Chilean regulation recognizes each technology's contribution to adequacy through heuristics. For wind and solar PV, the rule is the minimum between the lowest annual capacity factor of the last five years and the average contribution during the 52 peak hours of the current year. That mechanism breaks down once storage enters: a PV plant that charges a battery precisely during peak hours —and therefore matches net demand better— receives an accreditation of 0 MW, even though the combined system responds better to scarcity events than the plant alone.
This document discusses whether probabilistic methods can settle that problem, and what would have to be defined before applying them in Chile. ELCC (Effective Load Carrying Capability) measures how much additional demand the system can take on when a plant is added while holding the previous reliability level; ECP (Equivalent Conventional Power) looks for the equivalent thermal plant that would replace it without altering that level. Both are technology agnostic and rest on reliability metrics such as LOLP, LOLE or EENS.
What the 2025 estimates show
SPEC simulated the National Electric System for 2025 with a target LOLE of 7.6 hours/year, hydrology based on the last 60 years of records, renewable profiles from the last five years, and clustering of plants with at least 300 MW of accumulated capacity. Under marginal ELCC, solar PV accreditation falls to 1.7% of maximum capacity, against the 16% reference value for 2019, while reservoir hydro rises from 49% to 89%, wind from 15% to 22% and run-of-river from 48% to 59%; thermal units stay at 77%.
The cause lies in where the risk concentrates: loss-of-load probability clusters between May and August, in the 19:00 to 23:00 window, because of lower reservoir levels and lower solar output. With around 6,400 MW of installed PV, the system is highly reliable during sunlight hours. A near-zero marginal contribution does not mean a zero contribution: measured as average ELCC, the full PV fleet contributes close to 13% of its installed capacity.
Interactions that complicate the design
The exercise finds diminishing returns for storage: as storage penetration grows, its own marginal ELCC converges toward values close to zero, while accreditation for PV projects rises to levels around 10%. Sensitivity also appears with respect to stored water and to the month of maintenance: a thermal unit out of service during the months of highest LOLP receives lower accreditation.
What the document recommends defining
- An explicit reliability metric and target, whether by regulatory decision —such as the US one-in-ten criterion, LOLE of 2.4 hours/year— or through the economic route followed by ENTSO-E, combining VOLL and CONE.
- Whether accreditation uses marginal ELCC, which gives the correct signal about the need for new capacity but falls quickly and volatilely, or average ELCC, more stable but at risk of overvaluing and overcompensating.
- Modeling rules: representation of cascaded reservoirs, irrigation agreements and non-electric uses, clustering of units, Monte Carlo convergence limits, and a granularity that balances precision, computational load and administrative simplicity.