Options for Reducing Peak Load Michigan Retreat on Peak Shaving to Reduce Wasted Energy Dr. Stephen George Senior Vice President Nexant August 6, 2014
Options for reducing peak load among residential customers Pricing Static TOU rates Dynamic rates (CPP, VPP, RTP, etc.) Demand charges Interruptible rates Pay for performance programs Peak time rebates Load control Switches Programmable communicating thermostats (PCTs) Smart thermostats (Nest, Ecobee, etc.) 1
What do we know about time-variant pricing? Four decades of research and experience support what every Econ 1A student knows when prices go up, demand goes down The higher the price, the lower the demand (up to a point) People can and do produce meaningful load reductions without enabling technology Combining time-variant pricing with load control technology increases average demand response by 50% to 100% Make sure you are doing an apples-to-apples comparison across groups with and without technology customers with technology typically all have air conditioning whereas some without technology don t Even if the incremental impact is 50% or more, it s still important to ask whether the net benefits of adding technology exceed the cost Marketing time-variant pricing to customers that already have enabling technology, and vice versa, is much easier than selling pricing or technology alone 2
What do we know about time-variant pricing? (continued) If you ask customers if they want to go on a time-variant rate, most will say no if you find a way to get them on a timevariant rate and ask them if they want to change back to a standard rate, most will say no Nearly every survey done for TOU pilots shows high satisfaction by people on the rates (see next two slides) Getting customers to enroll on time-variant rates is challenging, but there are some important success stories APS has 50% of its residential customers on TOU rates based largely on point of sale marketing (catching customers when they sign up for services) works well and is low cost SRP has more than 20% on TOU rates SMUD got 16% to 18% to sign up through a single season of marketing You can enroll customers on a default basis with very low opt-out rates (~5%) and very little dissatisfaction (both ComEd and SMUD pilots found this) 3
Customer satisfaction was very high for all pricing plans in SMUD s SmartPricing Options (SPO) Pilot 100 90 80 70 How satisfied are you with your current electricity pricing plan? 89.4 87.1 86 84.4 80.3 87.1 Percent 60 50 40 30 20 10 19.7 12.9 14 15.7 10.6 12.9 0 Standard Rate Default CPP Default TOU-CPP Default TOU Opt-in CPP Opt-in TOU Very or Somewhat Satisfied Very or Somewhat Dissatisfied 4
One reason for the high satisfaction is that customers feel TOU rates give them greater opportunities to save money than standard rates 100 My current pricing plan provides me with opportunities to save money 90 80 76.5 74.1 70 60 61.4 57.7 59.5 Percent 50 40 33.3 30 23.6 20 10 10.3 12.1 15.2 9 9.3 0 Standard Rate Default CPP Default TOU-CPP Default TOU Opt-in CPP Opt-in TOU Strongly or Somewhat Agree Strongly or Somewhat Disagree 5
What do we know about time-variant pricing? (continued) In SMUD s SPO, customers on CPP rates produced about twice the load reduction as customers on TOU rates The average load reduction for default customers was about half as large as for opt-in customers However, default enrollment (95%) was about five times larger than opt-in enrollment (16 18%, which was pretty high by industry standards) Combining enrollment with average load impacts, aggregate impacts were about 3 times larger under default enrollment compared with opt-in enrollment (see next slide) 6
SMUD s SPO Pilot is the only pilot that tested default and optin enrollment on identical customer populations MW Reduction Between 4 and 7 PM 40.0 35.0 30.0 25.0 20.0 15.0 10.0 Using SMUD s SPO enrollment and load impacts, if rates were offered to 100,000 customers, default enrollment would produce aggregate load impacts 3 times larger than opt-in enrollment 5.0 11.6 34.5 3.7 10.8 5.6 14.8 0.0 CPP Average Event Day TOU Average Weekday TOU Average Event Day Opt-in Default 7 7
What do we know about pay for performance (PTR)? With Peak Time Rebates (PTR), individual customers are paid on the difference between their metered peak period load during an event and an estimate of what they would have used if an event had not been called the latter is called baseline or reference load Baselines are estimated using an individual customer s load on prior days that have similar weather conditions as event days For example, the top 3 out of the prior 5 or 10 days Many different baseline methods have been used and tested While it is possible to find a baseline that is accurate (unbiased) on average across a large group of customers and multiple event days, IT IS IMPOSSIBLE to find one that is highly accurate for all event days averaged across all customers or for many event days at the individual customer level Put another way, almost no individual customer will be paid accurately very often, if ever there is simply too much day-to-day variation in usage across days (see next slide) 8
Across days, customer loads vary a lot due to factors other than weather Actual Load on Proxy Event Day Estimated Load Using 3- in-5 Baseline 9
What do we know about pay for performance (PTR)? There is a difference between baseline error and impact (payment) error If the average actual load reduction is 10%, a 2% baseline error produces a 20% error in the load impact (and customer payment) a 5% baseline error produces a 50% impact (payment) error When average impacts are large (e.g., 10 to 20%), an unbiased baseline will over pay some customers and underpay others but on average the incentives paid will be about right for the load reductions obtained When average impacts are small (1 to 2%), as they are with default PTR, overpayments do not fully offset underpayments (because underpayments bottom out at 0) and the program can significantly overpay for the load reductions obtained (see next slide) 10
Aggregate payment error and costs decrease with larger demand reductions Aggregate Payment Costs and Error Best Unadjusted Baseline (3 in 10) Based on analysis done for ComEd (but same elsewhere) $ Paid / kw year Aggregate Payment Error Assumptions: $180 90% 100,000 customers (20%) $160 80% 5 events per year $ Paid per kw year $140 $120 $100 $80 $60 $40 $20 $0 $20 2% 5% 10% 20% 25% 30% 70% 60% 50% 40% 30% 20% 10% 0% 10% Aggregate Payment Error (%) Rebate of $1.00 per kw Baseline bias of -5% Larger reductions reduce impact estimation error and payment asymmetry inherent in PTR Percent impact reduction can be increased through: Target high responders for marketing Combine PTR with enabling tech Avg. Demand Reduction (%) Do opt-in with notification, not default without notification 11
What do we know about pay for performance (PTR)? The larger the average load reduction, the smaller the percent of total incentive payments made in error Personalized notification, opt-in enrollment and/or enabling technology increase average reductions Default PTR with mass media notification is highly unlikely to produce large aggregate demand reductions and is highly likely to significantly over pay for the reductions obtained Default PTR with default notification (on customers for whom a utility has phone numbers or email addresses) is likely to produce more aggregate load reduction than default enrollment without personalized notification Whether it will produce larger aggregate load reductions than a well marketed opt-in program has not been tested empirically 12
What do we know about load control? Load control has been around for decades and is a very effective approach to reducing peak demand There are a variety of different technology options Switches PCTs Smart thermostats And there are many different program and incentive types the IOU programs just within California are quite diverse PG&E pays a modest, one time incentive ($25 - $50) and uses the program for emergency purposes only (or as an enabling technology for their CPP program) 168,000 devices, mostly switches SCE historically paid large annual incentives (as much as $200) and used it only for emergencies 400,000 switches in 2010, mostly on 100% cycling SDG&E outsources DLC to a third party (Comverge) and also pays reasonably large incentives used more as an economic dispatch resource 40,000 devices 13
What do we know about load control? The key to a more cost effective AC load control program is targeting Not every customer is eligible and not every eligible customer is necessarily a cost effective customer Recruiting among customers that have a high likelihood of owning an air conditioner and having high usage will reduce marketing costs Another critical factor is efficiency in scheduling and installations Some programs loose up to a third of recruited customers due to scheduling issues and delays Scheduling problems higher with PCTs than with switches because of the need to enter the home when installing PCTs Historically, PCTs have been more expensive and have had much higher communication problems than switches But this may not be the case with more modern switches or with smart thermostats 14
What do we know about load control? Cycling/control strategy is also important People will sign up for very high cycling rates (including 100% cycling) if you pay enough but make sure that you aren t paying more and getting less customers in hotter areas are less likely to sign up for higher cycling rates Load impacts from adaptive cycling are greater than from standard cycling High sign up rates are possible using telephone and door-todoor marketing compared with direct mail marketing Enrollment rates are much higher when you are selling to customers on dynamic rates (and perhaps TOU rates) if cycling is offered as an enabling technology But it s important to avoid double counting load impacts for cost effectiveness analysis or when predicting what load reductions will be if the entire DR portfolio is called 15
What does the future of load control look like? The future may be quite different from the past Commercially marketed smart thermostats and gateway devices offer both an opportunity and a threat to traditional utility sponsored load control programs The opportunity is that customers are purchasing smart thermostats on their own for the enhanced functionality they provide Learning capabilities and mobile information and control These customer purchased devices can cost effectively deliver load reductions during peak periods if utilities team with device suppliers in the long run, they could replace the need for utility owned switches and PCTs Even without teaming, device suppliers can independently deliver savings from customers who are enrolled in pay for performance or dynamic rate programs The threat is that utilities may have less control over the resource or that these programs can cannibalize legacy programs, creating stranded costs 16
For comments or questions, contact: Stephen George Senior Vice President, Utility Services sgeorge@nexant.com Nexant, Inc. 101 Montgomery St., 15 th Floor San Francisco, CA 94104 415-777-0707