ISO Perspective

Figure 14.10depicts an example day-ahead forecast for CIMIS station 173 (Torrey Pines, California) for the day presented in Figure 14.9 (June 11, 2011). This forecast was created using WRF with cloud assimilation, initialized at 12 UTC on June 6, 2011. Hypothetically, this forecast would have been provided to the balancing authority a day ahead (June 10,2011,0900 PST) for planning the DAM.

image400

FIGURE 14.9 GOES satellite imagery: (a) June 11, 2011, compared to the intraday (0-24 h, initialized on June 11, 2011, 12 UTC) (b) and day-ahead (24-48 h, initialized on June 10, 2011, 12 UTC); (c) WRF-CLDDA irradiance forecasts in San Diego, California. This figure is repro­duced in color in the color section.

image401

provided to the ISO on June 10, 2011, showing bias-corrected irradiance with an 80% uncertainty interval: (a) probability of a large ramp event occurring; (b) predicted hourly ramp rate; (c) fore­cast cutoff at 1500 PST because the 36 h forecast horizon ends at that time.

On this day, the day-ahead forecast irradiance ramped up dramatically in the morning hours (0800-1000 PST; Figure 14.10a). Correspondingly, the hourly ramp-rate magnitude was greater than 5 W/m~2 (Figure 14.10c) and the proba­bility of a ramp event occurring increased to nearly 80%. For the ISO, this, in conjunction with demand forecasts, would be used to make initial production commitments in the DAM. Furthermore, the uncertainty limits should be considered to determine the likely minimum energy to be produced on this node. Since solar production was expected to be small prior to 0800 PST, more conventional energy would need to be procured in the morning hours. Addition­ally, the late-morning up-ramps indicated that extra production would be coming online and the ISO would have to adjust schedules downward accordingly. This scenario assumes constant demand. In reality, the solar up-ramp would be concurrent with a demand up-ramp and would therefore be convenient to the ISO.

As the real-time market approaches, an intraday forecast (current day) is provided and used to update production estimates (Figure 14.11). Compared to the initial commitments in the DAM, the more accurate intraday forecast is used to predict and account for potential energy imbalances in the RTM. Similar to the day-ahead forecast, the RTM forecast indicated that morning cloud cover would limit solar-energy production before 1000 PST. However, the updated forecast indicated that thick clouds would persist for much longer and irradiance would not increase significantly until 1000-1200 PST (Figure 14.11a). Correspondingly, large positive ramps were expected during these times (Figure 14.11b, c) as irradiance (and solar-energy production) increased. Additionally, a second, negative ramp was expected for the late afternoon hours (1300-1500 PST) as clouds returned. Since the RTM forecast differed significantly from the initial DAM commitments, the ISO could

image402

UTC on June 11, 2011.

modify unit-commitment and reserve plans based on the updated information. Such changes would be proportional to the change in expected irradiance from the DAM to the RTM (Figure 14.12).

In the early morning hours (0600-0800 PST), the day-ahead forecast was similar to but slightly smaller than the intraday irradiance forecasts (Figure 14.12). Since more solar energy was produced than was initially pre­dicted, a slight oversupply existed. Consequently, grid reliability was guaran­teed and no additional procurement was needed. However, some resources are wasted in procuring too much energy in the DAM. For the late morning hours (0900-1100 PST), the RTM forecast predicted much less irradiance than the DAM forecast, indicating that it was unlikely that the DAM commitment would

image403

FIGURE 14.12 Change in forecast irradiance from the initial DAM forecast to the RTM forecast on June 11, 2011.

be met. Therefore, energy had to be procured in the RTM. Depending on the RTM-to-DAM price ratio, the 0900-1100 hours could represent significant monetary loss for the plant operator (refer to Figure 14.2). If LMPrtm > LMPdam, the initial overprediction causes significant revenue loss. If LMPrtm < LMPdam, energy still needs to be procured but at a lower cost than the cost at which it was sold day-ahead, resulting in a profit.

Updated: August 23, 2015 — 10:47 pm