Operating-schedule inference

camber.schedule infers the building's actual weekly operating schedule from an interval load (or fan-status) series — what hours it really runs, versus what the schedule claims. It drives setback verification, demand-response eligibility, and onboarding (a detected schedule seeds the occupancy model).

Inference flow: detect_schedule thresholds the load, takes the per-hour majority into a WeeklySchedule, then compare_schedule diffs it against the stated hours.

flowchart LR
  load["Interval load_kw / fan-status series"] --> thr["Threshold = base + 0.5 x (peak - base), from 10th/90th pct"]
  thr --> mark["Mark each interval on/off"]
  mark --> maj["Majority state per hour-of-week (168 slots)"]
  maj --> sch["WeeklySchedule (days, occupied_fraction)"]
  stated["Stated schedule (weekday 9-5)"] --> cmp["compare_schedule"]
  sch --> cmp
  cmp --> out["extra_runtime_slots, n_missing, agreement"]

from camber.schedule import detect_schedule, compare_schedule

sch = detect_schedule(load_kw)  # threshold defaults to midway between base and peak
sch.days[0].start_hour, sch.days[0].end_hour  # Monday on-period
sch.occupied_fraction  # share of the 168 hour-of-week slots that are on

stated = [(d, h) for d in range(5) for h in range(9, 17)]  # weekday 9–5
cmp = compare_schedule(sch, stated)
cmp["extra_runtime_slots"], cmp["n_missing"], cmp["agreement"]

The method is transparent: mark each interval "on" above a threshold (default base + 0.5·(peak − base) from the 10th/90th percentiles), then take the majority state per hour-of-week across all weeks. compare_schedule returns the extra-runtime slots (running when it shouldn't — a setback opportunity), the missing slots, and the agreement fraction. Flags: threshold, on_level, min_fraction.