wattlab-eplus-demand-hourly

name	wattlab-eplus-demand-hourly
description	Run EnergyPlus hourly demand-management sims (shed, deadband, plant/HVAC off, precool load-shift) on the G14 Twin and feed ML/Unity DM twins. Use EnergyPlus-MCP as the IDF inspect/modify helper. Triggers on: demand management, load shed, load shift, precool, deadband, DR window, hourly kW, july_demand_profiles, Unity demand twin, synthetic DM training data.

EnergyPlus hourly demand management (MCP helper)

Question the farm answers: given today’s outdoor conditions, what is the hourly facility kW when we change HVAC controls?

Twin SoT

• Run: geo_b100_dual_ahu_shape_ops11 (Building 100 dual-AHU, G14 PASS)
• IDF: runs/.../model.idf (also packaged under vibe21 assets/twin_b100_ops11/)
• AMY: ...__stage_in/amy.epw

Runner

# Multi-day ML farm — native Windows EnergyPlus (preferred when Docker/WSL broken)
python /path/to/vibe_code_apps_21/tools/dm_hourly_farm.py --engine native --smoke
python /path/to/vibe_code_apps_21/tools/dm_hourly_farm.py --engine native --reuse-existing
python /path/to/vibe_code_apps_21/ml/tune_demand_hourly.py
python /path/to/vibe_code_apps_21/ml/writeup_demand_hourly.py
# Exe: C:\EnergyPlusV26-1-0\energyplus.exe  or ENERGYPLUS_EXE=

# Interim without EnergyPlus:
python .../dm_hourly_farm.py --from-seed-proxy

# Legacy single-day July portfolio (Docker)
docker exec vibe20 python /data/tools/july_demand_profiles_eplus.py --reuse-existing

Output (ML farm): reports/dm_hourly_farm/dm_hourly_rows.parquet
Writeup: reports/dm_hourly_farm/writeup/DEMAND_HOURLY_SURROGATE_RESULTS.md
Output (July seed): reports/full_parity_july_demand/july_demand_profiles.json
Playbook: tools/DEMAND_MANAGEMENT_HOURLY_PLAYBOOK.md
Unity massing: tools/export_unity_twin_manifest.py
Unity Editor project: vibe_code_apps_21/unity/<ProjectName>/

Implemented strategies

Mode	Story
setpoint_raise	Clg + DAT +°F in window
deadband_widen	DB ~5→10°F cooling-biased + DAT follow
chiller_off	CHW AvailabilityManager OFF
hvac_off	FanAvail + CHW OFF
precool_shift	morning precool then afternoon DB relax (load shift)
precool_chiller_off	precool then plant OFF through peak

EnergyPlus-MCP role

Use wattlab-energyplus-mcp / energyplus-mcp-dev to:

1. inspect schedules, AvailabilityManagers, Dump DAT, FanAvail before patching
2. modify when adding new DR knobs (DSP truncate, OA soak, capacity limit)
3. simulate / validate eplusout.csv Electricity:Facility Hourly
4. confirm comfort/unmet hours if labeling soft vs hard shed

docker exec -e WATTLAB_STUDIO_WORKSPACE=/data \
  -e WATTLAB_HOST_WORKSPACE=$HOME/wattlab_workspace vibe20 \
  wattlab energyplus-ensure

ML / Unity contract (keep narrow)

• Context: OAT/RH/solar/hour/dow/occupied
• Actions: precool_f, relax_clg_f, deadband_f, chw_avail, fan_avail, dat_delta_f
• Target: hourly facility_kw (+ optional cooling/fan kW, max zone T)
• Not in scope for DM twin v1: annual ECM package savings, FDD classifiers, gas-led ECMs

Excel Demand tab oracle: ECM_FULL_PARITY.xlsx. Product Excel long-term = open-fdd PyPI ECMJob.