Puerto Rico Generation

Methodology

Source — PREPA's operational generation feed (operated by Genera PR), the data behind genera-pr.com/data-generacion. A snapshot is captured about every 15 minutes and stored. "Now" figures read the most recent snapshot; history and hourly baselines read the stored snapshots.

Grid adequacy indicator — A supply-adequacy indicator computed from the feed, not an official PREPA/Genera declaration. It combines two tests: (1) Forecast balance — the feed's next-hour demand forecast vs. available capacity; a projected shortfall means load-shedding (relevos de carga) may be needed. (2) N-1 reserve — whether operating reserve covers the single largest online unit (the N-1 reserve benchmark). This is a reserve comparison, not a full contingency simulation. Critical = projected shortfall or near-zero reserve; Tight = reserve below the largest online unit; Adequate = reserve covers it.

7-day outlook — Sensadvisor's own projection of grid-supplied demand, since the feed itself forecasts only the next hour. A statistical model learns the typical daily load shape — a Fourier fit over hour-of-day plus a weekday/weekend term — from roughly the last few weeks of stored snapshots, and projects that shape forward to give a demand curve, a daily peak (the highest projected hour, which falls in the evening after rooftop solar fades), and projected daily energy. Weather is shown as context, not yet as a model input. The measured series is net of behind-the-meter rooftop solar, so over this short window apparent temperature correlates with solar output rather than clean demand and fits with the "wrong" (negative) sign; a calibrated weather sensitivity is being developed from the long monthly history and an explicit rooftop-solar adjustment before it drives the numbers. The apparent-temperature line and each day's high are therefore contextual. This is a model estimate, not an official forecast, and it sharpens as more history accumulates.

Hourly generation profile — For each hour of the day (AST), the shaded band is the 10th–90th percentile of generation across stored days and the solid line is the median; today's actual is overlaid. This shows whether the system is running above or below its usual level for the time of day, and — in the Reserve view — when reserves are typically thinnest (usually the evening peak, as solar fades and demand rises). The baseline currently spans roughly two weeks of data and sharpens as more accumulate; today is excluded from the "typical" band.

Reserves — Spinning reserve is synchronized capacity ready immediately; operating reserve is total reserve available. Both are the cushion between what is generating and what could be called on.

Units online vs. offline — The feed lists every generating unit each snapshot with its output. A unit is counted online when it is producing at least 1 MW and offline otherwise (idle, down, or — for solar — simply nighttime). The feed reports output, not the reason a unit is at zero, so "offline" does not distinguish a planned outage from a forced one from normal diurnal cycling.

◆ Critical units — Units that historically carry meaningful, dispatchable load — so their being offline is a real loss of dispatchable output, not a normal daily cycle. A unit is flagged Critical when, across the stored history, it is non-renewable, averages at least 20 MW when running, and runs at least half the time. Intermittent solar and wind are deliberately excluded (a zero at night is expected, not a loss). When a Critical unit is currently offline, its row is highlighted and the summary tallies the typical MW absent. The set of Critical units is curated by hand, but each one's "usually ~N MW" — its average output when running — updates automatically from the stored per-unit history, so the typical figures stay current as units change.

Plant & unit dispatch — Per unit: MW now (real power), MVar (reactive power; can be negative when a unit absorbs reactive power), and Cost ($/MWh). Where a running unit reports no price, the cost is shown in grey italics with a ~ prefix — imputed from the MW-weighted average of priced running units of the same plant type (the same method as the blended cost below). A unit whose type has no priced peer shows a dash.

Typical MW & last online — For each Critical unit the table shows its Typical MW (average output when running) beside its live output, and Last online — how long since it last produced at least 1 MW. A unit that is down then reads as a concrete loss against what it normally contributes, and the summary compares the typical output online now against the total the Critical fleet provides when fully in service. (Both Typical MW and Last online are read from a stored per-unit view that refreshes each snapshot; if it is unavailable, Typical MW falls back to a baked baseline and Last online shows a dash.)

Blended energy cost — The MW-weighted average price of everything generating at each snapshot: each unit's $/MWh weighted by its megawatts. A running unit that reports no price is imputed at the MW-weighted average price of other running units of the same plant type — so an unpriced gas turbine is valued like the gas turbines that do report a price, rather than being dropped and understating the figure. The small amount of fuel-free generation with no priced peer of its type is excluded. It rises when cheaper baseload is offline and pricier peaking units backfill, so it tracks the cost consequence of the same outages the units table shows.

Times shown in Atlantic Standard Time (AST, UTC−4).