Centralized Scheduling and Efficient Dispatch by the Market Operator

to be dispatched by the RTO/ISO. This allows for the market operator to ensure that the resource is flexible, but it does not guarantee how flexible the resource is. The first mechanism is

somewhat obvious but often overlooked. When suppliers participate in the pool market, the market operator will operate them at their most efficient operating point based on their offered bid-cost curve. The market operator minimizes the bid-production costs from all these bids to meet the energy and ancillary service demands subject to power system security and unit constraints. The LMP and ASCP, as discussed earlier, are calculated based on the marginal bid- based cost to provide energy or ancillary services. Therefore, the cost of supplying energy for a unit participating in the market and allowing for the market operator to commit and dispatch the supplier’s output should theoretically not be greater than the resultant price (reasons that costs can be higher than the price are discussed later in Section 4.3.4). When the price increases, the market operator gives the supplier a position that reflects that it is efficient to increase its output, and this allows the supplier to earn more revenue. When the price decreases, the market operator gives the supplier a position to reduce output, because it may be that the current output is no longer efficient when receiving the reduced price.

Under good electricity market design, the supplier output level should always reflect the changing prices and should avoid operating at levels that cost more to produce energy than the price they receive. Self-scheduled resources provide the market operator with the scheduled output before the market clears, and this schedule is fixed regardless of the price. During periods of high prices, the self-scheduled resource could miss out on additional profit. During low prices, the self-scheduled resource could lose money when the cost to supply energy is greater than the energy payments they receive. When substantial bilateral contracts are self-scheduled into the market, there may come a point at which the flexibility that is available to the market operator is insufficient, inducing a need for other mechanisms to obtain this flexibility. For example, a very high proportion of self-scheduled resources may drive the need for more expensive sources of flexibility, such as additional flexible generating capacity or storage. In cases such as this, the system may possess more flexibility than is needed; however, much of this flexibility may be stranded. It is important to note that the levels of physical flexibility may be sufficient; however, some of this may not be contractually available.

To illustrate the potential impacts of self-scheduling, we show a simple example. Table 4-1 shows a bid-in cost curve for a thermal generating unit which is taken from real bid-cost data. This bid-in cost curve reflects representative costs of thermal plants based on a convex, monotonically increasing incremental heat rate. We ignore no-load costs in this example for simplicity. The incremental cost in column 1 is the cost bid for the specific capacity represented in column 2. Therefore, the first 286 MW in this example will always cost (35*286) = $10,010.

Table 4-1. Hypothetical Thermal Plant, Piecewise Linear Cost Curve Incremental Cost ($/MWh) Energy/ Capacity Segment (MWh) 35.00 Up to 286.0 47.25 286.1–295.0 47.60 295.1–304.0 47.95 304.1–313.0 48.30 313.1–322.0 48.65 322.1–331.0 49.00 331.1–340.0 49.35 340.1–349.0 49.70 349.1–358.0 50.75 358.1–376.0 52.50 376.1–377.0

The cost data in this table forms the basis of how this resource would bid into the market. We next turn to the relationship between this cost data and LMPs and an examination of how various self-scheduling strategies compare to how the unit would be dispatched in the absence of

self-scheduling.

Table 4-2, column 2, shows a 12-hour period of LMPs. Scenario 1 (Market) allows the market operator to efficiently dispatch the resource every hour. For simplicity, ramp rates and other constraints that may cause inefficiencies are ignored.12 In nearly every time period, the unit’s output changes as a function of the LMP. This is in contrast to each of the self-scheduling

scenarios shown in Scenarios 2 through Scenario 5. In Scenario 2 (Min Self), the supplier simply schedules itself at its minimum capacity level for all hours. In Scenario 3 (Max Self), the supplier schedules itself at its maximum capacity. In Scenario 4 (Mid Self), the supplier schedules itself at a level in between its minimum and maximum capacity. Finally, in Scenario 5 (Lag LMP), the supplier uses the LMP from the previous hour to predict where it should schedule itself for the following hour.

Table 4-2. Twelve-Hour Example for Allowing the Market Operator to Efficiently Dispatch the Output of a Resource (Scenario 1) Versus Various Self-Scheduling Techniques (Scenarios 2–5)

Hour LMP ($/MWh) Scenario 1 Market Scenario 2 Min Self Scenario 3 Max Self Scenario 4 Mid Self Scenario 5 Lag LMP 1 $45.41 286 286 377 300 286 2 $49.65 349 286 377 300 286 3 $52.27 377 286 377 300 349 4 $51.37 376 286 377 300 377 5 $48.32 322 286 377 300 376 6 $46.45 286 286 377 300 322 7 $46.35 286 286 377 300 286 8 $50.97 376 286 377 300 286 9 $49.44 349 286 377 300 376 10 $44.70 286 286 377 300 349 11 $48.51 322 286 377 300 286 12 $51.13 376 286 377 300 322

Table 4-3 shows the revenue, cost, and profit results for all five scenarios. The profits from each of the self-scheduling cases are compared to Scenario 1: Market. Thus, the right-most column shows how much profit the supplier loses by not offering its flexibility into the market. Note that each of the self-scheduling cases results in lost profits compared to the market case. Even with intelligence in the self-scheduling strategy (Scenario 5), it would still lose out on $812 during a 12-hour period. Although the lost profit is small relative to the total profit, it will be highly dependent on the cost curve and prices during different time periods. For example, if the cost for the first segment of Table 4-1 (up to 286 MW) were $47 rather than $35, the profits lost would be the same as Table 4-3, but the total profits would be an order of magnitude less, making the relative profit loss much more significant.

Table 4-3. Revenue, Cost, Profit, and Profit Lost for Various Self-Scheduling Techniques

Scenario Total Revenue Total Cost

Total Profit (Total Revenue Minus

Total Cost) Profit Lost Scenario 1 Market $195,668 $147,313 $48,355 N/A Scenario 2 Min Self $167,326 $120,120 $47,206 $1,148 Scenario 3 Max Self $220,567 $173,594 $46,972 $1,382 Scenario 4 Mid Self $175,517 $128,079 $47,438 $916 Scenario 5 Lag LMP $190,452 $142,909 $47,542 $812