4.3 Traditional Market Design Elements That Impact Flexibility Incentives
4.3.2 Five-Minute Scheduling and Five-Minute Settlements
The real-time market scheduling intervals of all the regions that operate wholesale markets in the United States are shorter in time resolution than scheduling intervals of utilities prior to
restructuring. In fact, all market regions in the United States now schedule the real-time market and real-time output of resources that offer their flexibility at a 5-min interval, updated every 5 minutes. This allows for better pricing of actual conditions on a more granular scale and
provides incentives for resources that can follow the prices. Because ramp constraints are used in the market-clearing model to constrain the ability of a supplier to sell energy into the market when they do not have the flexibility to follow prices, the selection of supply into the energy market should be based on actual capability when ramp constraints, provided by the resources are based on physical ramp rates. For example, in the hypothetical example in Table 4-2, Hour 7 to Hour 8, the supplier changes its output from 286 MW to 376 MW. Although it may be
possible that this change could be made during 60 minutes, it may not be possible for most thermal plants to execute this ramp during a 5-min period. If the supplier had a ramp rate of 5- MW/min, it could reach only 311 MW in the next 5-min interval, resulting in $99 of lost profit (although ignoring the fact that it is now MW in 5 min rather than MWh). In this way, the 5-min dispatch provides an incentive for flexibility in response to quickly changing prices.
Most of the 5-min energy markets that are currently in place in the United States do a good job of extracting flexibility without resorting to a separate market for a specific product for ramping capability, as illustrated by the example above. Units that bid into the 5-min energy market are obligated to ramp to their set point by the time the market period begins. Because these set points are calculated so frequently, many units ramp a substantial portion of the time. This allows for the most economic provision of energy given the constraints on the transmission system, and units can take advantage of price volatility when they can ramp faster. However, in some instances, it may be that ramp constraints can give the opposite effect. A resource that is ramp constrained will not set the LMP, because a faster, more expensive unit would have to be used to
make up for the slower unit’s ramp limitation. Thus, the more expensive unit will set the price, while at the same time giving a higher revenue opportunity for the slower unit.
Milligan and Kirby (2010) provide a simple illustration of the issue depicting the real-time market. In this example, a single time period market assumption is used without any look-ahead function in the dispatch interval, so that the future expectations cannot affect the current energy pricing. The example is a simplistic power system with only two generators: a baseload unit that has a marginal cost of $10/MWh and a peaking unit that has a marginal cost of $90/MWh, as illustrated in Figure 4-1. During a steep ramp that is beyond the speed that the baseload unit can respond, but within its capacity range, the peaking unit is dispatched to cover the ramp and meet the load. After the baseload unit catches up, the peaking unit is shut off. However, the peaking unit sets the energy price at $90/MWh during the time it is used to meet the load during the ramp period. Although this is not a problem per se, the baseload unit also collects $90/MWh during this period, which does not provide the baseload unit any incentive to become more flexible, and in fact it may provide a disincentive. This topic should be studied further to see what
consequences may occur. For example, a multi-period dispatch looks ahead to ensure that units can meet upcoming ramping requirements. This can change the market prices and revenues for market participants, even if pricing is set only for the current interval. In fact, many U.S. market operators have or are developing proposals to move toward multi-period dispatch when solving the real-time market (O’Neill et al. 2011).
Figure 4-1. Simple example of ramp-limited resources and resulting prices
Although all of the U.S. markets have 5-min real-time energy markets that dispatch and price energy at 5-min intervals, not all of these markets settle at this granularity (Hirst 2001). Many settle the real-time markets based on the average hourly price of all intervals within that hour. Some areas, including NYISO and SPP, however, do settle at 5-min intervals. SPP states that 5- min settlement incents the submission of ramp capability by resources precisely because the capability to move quickly is rewarded by an LMP commensurate with the 5-min instructions.
SPP further explains that without this settlement feature, resources may be disinclined to offer all of their ramp capability, perceiving that they are not being fully compensated for the actions required.13
To provide an illustration of how the settlement period can have an impact on incentives for flexible operations, we develop a simple example. Table 4-4 shows 12 5-min LMPs, from real LMP data. Column 3 shows the average LMP for the hour, which is calculated based on the cumulative average LMP from the beginning of the hour to the current time period. We use the same incremental costs for the supplier as shown in the earlier example in Table 4-1. In Scenario 1 (Market), the supplier follows a schedule, as in Section 4.3.1, based on the most efficient output level that the market operator computes and directs each 5-min period. In Scenario 2 (Moving HourlyAverage), the supplier follows an output that is based on the current hourly average LMP (from column 3), because it gets updated throughout the hour. We ignore any impacts from uninstructed deviation penalties throughout this example. Finally, Scenario 3 (Perfect Knowledge) shows a hypothetical example of what the most efficient output would be from the supplier if it had perfect knowledge of the final average hourly price ($49.10).
Table 4-4. Incentive Differences When Settlements Are Done on an Hourly Average Versus 5 Min
Interval ($/MWh) LMP Current Hourly Average Scenario 1 Market
Scenario 2 Moving Hourly Average Scenario 3 Perfect Knowledge of Hourly Average 0:05 $73.68 $73.68 377 377 340 0:10 $41.87 $57.78 286 377 340 0:25 $43.48 $53.01 286 377 340 0:20 $44.17 $50.80 286 376 340 0:25 $45.75 $49.79 286 358 340 0:30 $46.69 $49.27 286 340 340 0:35 $46.73 $48.91 286 331 340 0:40 $45.91 $48.54 286 322 340 0:45 $61.25 $49.95 377 358 340 0:50 $47.88 $49.74 304 358 340 0:55 $47.88 $49.57 304 349 340 1:00 $43.85 $49.10 286 340 340
An examination of the different dispatches in the table shows that the maximum flexibility is achieved in Scenario 1 (Market). When maximizing profit based on the anticipated or predicted hourly settlement (Scenario 2 and Scenario 3), the unit provides the incorrect level of or no flexibility. It is a simple extrapolation of this scenario that would illustrate similar behavior if the assumption of perfect foresight is relaxed and the unit bids to another hourly average price level. Because price changes in each of the 5-min periods, this is an indication that the system needs a varying level of output; otherwise the price would have remained constant throughout the hour. Next, we turn to an examination of the profits earned in these scenarios.
Table 4-5 presents the total revenue, total cost, and total profit for each of these scenarios with both 5-min settlement and hourly average settlement procedures. If settlements are based on the hourly average, as they are in many markets today, the supplier will earn more profit by
producing output differently than the dispatch schedule that was given by the market operator. This would result in output levels that are not the most efficient and could potentially result in reliability issues. In the hourly settlement case, the profit almost doubles when the supplier follows the hourly average price compared to the market schedules ($4,068 compared to $2,443). This shows that even though 5-min scheduling is present in almost every U.S. energy market, it is important that the settlement interval length follow the same interval length as the scheduling to incentivize suppliers to provide the flexibility that is needed by the market operator. On the other hand, numerous uninstructed deviation penalties and ex-post pricing rules may also
incentivize the supplier to follow the efficient schedules when hourly average prices are used for settlements. There could also be other reasons that require hourly settlements, such as data retention and storage, as well as a desire to limit market complexity. However, from this
simplified example, it appears likely that settlements that match the interval length are the most efficient for extracting the desired flexibility needed from market participants.
Table 4-5. Revenue, Costs, and Profits of Different Scenarios With 5-Min Settlements Versus Average Hourly Settlements
Scenario Total Revenue Total
Cost
Total Profit
(Total Revenue Minus Total Cost) 5-Min Settlements Hourly Average Settlements 5-Min Settlements Hourly Average Settlements Scenario 1 (Market) $15,208 $13,338 $10,895 $4,313 $2,443 Scenario 2 (Moving Hourly Average) $17,479 $17,441 $13,373 $4,106 $4,068 Scenario 3 (Perfect Knowledge of Hourly Average) $17,134 $17,134 $13,053 $4,081 $4,081