Approaches to Facilitating the Development Of Price-Responsive Demand Robert Earle Ahmad Faruqui

Approaches to Facilitating the Development Of Price-Responsive Demand Robert Earle Ahmad Faruqui Sam Newell The Brattle Group October 22, 2008

Agenda • I. Premise • II. Approaches to Facilitating Price-Responsive Demand • III. Integration of Approaches with Wholesale Market Design • IV. Advantages and Disadvantages of Each Approach • V. Conclusions

Premise • New England can benefit from Price Responsive Demand (PRD) • Can make energy markets more competitive, compared to inelastic demand • Achieves efficiencies by making consumption/production decisions and pricing better reflect both the marginal value and marginal cost of power • Could reduce the need for administrative market mitigation and price caps • Reduces the need for generation capacity • The Challenge for New England • How to increase the amount of PRD to its economic potential and integrate it into the DA/RT energy markets more efficiently?

There are a variety of approaches to providing PRD: a taxonomy • Direct Load Control (DLC) • Customer end uses are directly controlled by the utility and are shut down (or reduced) during emergencies • Air conditioning DLC is presently a popular approach • Applies to residential and C&I classes • Typically thought of as useful for reliability DR programs, but can also be used for PRD (aka, economic DR) • Indirect Load Control (ILC) • Customer end use controlled by the customer but given a signal by a third party provider (aka, CSP in PJM) • Interruptible Service • Customers agree to reduce consumption to a pre-specified level, or by a pre-specified amount, in emergencies • Applies only to C&I class • Dynamic Pricing • Includes time-varying rates that can be called upon during emergencies or high-priced hours to encourage reductions in peak loads • At the very least, interval meters and a communication system must be in place to offer these rates. Examples include Critical Peak Pricing (CPP) and Real Time Pricing (RTP) • Advanced metering infrastructure (AMI) is the leading choice for implementing dynamic pricing • Applies to residential and C&I classes • Likely the wave of the future – RTOs will need to address the integration of PRD into wholesale market design • Recent CPUC decision will make dynamic pricing the default rates in California

The motivation for dynamic rates • For most utilities, the top 100 hours of the year can account for more than 10 percent of system peak demand and can be very costly to serve • For ISO-NE, the top 100 hours are 13% of peak • 10% of the peak in ISO-NE occurs in 60 hours • Innovation in utility rate design seeks to lower costs and improve system efficiencies by targeting this critical-peak load • Traditional retail rate designs fail to accomplish the objective • Customers are unaware of the dynamic cost of providing electricity (no connection between flat retail rates and varying wholesale prices) • Non-time varying rates leading to over-consumption during peak periods • Even time-of-use (TOU) rates, while time varying, do not differentiate between regular peak hours and the highest cost hours of the year • But the majority of benefits of dynamic rates occur in the small number of critical peak hours • Dynamic pricing provides price signals that can vary as the system needs vary

This results in a menu of innovative rate designs New (Dynamic) Rates Critical Peak Pricing (CPP) Peak Time Rebate (PTR) CPP-Variable (CPP-V) Variable Peak Pricing (VPP) Real Time Pricing (RTP) Existing Rates Flat Rate Inverted Tier (Blocks) Seasonal Time-of-day (TOD)

Two Leading Approaches are CPP and PTR • Dynamic rates such as CPP and PTR provide demand response during peak conditions, which can result in short-term and long-term savings • Critical Peak Pricing • When the power system encounters critical conditions, the peak-period price rises to much higher but known levels, either on a day-ahead or day-of basis • Could be structured as CPP plus TOU (a CPP rate, a peak rate, and off-peak rate) or pure CPP (a CPP rate and an off-peak rate) • Peak Time Rebate • Customers stay on their current retail rate • Customers receive rebates for kWh reductions during critical peak events • The rebate is equal to the difference between the critical peak rate or the real-time price in the wholesale market and the “current” retail rate • The reduction in consumption during critical peak events must be calculated from a “baseline” consumption level

Illustration of a CPP Rate • Critical peak and off-peak prices should reflect system costs • Revenue neutral • 10-20 critical days during summer, 4-6 hours per critical event (60-120 hours total) • Can also be layered on a TOU rate or an inverted tier rate

The PTR is an inverse CPP rate

A key challenge for integrating dynamic retail prices into an RTO wholesale market involves the LSE callable structures • Some pricing structures are static and will influence the load forecast, but little else: • Prices are known ahead of time • TOU Influences pattern of consumption, but not dynamically • Other pricing structures are dynamic, but not callable by the LSE, e.g., real time retail rates • Prices not known until day ahead or real time – or in some designs after the fact • Influences pattern of consumption dynamically, but LSE simply passes the prices on • Affects the RTO’s load forecast • There are other forms of RTP where callability is possible • LSE callable structures involve: • The LSE calling an “event” • End-user’s prices changing as a result of that event • CPP, PTR • Does LSE callability mean RTO dispatchability?

Available Approaches to PRD • PRD can occur in three ways (along the lines of our previous business models): • 1. No institutional involvement of the RTO (“No Curves”) • Consumption is reduced in response to high prices (or anticipated high prices) • Australia is an extreme example of this where demand bids are vertical • Less benefit than integration into the wholesale market because price elasticity of demand effects are indirect, not direct • Not recommended as an option for RTOs to rely on – included for completeness • 2. RTO has mechanisms for bidding of PRD through demand curves (“Demand Curves”) • The ability to submit sloped demand curves into the day-ahead energy market allows for this • However, characteristics of the particular dynamic retail rate program and the RTO’s market rules must be matched up • Example: If the notification time for a retail day-ahead dynamic price program occurs before market close, then integration of that retail program into a day-ahead demand bid is not possible • 3. RTO makes payments to customers for reducing load (“Supply Curves”) • Necessitates creation of uplift in order to fund payments • Creates baseline issues

In theory, most dynamic retail rate programs could be integrated into RTOs • The question is: what accommodation/facilitation will have to be made, and will it be worth it? Let’s work through the issues with a specific example • Dynamic pricing in the form of critical peak pricing (CPP) or the like may likely be one of the greatest sources of PRD in the medium to long term • CPP typically consists of a base rate with 12 to 15 critical peak events that can be called during the year (usually restricted to summer months) • During those critical peak events, the price to end users rises and consumption falls

CPP under the “No Curves” Approach • Without facilitation on the part of the RTO, CPP can be used by the LSE to mitigate high prices • Call events when the LSE thinks prices are going to be high • Standard valuation methods exist to facilitate this • Buy less from markets/sell more of long position • This method does not maximize the usefulness of PRD • Doesn’t translate directly into demand elasticity in the RTO market • LSE must guess what prices are going to be – no direct interaction or price triggers • No ability to set prices (though this ability may be limited under other approaches as well) • To what degree will calling CPP events affect the predictability of load and, therefore, reliability/operating reserves/imbalance markets?

CPP under the “Demand Curves” Approach • Incorporating CPP as part of the demand curve bid would allow the LSE to set a reservation price for when it calls critical peak days • Set reservation price to reflect value of the call • Similar methods to evaluate when to call events as with the “No Curves” approach • Allows demand elasticity to play directly in the RTO market and perhaps let demand set the price • Translates into some amount of load reduction • The amount of load reduction is not likely to be exact due to varying end-user responsiveness • Is the inexactness of response analogous to wind? • Would need coordination with RTO on timing/locational issues/imbalance settlement • Program design at the end-user level would enhance usefulness at the RTO • Ability to call in blocks • Ameliorate locational issues • Not have to use resource as “all-or-nothing” • Timing of announcement of calls • Enabling technology would not only increase response, but likely stabilize it

CPP under the “Supply Curves” Approach • Paying LSEs for load reductions by calling a critical day as part of a CPP program would also allow the LSE to set a reservation price for when it calls critical peak days • The characteristics of the Supply Curves Approach is similar to the Demand Curves approach • Scheduling issues are generally the same • However, the primary differences on the end-user program design level is that the CPP program would require customer baseline (CBL) measurement • Don’t need a CBL for CPP if one of the first two approaches (i.e., No Curves or Demand Curves) are used • At the RTO level, the issues include: • Verification of CBL • Level of payments for the reduction

Summary of Advantages and Disadvantages • Dynamic pricing likely has the greatest resource potential for PRD and can work with all of the approaches (no curves, demand curves, supply curves). Except for a few leading programs and some pilots, however, it is only slowly beginning to take off. Barriers are: • Reluctance on the part of state regulators to expose all but largest customers to volatile prices. • Need AMI to provide interval data and 2-way communication for small customers. • In the absence of widespread dynamic pricing, various wholesale programs have been developed to enable the marketing of “negawatts” (supply curves) as a “no lose” way to expose customers on fixed rates to market prices at the margin • Approaches include LSEs offering DLC negawatts into RT markets, CSPs offering DLC (and indirectly controlled) negawatts into RT markets as self-scheduled or dispatchable, similar to large wholesale customers. DA participation is rare. • ISO must facilitate through programs that provide for: • Incorporation of DR negawatts into scheduling, dispatch, and settlement systems, similar to generation • Need to accommodate diversity of characteristics (e.g., MDT, intermittency, block loading) • No DR sets RT prices in other RTOs due to lack to telemetry, zonal prices, uncertain customer baseline until after-the-fact • Baseline methodology (sometimes performed by CSPs or distribution utilities) • Funding mechanism (i.e., LSE or all customers pay gross), sometimes including a subsidy • Unless customers have a way to take title to their baseline usage, negawatt approaches may not be the best option • Baselines are subject to gaming and uncertainty • Side-payments from other customers to fund payments to negawatts questioned if baseline is uncertain

Key Issues in Facilitating PRD • The “No Curves” approach basically leaves PRD outside the market • Requires nothing be done at the ISO! • “Demand Curves” and “Supply Curves” approaches share some issues: • Price setting • Dispersed response at the retail level requires aggregation on a zonal- rather than a nodal-basis, resulting in issues with setting LMP • Block loaded resources • Variability of reaction • Response amount is not exact for either DLC or pricing programs • In the California Statewide Pricing Project, confidence intervals range from ±6% to ±18% depending on pricing program • Issues with imbalance payments/penalties • Is demand response analogous to wind in this regard? • Conformation of bidding rules to program characteristics and timing

Key Issues with the Supply Curves Approach • Double payments and the price formation problem – as presented by Dr. Chao • PJM has attempted to avoid this by only paying the difference between LMP and retail rate • This approach is plausible if the retail rate is known, but where there is retail choice this may be difficult • Customer base line issues – theory described by Dr. Chao • AKA, Measurement & Verification • Supply curves approaches require estimation of a customer baseline (PTR, for example)

AMI Infrastructure Issues • Enabling dynamic pricing through AMI has costs • Most of the benefit from AMI comes through operational benefits • Existing infrastructure may be an issue Summary of California Utility AMI Costs Notes: Figures are rough estimates based on information in utility filings PG&E costs include initial $1.7 billion investment and $0.6 billion upgrade PG&E customers includes 5 million electric and 4 million gas customers Net cost calculation assumes 65% of costs are offset by operational benefits

Develop Vision of Endgame • Initial vision, subject to revision: • Every customer segment should provide price-responsive demand primarily through dynamic rates (very efficient and avoids awkward baseline/M&V issues) • Enable dynamic rates to be schedulable day-ahead through demand bids • Enable demand response that can be dispatched in RT (such as DLC) • To the extent feasible, enable price setting by price-responsive demand • May have to restrict to large customers depending on feasibility of overcoming difficulties for smaller customers • Smaller customers are dispersed • Explore how to make negawatts (supply curve approach) more cost effective and more useful

Approaches to Facilitating the Development Of Price-Responsive Demand Robert Earle Ahmad Faruqui