AMI/Demand Response: Getting It Right the First Time

Deck: 

Each DR portfolio will have a different set of AMI needs, based on overall technology infrastructure.

Fortnightly Magazine - September 2006

Advanced metering infrastructure (AMI) evaluations will benefit greatly from creating an appropriate DR portfolio as part of the overall solution.

In the Energy Policy Act of 2005 (EPACT), Congress sent a strong message to electric utilities, consumers, and industry regulators that they need to get serious about advanced metering, time-based rates, and demand response (DR).

To underline this point, EPACT states:

“It is the policy of the United States that time-based pricing and other forms of DR, whereby electricity customers are provided with electricity price signals and the ability to benefit by responding to them, shall be encouraged, the deployment of such technology and devices that enable electricity customers to participate in such pricing and DR systems shall be facilitated, and unnecessary barriers to DR participation in energy, capacity and ancillary service markets shall be eliminated.”

However, a number of state public utility commissions (PUCs) and electric utilities were evaluating the implementation of advanced metering infrastructure (AMI) even before EPACT was signed into law. In many cases, a significant portion of the economic justification for installing AMI systems is attributed to potential benefits derived from DR applications. AMI can be an enabling technology for DR efforts, and including DR benefits in the AMI business case is both reasonable and prudent.

Unfortunately, precious little time has been spent identifying and determining the degree to which DR will benefit the market, nor to the types of DR applications that will provide those benefits. Absent this understanding, it is difficult to scope the functional requirements of the AMI. Doing so will limit the overall benefits of the total AMI business case severely.

Table 1 shows the level of AMI functionality needed with various type DR applications.

For example, an AMI system should be designed with real-time, two-way data communication if the DR product is intended to support spinning reserves. An AMI system might have only daily data communication if DR is intended to provide a capacity backstop alone. Either of these situations is reasonable, but they would have different designs and, most certainly, different costs and benefits.

It is difficult to conclude the required functionality from an AMI absent an analysis of the benefits various DR efforts can provide. This may be analogous to putting the cart before the horse. In light of goals stated in EPACT, the industry is likely to spend millions of dollars evaluating new AMI. Having DR enabling technologies like AMI almost certainly will be beneficial to the electric industry. However, the AMI evaluations will benefit greatly from creating an appropriate DR portfolio as part of the overall solution. Each DR portfolio will have a different set of AMI needs. These need to be accounted for when determining the overall technology infrastructure.

The Contribution of Demand-Response Resources

Any efficient market depends upon the free interaction of demand and supply to achieve market balance. DR is the ability of the customer to influence their electric loads at key times. These actions are a viable option for peak-load reduction and for providing pathways for demand and price elasticity and high price mitigation, not to mention just keeping the lights on.

Barriers to DR exist in many electricity markets. Often, these stem from a history of regulated retail pricing based on a supply-side structure designed to deliver customers all the electricity they want at a fixed price. Now, there is a unique opportunity to bring the demand side back into the electricity market with potentially large efficiency gains. Three factors contribute to this current opportunity associated with DR:

• Technology advances in communication and controls;
• Increasingly favorable economics for DR resources because of rising fuel costs; and
• Uncertainty in the costs of future environmental mitigation associated with fossil fuel plants.

In addition, many parts of the country that have had excess demand for the past four years are projecting near-term shortfalls.

As a result, this is a time to explore a complete solution. A complete solution would incorporate the development of an AMI that would enable and enhance demand-side options, and allow for the development of DR as part of a portfolio of options to meet future electricity demand. Equally as important, expanding the DR portfolio also helps manage system risks going forward. This more diversified portfolio of both supply-side and demand-side resources is more robust in its ability to handle future shocks and mitigate the costs associated with the stress events all system planners worry about—a hot summer, with record peak demands, combined with a major plant outage and transmission constraints.

These once-a-decade events are characterized as low-probability, high-consequence. DR can go a long way in mitigating the adverse consequences of events to the substantial benefit of customers who eventually must pay for them.

Applying a Resource Planning Framework to DR

We developed a case study modeling effort for valuing demand-response resources (DRR) using a resource-planning context, and we examined changes in system costs with and without DRR included in a portfolio of resources. The difference in system costs over a 19-year time horizon provides an estimate of the value of DRR for the electric system. The specific model used for this effort was New Energy Associates’ Strategist Strategic Planning Model.3

The base case for the model realistically represents an electricity market that allows for appropriate tradeoffs between resources—both supply-side and DRR—and which is able to address issues such as off-system sales/purchases and system constraints (e.g., transmission constraints).4

Model Inputs

One-hundred cases were created as data inputs to the strategist model. These were calculated to represent a wide variety of possible futures. Monte Carlo methods were used to create these different future cases that represent the uncertainty in key future inputs. To accomplish this, we identified a number of pivotal factors and dimensioned the uncertainty around these factors. Data was provided for the years 2005 to 2023. In addition, we developed data sets for four DR programs as inputs to the model.

The key input variables around which we dimensioned uncertainty are:

• Fuel prices: natural gas, residual oil, distillate oil, and coal;
• Peak demand;
• Energy demand;
• Unit outages; and
• Tie-line capacities.

We included four DRR products as potential resources to meet future system needs, in combination with the full range of supply-side options. The four DRR programs are:

Interruptible Product: A known amount of load reduction based on a two-hour call period. Customers are paid a capacity payment for the megawatts, with penalties if megawatt reductions are not attained.

Direct Load Control Product: A known amount of load reduction with 5 to 10 minutes of notification. This is focused on mass market customers. As a result, it has a longer ramp-up time to attain a sizeable amount of megawatt capacity.

Dispatchable Purchase Transaction: A call option where the model looks at the “marginal system cost” and decides to “take” the DRR offered when that price is less than the marginal system cost. This program can also be classified as a day-ahead pricing program.

Real-Time Pricing (RTP) Product: This program poses a challenge in that there is no feedback loop built into the model that looks at the marginal hourly cost and the demand for that same hour. As a result, two pricing products are examined:

    • One is a peak-period pricing program that produced a reduction in peak demand and little impact on load in other hours. This is similar to a critical peak pricing product, with the overall monthly and annual energy demand largely unaffected.

    • The second is a standard RTP program that produces a reduction in peak demand and also an overall energy efficiency effect, resulting in reductions in weekly, monthly, and annual energy demand. This is consistent with the RTP literature.

We then use data from each product design to develop inputs to the strategist model, such that each program could be treated consistently by the model. All dollar values are inflated at a rate of 2.5 percent per year. The following data is supplied for each product for the years 2005 to 2023:

Case Study Results

Results from these analyses include the following:

Uncertainty: In the base case, the overall uncertainty in total system costs for each year (100 cases per year) is quite large across these cases—indicating that the uncertainty in the modest number of variables selected results in a wide range of net system costs for each year in the 20-year planning horizon. On average, the range is 100 percent, i.e., the highest cost in the range is roughly double the lowest cost for almost every year in the planning horizon.

Hourly Costs: On a peak-demand day with additional system stresses, such as 10 percent of generating capacity being offline, savings in marginal production costs are substantial. The addition of DRR to the system greatly reduce the “peakiness” of the hourly costs, reducing the maximum by more than 50 percent. For example, in one peak day in July the total cost savings are $24.5 million.

Capacity Charges: A substantial percentage of new capacity charges is deferred by the model because of the DRR availability. This amounted to savings of $892 million (2004 dollars) over the 20-year period.

Savings in Each Year: DRR provides significant benefits in those years in which it is used. While DRR provides considerable amounts of benefits on select days, there is a cost to building and maintaining the DRR capacity which is paid for in every year and in every case, even if DRR is not used. This results in there being some cases where there are costs but no savings from DRR. Looking at the 100 cases individually, in the scenario with DRR but no RTP, 36 percent of the 100 cases show savings in total system net present value (NPV) compared with the base case, and with the full RTP scenario, 97 percent of the cases show savings.

DRR Capacity Usage: Large amounts of DRR are used about once in every four years. Across all resource scenarios, small amounts of DRR are used in most years in the planning horizon, with near capacity use of DRR happening infrequently. The amount of DRR called upon did not vary much across the three scenarios, e.g., the “with full RTP” resource option only resulted in a 10 percent reduction in DRR hours called across the 20-year planning horizon. As a result, the callable DRR retain their value as a hedge against extreme events even with pricing options that result in better utilization of system resources across all hours.

Cost-Risk Profile: There is a change in the risk profile associated with the planning scenarios with the addition of DRR. There are significant savings when looking at value at risk (VAR) at the 90th percentile (VAR90) and at the 95th percentile (VAR95). The VAR90 is the reduction in costs averaged across the 10 percent worst-case outcomes, i.e., the highest cost futures. Results for the three scenarios are shown below.

Loss of Load: The addition of DRR decreases the loss of load (LOL) hours substantially across all cases. The base case has an average value for loss of load hours of 7.64 hours across the cases, but values for some individual cases are as high as 30 hours. For the DRR with Peak Pricing, the average LOL hours averaged across all cases is lowered to 0.33 hours. The magnitude of the savings due to enhanced reliability across all the years in the planning horizon could be quite high, but no estimate has been calculated at this time and this estimate may vary by the number of customers impacted and the characteristics of different systems.

Total System Cost: Overall, the incorporation of DRR results in some reduction in the average total system cost net present value (NPV) in all three scenarios with DRR (callable DRR, DRR with CPP, and DRR with standard RTP). In the scenario with the standard RTP program, savings are about 3.5 times those in the scenario with the critical peak pricing program, and similarly, savings in the scenario with the critical peak pricing program are approximately twelve times those with only the callable DRR programs.

Incremental System Cost: As the system being studied is a very large system, it is meaningful to look at the incremental costs of meeting energy demand, as opposed to a percentage of the total system cost. On average, the savings in incremental costs due to DRR (year on year) are 10 percent for the scenario with peak pricing and 23 percent for the scenario with standard RTP. For the scenario with the standard RTP program there is a range of savings from -73 percent to +320 percent, and in 53 percent of the cases the incremental costs in the callable DRR scenario are less than or equal to those in the base scenario. In a few cases the DRR provides large reductions in incremental costs.

Overall, this case study shows that a Monte Carlo approach, coupled with a resource planning model, can address the value of DRR given uncertainties in future outcomes for key variables, and can also assess the impact DRR has on reducing the costs associated with low-probability, high-consequence events. In this case study, the addition of DRR to the resource plan reduces the costs associated with extreme events, and it reduces the NPV of total system costs over the planning horizon. This is an important finding. It can be compared to being paid to buy life insurance. Not only does DR reduce the expected or mean net system costs of meeting load growth, but it also greatly reduces the impacts of adverse events by between $1 billion and $2.5 billion—a considerable sum and a sizeable reduction in risks to ratepayers.

The outcomes of this case study are illustrative of both a method for assessing DR portfolios along supply-side portfolios, and the potential magnitude of the benefits. The role of DR resources is important for both cost effectiveness and risk management. As a result, any AMI structure developed should align with the development of DR resources.

As a final note, the total DRR capacity across all the DR options is approximately 15 percent of system-peak demand in 2015. A large DRR capability initially was viewed as appropriate for this case study. As the results section indicates, this level of DRR capability was found to be an overbuild for this system, i.e., DRR values of between 7 and 10 percent of total system peak would probably have been more appropriate for this system. This indicates that any resource will have diminishing returns at some level and, as with any resource, it can be overbuilt.

The electric power industry likely will invest billions of dollars in new AMI systems over the next 10 years. All energy consumers will bear the burden of these expenses, but they also will receive the benefits from the investment. To ensure these consumers receive the benefits that they deserve, the industry should spend the relatively small time and financial investment required to complete a proper analysis of the DR products and benefits. Absent this analysis, we may end up investing in system functionality that will not yield the benefits that we need and expect.

 

Endnotes:

1. Eric Hughes with New Energy Associates (EHughes@newenergyassoc.com) assisted with all of the resource planning model runs and provided insights regarding the interpretation of results.

2. The base case system was developed using data compiled by New Energy Associates, based on publicly available information for a selected region in the National Electric Reliability Councils (NERC), i.e., the Mid-Atlantic Area Council (MAAC) region. The initial data came from the Platts-McGraw Hill Base Case database for the region with some adjustments to the data based on New Energy Associates’ and Summit Blue’s experience.