As we move through the summer months and solar production becomes increasingly important to power market operations, forecasting accuracy plays a critical role in helping market participants prepare for volatility and manage risk. With this in mind, we reviewed the performance of the Enverus day-ahead solar forecast across the two U.S. ISOs with the highest solar generation: ERCOT and CAISO.
The Enverus day‑ahead STPF solar generation forecast outperformed ISO forecasts across the two ISOs with the highest solar generation during the May–June 2026 period (using day‑ahead rolling forecasts published at 8 AM). The strongest improvements were observed during periods of elevated weather-driven variability, including heat waves and significant cloud-cover events, where ISO forecasts exhibited considerable bias. Table 1 summarizes total MAE and CAP_MAE for Enverus forecasts versus ISO forecasts, with Enverus outperforming the ISOs in every month analyzed.
| Region/ISO | Month | Total MAE (MW) | Cap_MAE (%) | ||
|---|---|---|---|---|---|
| Enverus | ISO | Enverus | ISO | ||
| ERCOT | May | 1214.05 | 1221.95 | 3.54 | 3.57 |
| ERCOT | June | 1144.8 | 1403.28 | 3.34 | 4.10 |
| CAISO | May | 971.43 | 1081.56 | 4.22 | 4.70 |
| CAISO | June | 708.64 | 930.02 | 3.08 | 4.04 |
Table 1. Total MAE: May–June 2026
Extreme Weather Period Analysis
The Enverus STPF (Short-Term-Production-Forecast) forecast demonstrated its largest performance advantage during periods of challenging weather conditions, including the late-June ERCOT heat wave and an early-May CAISO solar generation decline associated with extensive cloud cover.
ERCOT Heat Wave (June 18-30, 2026)
ERCOT experienced a prolonged heat wave from June 18–30, with peak system demand approaching 82 GW on multiple days. During this 13-day period, the Enverus day-ahead solar forecast achieved an MAE of 1,060.64 MW, substantially outperforming the ISO forecast MAE of 1,704.50 MW. Enverus delivered lower forecast error on 12 of the 13 days analyzed. During this event, the ISO forecast consistently overestimated solar generation, while the Enverus forecast more accurately captured the solar generation pattern.
| Period | Total MAE (MW) | Cap_MAE (%) | ||
|---|---|---|---|---|
| Enverus | ISO | Enverus | ISO | |
| June 18 – June 30 | 1060.64 | 1704.5 | 3.10 | 4.98 |
Table 2. One day‑ahead ERCOT system‑wide solar forecasts vs. ISO
CAISO Low Solar Output Event (May 5–6, 2026)
CAISO solar generation declined significantly on May 5–6 relative to surrounding days. Weather conditions across portions of California during this period included widespread marine-layer cloud cover and overcast conditions, which likely contributed to the reduction in solar output. Across the two-day event, the Enverus day-ahead solar forecast achieved a combined MAE of 411.29 MW, substantially outperforming the ISO forecast MAE of 1,628.98 MW. Enverus delivered lower forecast error on both days analyzed. The ISO forecast significantly overestimated solar generation during the event, particularly on May 6, whereas the Enverus forecast more accurately anticipated both the magnitude and timing of the reduction in solar output.
| Period | Total MAE (MW) | Cap_MAE (%) | ||
|---|---|---|---|---|
| Enverus | ISO | Enverus | ISO | |
| May 5 | 306.58 | 1294.35 | 1.33 | 5.62 |
| May 6 | 516.01 | 1963.60 | 2.24 | 8.53 |
Table 3. One day‑ahead CAISO system‑wide solar forecasts vs. ISO
Figure 1 illustrates the sharp decline in CAISO solar generation on May 5–6, followed by a recovery beginning on May 7. The Enverus forecast accurately anticipated the magnitude of this decline compared to the ISO forecast, resulting in substantially lower forecast error during the event.
ERCOT 4CP Call-Out – June 2026
The Coincident Peak (CP) in ERCOT for June occurred during Hour Ending 18 on June 18. Enverus accurately identified this interval one day in advance, giving customers valuable time to prepare and respond. Earlier in the month, our latest demand forecasts and analysis indicated that the peak demand was more likely to occur during the second half of June. As a result, we did not issue any CP alerts at the early beginning of the month.
In total, Enverus issued:
- 5 CP-related calls throughout June
- Covering 8 high-demand hours
This means we issued calls for 16.67% of the days, and just 1.11 % of the total hours of the month of June.
On June 19, the Enverus day-ahead load forecast achieved a daily MAPE of 1.12%, compared with 1.76% for the ISO day-ahead load forecast, reinforcing the accuracy of our CP call made the previous day. These results highlight the precision and adaptability of our forecasting tools.
Conclusion
Across both renewable generation and load forecasting applications, Enverus day-ahead forecasts consistently demonstrated superior accuracy relative to ISO forecasts during May and June 2026.
The performance advantage was particularly evident during periods of elevated weather-driven volatility, including the late-June ERCOT heat wave and the early-May CAISO solar generation reduction. In ERCOT, Enverus successfully identified the June Coincident Peak one day in advance with very few calls. The strong performance of the day-ahead load forecast during this period further reinforced the reliability of the forecasts used to support CP decision-making.
These results demonstrate the value of our forecasting tools and analysis, particularly during high‑volatility periods, ultimately supporting improved market outcomes, risk management, and peak-demand response strategies across North American power markets.
Note: CAP_MAE represents the Mean Absolute Error (MAE) scaled by the maximum observed value (CAP) in the dataset. This scaling produces a relative error measure, ensuring that the system’s overall scale and the magnitude of values during the evaluation period do not distort the error metric.
