retail and whole-sale level participation, which creates a barrier of en-try for DERs already participating at the retail level toenter the wholesale market, despite the creation of ag-gregation programs. For example, under CAISO, if aresource is connected by Rule 21 (for which rules varybetween utility distribution companies) and wants tonow participate in wholesale, it must reapply underWDAT, which is structured for conventional genera-tors and often allocates the cost of technologies to theresource. Second, there is no standardized commu-nication protocol across retail and wholesale spaces,
R Haider et al.:
Preprint
Page 7 of 14SO-Centric Retail Market limiting DER participation in aggregations or underVPP models. Third, wholesale programs which al-low DERs to provide services to the grid are limiting,in that they often require all-year and 24/7 participa-tion. Under such models, resources cannot step outof the market when desired , unless they enroll in DRprograms. This is limiting for DERs, especially thosewhich are behind-the-meter and serve on-site load: astorage device discharging locally (i.e. not injectingpower to the grid) when the LMP is negative mustmake payments to the wholesale market (Gundlachand Webb, 2018).
4. Proposed Retail Market Design
As DER penetration continues to increase, better coor-dination of these resources is warranted. There is a need forincreased temporal and locational granularity in electricitypricing, innovative ancillary products, an expansion of mar-ket derivatives to include more grid services, and an align-ment of retail and wholesale markets through coordinatedtariff structures and market clearing schemes, and are theemerging responsibilities for DSOs (Anisie et al., 2019). Inour market design, the DSO is responsible for overseeingthe participation of DERs in a retail market, through whichDERs are scheduled and remunerated at real-time prices.The market is composed of a (1) real-time energy marketwhich schedules DERs and determine market settlements;and (2) an ancillary services market which balances loadacross primary feeders. In this paper, we limit our focus tothe energy market.The DSO is composed of two entities, the Workers (DSO-W) and Representatives (DSO-R), which reside at the sub-station and primary feeder respectively. The DSO-Rs over-see the energy market and aggregate data of the DERs un-der their purview; and the DSO-Ws operate the ancillarymarket and aggregate information from the DSO-Rs. Whilethe DSO acts as a data aggregator, it does not bid into theWEM on behalf of its DERs like an aggregation company ortransmission level resource. Rather, the DSO can be viewedas a proactive utility in the sense that it accepts the Loca-tional Marginal Price (LMP) as traditionally determined bythe WEM, and optimally makes use of the DERs within thedistribution network to maximize economic efficiency andother network-level objectives. In doing so, the DSO re-quests service from the WEM only for net loads beyond theDER capabilities, and compensates/charges the DERs fortheir services/usage at the d-LMP.A schematic of the operation of the proposed retail mar-ket is shown in Fig. 3.
The energy market is a highly distributed local real-timemarket carried out by the DSO-Rs. The market operates atthe primary feeder level (4 to 35 kV level); any DERs and This rule limits arbitrage opportunities, which aligns with ISO/RTOsneed for reliable and transparent market participation.
DSO for Feeder 1
Successive market bids through PAC Successive market bids through PAC ! ",$ ! ",% ! ",&'$ ! ",& DSO for Feeder L ! (,$ ! (,% ! (,&'$ ! (,& Coordinated by Distribution System Operators
Wholesale Electricity Market
Coordinated by Independent System Operator ) " ) ( LMP d-LMPd-LMP ) Retail Market
Bulk Energy System (Transmission)
Wholesale Market
Distribution Grid (Primary feeder)
Retail Market
Figure 3:
Proposed retail market structure uncontrollable loads at the secondary feeder level and beloware represented through aggregators, through which they canparticipate in this retail market. Each primary feeder has itsown DSO-R which oversees the energy market. To simplifythe discussion, we model every bus in the physical networklayer as an independent agent participating in the marketlayer, which represents all the DERs located at and/or belowthat node . These DERs include both behind-the-meter re-sources and those connected directly to the distribution grid,including DR, DGs, and storage.Each agent is equipped with the necessary computationaland communication infrastructure to participate in the mar-ket, which is built upon a distributed optimization algorithmcalled PAC (for technical details see Haider et al. (2020);Romvary et al. (2020); Romvary (2018)). Using this algo-rithm, each agent self-schedules to minimize its expenses(equivalently maximize profit) while subjected to networkconstraints such as voltage limits, thermal line limits, andother DSO-level objectives, which are modeled through anon-linear convex optimal power flow formulation. The DERdispatch schedules and d-LMPs are determined by repeatednegotiations between neighbouring agents using peer-to-peercommunication, which are carried out autonomously usingPAC. During every negotiation, the PAC algorithm requireseach agent to communicate its proposed load or generationsetpoint and variables pertaining to the network’s physical This is not a technical limitation of the proposed structure. Multipleneighbouring nodes can choose to be represented by the same agent, whichwould then have access to all required operational data and pricing infor-mation. A detailed discussion of such an agent is beyond the scope of thispaper.
R Haider et al.:
Preprint
Page 8 of 14SO-Centric Retail Market constraints (such as voltage and current) to its neighbours.These interactions will determine the d-LMP (retail cost ofelectricity per kW) of each agent, based on marginal cost ar-guments. After reaching an agreement with its neighbours,each agent enters into a bi-lateral agreement with its DSO-R, committing to deliver or consume the decided amount ofpower, at the d-LMP. The net load consumed by an agentwill be charged at this d-LMP, and equivalently, the net gen-eration by an agent will be remunerated at the d-LMP. Pay-ments will be made to/from the DSO-R. The above transac-tions proceed in a parallel fashion across all agents report-ing to a DSO-R. It is recommended that this retail energymarket has a shorter clearing period than the WEM clear-ing. As non-dispatchable resources displace conventionalgeneration, market clearing times must be faster and moreflexible, in order to better reflect the highly temporal natureof renewable resources and to accommodate updated fore-casts (NASEM, 2021; Poplavskaya and Vries, 2019). Forexample, the WEM in ISO-NE clears every 5 minutes, so theproposed retail market could clear every minute. The mar-ket clearing time however, can be freely selected to suit theneeds of each ISO/RTO. It should be noted here that our pro-posed retail market includes a market for both real and reac-tive power. Although reactive power markets don’t currentlyexist, even in the WEM - due to issues of price volatility andmarket power concerns - the increase in DER penetrationtogether with enabling technologies such as smart invertershas the potential to realize efficient reactive power marketdesigns.
We benchmark the proposed market operation againstfour operating models, wherein the utility purchases powerfrom the WEM at the wholesale price and sells to customersat a fixed retail price, as is currently done in the US. The firstof these is a ‘Traditional’ model where there is no DER uti-lization. The ‘No Export’ model realizes DR as continuousDER operation, rather than only during specific call win-dows which are a limitation of the programs introduced inSec. 3.2, and retain the no-export rule for behind-the-meterresources from wholesale market participation models (seeSec. 3.4.1). Retail compensation schemes NEM and NEBdiscussed in Sec. 3.3 are used in the ‘Retail_M’ and ‘Re-tail_B’ models.
Traditional:
There is no DER utilization within the net-work. All load is serviced by the utility.
No Export:
DG resources are used to offset local load andcannot export excess generation to the grid (excess genera-tion is curtailed). All load from customers without DGs andany excess load of DG owners is serviced by the utility.
Retail_M:
DG resources are used to offset local load andcan export excess generation to the grid. Compensation forDGs follows NEM, at a fixed retail purchase rate. Any ex-cess network load is serviced by the utility.
Retail_B:
DG resources are used to offset local load and canexport excess generation to the grid. Compensation for DGsfollows NEB, at a fixed retail purchase rate. Any excess net- work load is serviced by the utility.We use several metrics to validate the market performanceusing different stakeholder perspectives. They include therevenue for a DER owner, the cost for a customer consumingelectricity, and the net revenue for the DSO. These metricsare calculated as follows. Real and reactive power are de-noted as 𝑃 and 𝑄 , with superscripts G and L for generationand consumption respectively. Subscript 𝑗 denotes the j-thagent participating in the market. The wholesale LMP is de-noted as 𝜆 P , retail electricity prices as 𝜇 Pretail , retail purchaserate as 𝜇 Pretail-G , and the d-LMP for an agent 𝑗 as 𝜇 P 𝑗 and 𝜇 Q 𝑗 .The baseline load for an agent 𝑗 is denoted as 𝑃 L0 𝑗 and 𝑄 L0 𝑗 .With this notation, we define the following quantities. Payment made to WEM, for purchasing power: WEM = 𝜆 P ∑ 𝑗 𝑃 L 𝑗 Revenue earned from loads without proposed market: baseload = ∑ 𝑗 𝜇 Pretail 𝑃 L 𝑗 Revenue earned from loads with proposed market: marketload = ∑ 𝑗 ( 𝜇 P 𝑗 𝑃 L 𝑗 + 𝜇 Q 𝑗 𝑄 L 𝑗 ) Remuneration to distributed generators: gen = ∑ 𝑗 ( 𝜇 P 𝑃 G 𝑗 + 𝜇 Q 𝑄 G 𝑗 ) where for traditional and no export cases 𝜇 P = 0 and 𝜇 Q = 0 ,for Retail_M and Retail_B cases 𝜇 P = 𝜇 Pretail-G and 𝜇 Q = 0 ,and for the proposed market 𝜇 P = 𝜇 P 𝑗 and 𝜇 Q = 𝜇 Q 𝑗 . Remuneration to flexible loads: flex = ∑ 𝑗 ( 𝜇 P 𝑗 ( 𝑃 L0 𝑗 − 𝑃 L 𝑗 ) + 𝜇 Q 𝑗 ( 𝑄 L0 𝑗 − 𝑄 L 𝑗 )) The metrics are then defined as:
Revenue for DER owner: flex and gen Cost for consumer: xload Net revenue: = xload − flex − gen − WEM
In the numerical exercise that follows, the market opera-tion has been simulated over a 24 hour period, on the IEEE-123 node network, which is a primary distribution feedermodel. The network data was modified to be a balanced3-phase distribution network, and DERs were added to thenetwork (Haider et al., 2020). All loads are assumed to becapable of DR (in real power). About 10% of the nodes inthe grid are assumed to have local generating capabilities,with almost 70% of the total network load capable of beingmet by the total nameplate generation. Market reports fromISO-NE operations provide the five-minute approved LMPs(wholesale price) (ISO-NE, 2020b), and five-minute totalrecorded electricity demand from which the time-dependentdemand ratio 𝛼 ( 𝑡 ) is calculated, for August 25, 2020 (ISO-NE, 2020a). Load data from the IEEE datasheet providesthe upper bound on load forecast, with the real-time forecastvarying as per 𝛼 ( 𝑡 ) . Retail data from Eversource in Mas-sachusetts is used for the benchmark scenarios, with 𝜇 𝑃 retail =$0 . /kWh (generation service charge for basic service),and 𝜇 𝑃 retail-G = $0 . /kWh (Class I solar/wind under Resi-dential R-1 tariff) (Energy, 2020).Results from the simulation are presented in Figures 4-7. R Haider et al.:
Preprint
Page 9 of 14SO-Centric Retail Market
The normalized retail prices from the proposed energy mar-ket, which are calculated using PAC, are shown in Fig. 4.There is a high locational variation in both prices, with 𝜇 𝑃𝑗 prices varying by a factor of 2 within the same period. Theresults also show temporal variation for prices at the samenode, with higher variation in 𝜇 𝑄𝑗 which sees a factor of 3.8between highest and lowest prices throughout the day. Thetemporal variation of 𝜇 𝑄𝑗 roughly follows the demand ratio 𝛼 ( 𝑡 ) , with higher prices during higher load periods. Thisis likely because the DG units were configured to provideonly real power and all reactive power load must be met bythe utility purchasing power from the WEM; to stabilize theprices, all DGs must have reactive power capabilities, suchas solar PV with smart inverters. This is in line with the re-vised Rule 21 interconnection procedures in CAISO, whichrequire DG units to be equipped with smart inverters priorto their approval. In comparison, there is lower temporalvariability in 𝜇 𝑃𝑗 ; this is likely due to the modeling choices(DGs with continuous output and a fixed percentage of cur-tailable load at all times of the day) and the smooth demandcurve. More realistic data including the variability in renew-able generation such as a day with passing cloud cover, net-works with high loading conditions, and more granular mod-eling of DR capabilities may increase the volatility of thereal-time price. The retail market allows the DSO to pricethese spatial-temporal variations and realize the true valueof energy services provided by DERs. It may not be desir-able to expose customers to these volatile prices, which canbe remedied by more traditional TVP techniques, which av-erage prices over a period such as an hour.The aggregated hourly schedule determined by the en-ergy market is shown in Fig. 5. The forecasted network loadis serviced by the utility purchasing power from the WEM(in grey), DGs serving both onsite load and exporting powerto the grid (in green), and curtailment from demand response(in blue). The graph also shows the total power loss due toelectrical resistance (in burgundy). The wholesale price 𝜆 is also plotted (black line). The maximum LMP coincideswith peak network load in hour 17, during which both DRand DG utilization is at a maximum. Periods of low demandand low wholesale prices have lower resource utilization, aspurchasing power from the WEM is comparable to remuner-ating a DER at 𝜇 𝑃𝑗 , with an average LMP of $0.0267/kWhand average 𝜇 𝑃𝑗 of $0.0291/kWh. The aggregated resourceutilization for each market operation benchmark and the pro-posed retail market is shown in Fig.6. The dashed line showsthe total load serviced under the proposed market operation,which is lower than the benchmark cases which do not haveDR enabled. Both the Traditional and No Export scenariosfail to utilize DERs, and while the Retail_M scenario doesuse DGs, there is no coordination of resources to achieveeconomic and energy efficiency.A detailed comparison of the cost of market operation isshown in Fig. 7. Both the Traditional and No Export sce-narios result in large profits for the utility, due to the largedifference between the retail and wholesale prices of elec-tricity. Both Retail_M and Retail_B result in a loss for the (a) d-LMP Real Power.(b) d-LMP Reactive power Figure 4:
Locational-temporal variation in retail price, usingPAC algorithm and proposed market utility. While these retail compensation structures are cur-rently used in US electricity markets, the high retail pur-chase rate means the utility is not only overcompensating theDGs, but that under high penetration of these DG resources,this participation model becomes uneconomical. One op-tion is to provide lower purchase rates, however decidingthe value of the energy service being provided is challeng-ing. Another option is to enable participation at the whole-sale level, but this continues to be challenging for small re-sources, even through aggregator models. Most notably inFig. 7, all the quantities for the retail market scenario are sig-nificantly lower than of the Traditional and No Export case,and only comparable to the Retail_M/Retail_B cases for thecost of electricity from the WEM. Despite serving the sameload, the proposed retail market is able to do this at a muchlower retail cost: an average of 0.0291 $/kWh, compared tothe current utility retail price of 0.114 $/kWh. Rather thansimply making a large profit, the proposed DSO is buildingsocial equity and redistributing wealth through socializationof the profit. With the retail market, the true value of energyis recovered, which results in lower electricity costs for con-sumers and lower compensation for DERs, while ensuringpower balance and economic efficiency in the market.
To implement the proposed retail market and support thesecure bidding between agents, communication infrastruc-ture in the distribution grid is needed. The existing com-munication infrastructure in the bulk energy grid, which areused to support operations and wholesale market functions,consists of optical fibers connecting control centers to sub-
R Haider et al.:
Preprint
Page 10 of 14SO-Centric Retail Market
Figure 5:
Schedule for resources using proposed RM and PAC.The forecasted load is serviced by the utility, DGs, and DRcurtailment. Additional power is purchased due to losses inthe network.
Figure 6:
Resources utilized across different market scenarios.The forecasted load is serviced by the utility, DGs, and DRcurtailment. stations, and multiple protocols like Synchronous OpticalNetworking (SONET) which enable fast and secure grid com-munications. At the sub-transmission and distribution level,Wide Area Networks (WAN) or LTE based networks areused in conjunction with various protocols such as DNP3(legacy) or IEC 61850, to measure and control the grid. Mar-ket functions, such as submitting bids, are typically performedover the Internet, using various authentication mechanismsto ensure security.To implement the retail market and realize the successiveautonomous bidding mechanism of the PAC algorithm foreach agent, the existing communication infrastructure can befully utilized. Currently, SCADA networks provide limitedcommunication and control in distribution systems; how-ever, this is rapidly changing with the proliferation of AMIsystems. While AMIs were initially deployed to aid in gridoperations, their use has evolved to support market functionssuch as TVP (Foster et al., 2019). Further, with the adventof the Internet of Things (IoT), connectivity and communi-cation with grid edge resources and loads is enabled withoutthe need to build additional infrastructure, providing morevisibility and control capabilities. Various communicationprotocols such as ZigBee, Modbus, IEEE Std 802.15.4, andPLC standards allow for communication with AMI devices. Recently, the IEEE 2030.5 Smart Energy Profile (SEP) 2.0standard, which provides a framework for monitoring andcontrol of DER assets, has been gaining traction with gridoperators, and has been suggested as the standardized com-munication protocol for DER aggregation programs. For ex-ample, CAISO outlines in their Common Smart Inverter Pro-file (CSIP) how SEP should be implemented to meet Rule 21,requiring DERs to have monitoring and reporting capabili-ties, and grid support functionalities such as Volt-VAr Con-trol (VVC) (Tansy et al., 2018). With these technologies, aPAC-based retail market can be realized by leveraging thegrid-edge intelligence and connectivity of resources.
5. Conclusion and Policy Implications
A retail market operating within the distribution grid en-ables the participation of small scale resources, which in-clude, among others, DG, DR, and storage including EVs.These DERs can be compensated for the services they pro-vide at a real-time rate with such a market, based on theirmarginal cost of operation and current grid conditions. Suchan approach also allows resources to operate more dynami-cally and eliminates the no-export rules for behind-the-meterDERs , so they can generate, reduce load, or even increaseload as needed by the network. As DER penetration contin-ues to increase, technology costs reduce, and subsidies forthese resources are removed, new incentives for DER partic-ipation in markets is required. This can be achieved throughnew revenue streams from retail markets. In this paper weproposed a retail market structure using a distributed opti-mization algorithm capable of solving for the optimal dis-patch and d-LMPs, while leveraging grid-edge intelligenceand peer-to-peer communication (Haider et al., 2020; Rom-vary et al., 2020).The proposed energy market can also be augmented withan ancillary market (Haider et al., 2020), by allowing DSO-W to coordinate DSO-Rs to ensure balance of supply and de-mand under service disruptions commitments. In doing so,the flexibility of the DERs under the purview of the DSO-Rare able to participate in balancing and are compensated fortheir fast reacting capabilities. These distributed resourceswith computational abilities can then be utilized for grid re-siliency in the face of large outage events, extreme weather,and cyber attacks. Designing these derivatives and under-standing the operational overlap between energy, ancillary,and ‘voltage’ services is necessary to fully realize smart co-ordination of DERs and leverage their flexibility.The hierarchical structure also lends itself to localizedenergy markets within the secondary feeder, realized throughtechnologies such as blockchain-enabled peer-to-peer energytrading. At this level, submetering can be used for load dis-aggregation, particularly for co-located resources such as EVsand loads. This notion of hierarchical markets and opera-tions requires coordination between interfacing markets: thelocal and retail market at the interface of the primary feeder,and the retail and wholesale market at the interface of the where technically feasible, i.e. grid constraints are satisfied R Haider et al.:
Preprint
Page 11 of 14SO-Centric Retail Market
Payment by Loads – [ Payment to WEMPayment to DGsPayment to DRs ] ++ = Net RevenueforUtility/DSO $$$ $$
Figure 7:
A breakdown of the net revenue calculation under all scenarios, with all revenue and cost in USD. The kW value usedto calculate the amount is annotated on the bars - these are the load serviced by the utility, power purchased from the WEM,load serviced by the DGs, and load curtailment by DRs. transmission substation. Coordination can be realized bycommunicating net load/generation and forecasts, sharingoperating status, or even bidding mechanisms. In this lattercase, at the DSO-TSO interface the DSO acts as an aggre-gator, bidding the net load/generation for the network intothe WEM. Co-optimization can be carried out through it-erative schemes, to determine the optimal schedule of bulkresources and DSO aggregators, and determine the LMP atthe DSO node. In this way, the DSO is no longer a pricetaker but an active market player capable of setting whole-sale prices, and the DERs can better respond to bulk energysystem changes.From a regulatory perspective, an aggregator participa-tion model for DERs into the wholesale market needs to befurther analyzed, especially with FERC Order 2222: shouldDERs be able to participate only in the retail market, or shoulddirect wholesale participation be allowed, and if so, can re-sources participate in both markets simultaneously? If ableto participate in either market, can resources freely move be-tween them, or will they be required to provide 24/7 ser-vice to a single market for a duration of time, say a year?Is a reactive power market with contained volatility realiz-able? In each of these participation models, resource ad-equacy and market fairness come into question: are partici-pation models centered around the flexibility of DERs (as re-quired by FERC Order 841) fair to traditional bulk resourceswhich cannot readily step in and out of markets, and are be-ing pushed out of the energy market by lower cost renew-ables - and, in future, DERs - but are still needed on standbyto provide fast ramping? Alignment is also needed with bal-ancing and reserve markets to integrate DERs alongside aretail market, else market gaming between intraday and bal- ancing markets will persist (Just and Weber, 2015). Our pro-posal for an allocation of tasks among the wholesale andretail markets is this: The need for new or updated capac-ity models and compensation for standby generators mustbe addressed at the wholesale level. The task of DER in-tegration and compensation must be addressed at the retaillevel. There needs to be appropriate coordination betweenthese two markets. Our proposed retail market is a first stepin answering all of the above questions.A few statements need to be made regarding the lowercosts for customers that can be realized using the proposedretail market. This is in sharp contrast to the current regula-tory structures in which retail prices are increasing despitethe drop in levelized cost for renewables and decreasing –and sometimes even negative – wholesale prices (Murray,2019). Efficient pricing and resource coordination at the dis-tribution level permit these lower costs. As we look towardsreal-time pricing, however, we must also consider rate equityacross different socioeconomic classes, to ensure fair accessof electricity (Burger et al., 2019). Regulator concerns aboutexposing customers to the price volatility of real-time ratescan be alleviated by ensuring tariff designs are equitable,and by employing hybrid TVP models such as Block-and-Index pricing to enable risk hedging. The lower retail pricealso results in a lower revenue stream for the DSO. How-ever, we note that this is not endemic to the proposed re-tail market, but rather a reality of the modern electricity gridwith negative electricity prices (frequently occurring in USstates of California and Texas, and Germany), high renew-able curtailment, and unprecedented ramp rates as in the fa-mous California ‘duck curve’. This is already manifesting insystems with high renewable penetration at the transmission
R Haider et al.:
Preprint
Page 12 of 14SO-Centric Retail Market level: the share prices for the three largest utilities in Ger-many have dropped by 45% to 66% between 2010 and 2016,and other utilities in Western Europe have similarly lost mar-ket value over the past decade (MITEI, 2016). While thismay seem concerning for the future of the utility as we knowit, it is a reality stemming from the misalignment of incen-tives present in the current utility business model. The utilitymodel and corresponding rate structures must be redesignedto shift away from a commoditized market with capital ex-penditures and energy sales as the main revenue stream, andtowards performance-based ratemaking (PBR) where util-ity revenue is instead based on achieving performance met-rics and other non-investment factors. Although the com-pensation mechanism for resources, in particular generators,will likely become increasingly complex, this new businessmodel can help realign revenue with state RPS and energygoals, by supporting utility investment into NWA and moreefficient grid utilization. Retail electricity prices then bet-ter reflect both the quality of service for customers, and theperformance and responsiveness of utilities to governmentmandates (Littel et al., 2017; Aggarwal, 2018). The extent towhich the composition of revenue stream will change and theresulting decrease in retail electricity prices depends heavilyon the geographical location. Factors such as renewable pen-etration at the transmission level, DER penetration, interde-pendence of electricity prices and commodity prices (such asnatural gas in the Northeast US), climate, wholesale marketstructure and capacity procurement, and regional RPS goalsand policies can result in different performance metrics fora single retail market design. Of equal importance is an-alyzing how regulatory and policy changes can impact thebusiness model structure for providing electricity services(Burger and Luke, 2017), and ensuring high enrollment ofDERs to increase market liquidity (Weber, 2010).Finally, the design of the retail market must uphold andsupport both state and federal policy objectives. The optimalmarket design requires consideration of both short- and long-term incentives for all market participants (Weber, 2010).More analysis is needed to determine how the proposed mar-ket structure can promote investment into energy efficiency,grid reliability, and clean energy. Assessing the impact ofcarbon pricing and environmental costs, and accounting forexternalities such as air quality and healthcare costs is alsonecessary (Bell and Gill, 2018). The interaction between theelectricity and natural gas markets also needs to be better un-derstood, especially in areas where gas is used for both bulkelectricity production and home heating, as in the NortheastUS. Another interesting concept is the realization of a ‘ther-mal market’ whereby DR is also enabled for thermal loads,such as space and water heating, which traditionally rely ongas. As we look to electrify more sectors of the economy,including heating, transportation, and manufacturing, the in-teraction of these networks must be accounted for. A smartcity approach can better integrate electricity consumption,EVs, and thermal loads, to achieve higher operational effi-ciency and lower costs.
References
Aggarwal, S., 2018. America’s utility of the future forms aroundperformance-based regulation. Forbes .Anisie, A., Boshell, F., Ocenic, E., 2019. Future Role of Distribution Sys-tem Operators: Innovation Landscape Brief. Technical Report. Interna-tional Renewable Energy Agency.Bell, K., Gill, S., 2018. Delivering a highly distributed electricity system:Technical, regulatory and policy challenges. Energy policy 113.Borenstein, S., 2002. The trouble with electricity markets: understandingCalifornia’s restructuring disaster. Journal of economic perspectives 16,191–211.Borenstein, S., 2014. Money for nothing? Energy Institute at HAAS .Bös, D., 2015. Pricing and price regulation: an economic theory for publicenterprises and public utilities. volume 34. Elsevier.Burger, S., Schneider, I., Botterud, A., Pérez-Arriaga, I., 2019. Fair, Eq-uitable, and Efficient Tariffs in the Presence of Distributed Energy Re-sources. pp. 155–188. doi: .Burger, S.P., Luke, M., 2017. Business models for distributed energy re-sources: A review and empirical analysis. Energy policy 109.CAISO, . Summary comparison matrix. .CAISO, 2016. Distributed energy resource provider participa-tion guide with checklist. .Energy Solutions Center, C., . Net metering and net billing. https://cleanenergysolutions.org/instruments/net-metering-net-billing
Chernyakhovskiy, I., Tian, T., McLaren, J., Miller, M., Geller, N., 2016.U.S. laws and regulations for renewable energy grid interconnections.Paso Electric Company, E., 2020. El Paso Electric Company’s 2020 loadmanagement program. .CPower, 2020. Understanding PJM capacity demand response changes. https://cpowerenergymanagement.com/help/pjm-dr-changes/
DSIRE, 2019. Net Metering Policies by State. Technical Report. Databaseof State Incentives for Renewables and Efficiency.Eckhouse, B., Martin, C., 2019. Residential solar and storage to participatein New England wholesale energy capacity market. Bloomberg NewsEditors .EERE, U., . Demand response and time-variable pric-ing programs. .EIA, 2011. Wholesale power price maps reflect real-time constraints ontransmission of electricity. .Energy, E., 2016. NHPUC no. 9 – electricity delivery; public ser-vice company of New Hampshire DBA Eversource Energy. .Energy, E., 2020. Eastern Massachusetts net metering credit pric-ing. .of Energy Efficiency & Renewable Energy, O., . Time-of-use rates. .Faruqui, A., 2012. The ethics of dynamic pricing, in: Smart Grid. Elsevier,pp. 61–83.FERC, 2018. 2018 Assessment of Demand Response and Advanced Me-tering. Technical Report. Federal Energy Regulatory Commission.FERC, 2020. Participation of Distributed Energy Resource Aggregationsin markets operated by Regional Transmission Organizations and Inde-pendent System Operators.Foley, E., Blomberg, M., Kakley, M., 2019. New England’s forward
R Haider et al.:
Preprint
Page 13 of 14SO-Centric Retail Market capacity auction closes with adequate power system resources for2022-2023. .Foster, B., Bialecki, T., Burns, D., Kathan, D., Lee, M.P., Peirovi, S., 2019.2019 assessment of demand response and advanced metering. .Gerard, H., Puente, E.I.R., Six, D., 2018. Coordination between transmis-sion and distribution system operators in the electricity sector: A con-ceptual framework. Utilities Policy 50, 40–48.Gheorghiu, I., 2019. Residential solar+storage breaks new ground as Sun-run wins ISO-NE capacity contract. UtilityDive .Gundlach, J., Webb, R., 2018. Distributed energy resource participationin wholesale markets: Lessons from the California ISO. Energy LawJournal doi: .Haider, R., Baros, S., Wasa, Y., Romvary, J., Uchida, K., Annaswamy,A.M., 2020. Towards a retail market for distribution grids. IEEE Trans-actions on Smart Grid doi: .Hinson, S., 2019. What California’s Rule 21 gets right and wrongfor residential solar. .Hogan, W.W., 2010. Fairness and dynamic pricing: comments. The Elec-tricity Journal 23, 28–35.IRENA, 2019. Innovation Landscape for a Renewable-powered future: So-lutions to integrate variable renewables. Technical Report. InternationalRenewable Energy Agency.ISO-NE, 2019a. About demand resources. .ISO-NE, 2019b. Response to letter dated September 5,2019 regarding par-ticipation of distributed energy resource aggregations in markets op-erated by regional transmission organizations and independent systemoperators. .ISO-NE, 2020a. Energy, load, and demand reports. .ISO-NE, 2020b. Pricing reports. .ISO-NE, 2020c. Results of the annual forward capacity auctions. .John, J.S., . ‘Enormous step’ for energy storage as court up-holds FERC Order 841, opening wholesale markets. Green-TechMedia URL: .Just, S., Weber, C., 2015. Strategic behavior in the German balancing en-ergy mechanism: incentives, evidence, costs and solutions. Journal ofRegulatory Economics 48.Littel, D., Kadoch, C., Baker, P., Bharvirkar, R., Dupuy, M., Hausauer,B., Linvill, C., Migden-Ostrander, J., Rosenow, J., Xuan, W., Zinaman,O., Logan, J., 2017. Next-Generation Performance-Based Regulation.Technical Report. NREL.MISO, 2020a. Der markets workshop. https://cdn.misoenergy.org/20200331%20DER%20Workshop%20MISO%20Summary%20Presentation440314.pdf .MISO, 2020b. MISO and DER: Framing and discussion document. https://cdn.misoenergy.org/DER%20Framing%20Report%202019397951.pdf .MITEI, 2016. Utility of the Future. Technical Report. MIT Energy Initia-tive.Murray, B., 2019. The paradox of declining renewable costs and risingelectricity prices. Forbes .NASEM, 2021. Legal and Regulatory Issues that shape the Electric System.The National Academies Press.Nichols, J., Lehman, S., 2019. Having a capacity supplyobligation lesson 2c: Introduction to capacity resources. .NIST, 2017. Transactive energy: An overview. . Nudell, T.R., Annaswamy, A.M., Lian, J., Kalsi, K., D’Achiardi, D., 2019.Electricity Markets in the United States: A Brief History, Current Oper-ations, and Trends. Springer International Publishing, Cham. pp. 3–27.doi: .Pepper, T., 2013. FERC settles investigation concerning demand responseproducts in PJM. Washington Energy Report .PJM, . Retail electricity consumer opportunities for demand response inpjm’s wholesale markets. .PJM, 2020. 2019 distributed energy resources (DER) that participatein PJM markets as demand response. .Poplavskaya, K., Vries, L.D., 2019. Distributed energy resources and theorganized balancing market: A symbiosis yet? case of three europeanbalancing markets. Energy policy 126.Puckett, J., 2020. Time to make net metering a net positive. Real ClearEnergy .Ritchie, E.J., 2016. The solar net metering controversy: Who pays for en-ergy subsidies? Forbes .Romvary, J., 2018. A proximal atomic coordination algorithm for dis-tributed optimization in distribution grids. Ph.D. thesis. MassachusettsInstitute of Technology.Romvary, J., Ferro, G., Haider, R., Annaswamy, A.M., 2020. A distributedproximal atomic coordination algorithm. IEEE Transactions on Auto-matic Control (provisionally accepted) .Ruester, S., Schwenen, S., Batlle, C., Pérez-Arriaga, I., 2014. From distri-bution networks to smart distribution systems: Rethinking the regulationof European electricity DSOs. Utilities Policy 31, 229–237.Smith, J.T., Patty, G., Colton, K., 2018. Net metering in the States: A primeron reforms to avoid regressive effects and encourage competition. TheCenter for Growth and Opportunity .Tansy, T., Nelson, R., Moy, K., Martinez, S., 2018. Analysis Reportof Wholesale Energy Market Participation by Distributed Energy Re-sources (DERs) in California. Technical Report. SunSpec Alliance.Trabish, H.K., 2020. Demand response failed California 20 years ago; thestate’s recent outages may have redeemed it. UtilityDive .Ulmer, A., Collanton, R.E., Ivancovich, A., Weaver, W., 2018. Post-technical conference comments of the California Independent SystemOperator Corporation.Vlerick Energy Centre, 2020. Outlook on the European DSO Landscape2020. Technical Report. Vlerick Business School. https://home.kpmg/content/dam/kpmg/pdf/2016/05/Energy-Outlook-DSO-2020.pdf .Weber, C., 2010. Adequate intraday market design to enable the integrationof wind energy into the european power systems. Energy policy 38.Wolak, F.A., 2006. Residential customer response to real-time pricing: TheAnaheim Critical-Peak Pricing experiment .Wolfram, C., 2017. The problem with demand response. Energy Instituteat HAAS .Wood, L.V., 2016. Why net energy metering results in a subsidy: Theelephant in the room. Brookings .XcelEnergy, . Texas solar and private generation. .Yoshimura, H., 2018. Statement of Henry Yoshimura, ISO New EnglandInc. .Zinaman, O., Miller, M., Adil, A., Arent, D., Aggarwal, S., Bipath, M.,Linvill, C., David, A., Kauffman, R., Futch, M., Arcos, E.V., Valenzuela,J.M., Martinot, E., Bazilian, M., Pillai, R.K., 2015. Power Systems ofthe Future: A 21st century power partnership thought leadership report.Technical Report. NREL.
R Haider et al.: