Grid-Aware Energy Load Shifter

Shift heavy residential loads to the cheapest electricity hours using Home Assistant energy data.

Quick Start

CODEBLOCK0

Connection

Two paths to reach Home Assistant:

1. 1. MCP (preferred): If the HA MCP server is configured, use mcporter call homeassistant.<tool> directly.
2. REST API: Use python3 {baseDir}/scripts/ha_bridge.py. Requires HA_URL and HA_TOKEN environment variables.

Security

Required credentials:

Least-privilege recommendations:

• - Create a dedicated Home Assistant user account for this skill (e.g. openclaw-energy)
• Generate a Long-Lived Access Token from that account only
• Limit the account's entity access to energy-related entities if your HA setup supports entity-level permissions
• Test with read-only commands first (discover, energy-summary) before enabling device control

Domain allowlist: The call-service command restricts actions to energy-related domains only: switch, automation, script, climate, water_heater, input_boolean, input_number, number. All other domains (e.g. lock, alarm_control_panel) are blocked with exit code 2.

Commands

All commands output JSON to stdout.

Load-Shifting Workflow

Follow these steps when asked about energy optimization:

1. 1. Discover available energy entities: run discover or INLINECODE32
2. Read prices: Check pricing entities' state and attributes — look for:

1. 3. Identify deferrable loads: Find switch.* entities for schedulable devices (EV charger, pool pump, dishwasher, washer/dryer, water heater)
2. Find the cheapest window: Scan hourly prices for the contiguous N-hour block with the lowest sum (N = estimated run time of device)
3. Execute: Call switch/turn_on at the optimal time, or automation/trigger if the user has an existing automation

Interpreting Price Data

Different integrations expose prices differently:

• - Hourly arrays (Nordpool, ENTSO-e, Octopus): Read today/tomorrow attributes → find cheapest hours
• Price level (Tibber): Read price_level → act when CHEAP or VERY_CHEAP
• Real-time (Amber Electric): Read 5-minute pricing → shift loads immediately when cheap
• Utility meter tariffs: Read sensor.*_peak vs sensor.*_offpeak → user's HA automations switch tariffs at configured times
• Static TOU: Read current_price attribute → compare against historical average

Cost Savings Estimate

When recommending a shift, show estimated savings:

CODEBLOCK1

Solar Self-Consumption

If solar sensors exist, align loads with peak production:

• - Read sensor.forecast_solar_* or sensor.solcast_* for today's forecast
• Shift loads to hours with highest expected production
• This avoids grid import entirely — savings = full retail rate × kWh shifted

HVAC Pre-Conditioning

HVAC is the largest residential load (40-50% of electricity). Pre-cool or pre-heat during cheap/solar hours so the home coasts through expensive peak periods:

1. 1. Read climate.* entities for current HVAC mode and setpoint
2. During cheapest window: lower cooling setpoint by 2-3F (pre-cool) or raise heating setpoint by 2-3F (pre-heat)
3. During peak window: raise cooling setpoint by 2-3F to coast on thermal mass
4. Savings estimate: 1.5-3 kW shifted × price differential × hours

Water Heater Scheduling

Electric water heaters (4.5 kW typical) are ideal deferrable loads:

1. 1. Find switch.water_heater or water_heater.* entities
2. Heat during cheapest/solar window to full temperature
3. Turn off during peak hours (tank maintains temperature for 4-6 hours)
4. Savings estimate: 4.5 kW × price differential × 3-4 hours/day

Battery Arbitrage

If home battery entities exist (sensor.battery_soc, sensor.powerwall_*, sensor.enphase_*):

1. 1. Read current state of charge and charge/discharge rate limits
2. Charge from grid during cheapest hours (or from solar)
3. Discharge to home during peak price hours to avoid grid import
4. Advanced: If battery supports grid export and VPP enrollment, discharge to grid during extreme price events ($2,000+/MWh)
5. Savings estimate: batterycapacitykwh × (peakrate - valleyrate)

Demand Response / VPP Integration

For homes enrolled in utility demand response or virtual power plant programs:

1. 1. Read demand response signal entities (if available via HA integration)
2. When DR event active: shed non-critical loads, pre-cool/pre-heat, discharge battery
3. Estimate DR payment: kW reduced × event duration × program rate

Entity Reference

For detailed entity patterns across providers, read: energy_entities.md