A complete HEMS is built in four progressive layers: Layer 1 Monitoring, Layer 2 Basic Automation, Layer 3 Predictive Control, and Layer 4 AI Optimization. Building incrementally allows homeowners to demonstrate value at each layer before investing in the next.
Layer 1 — Monitoring: Install CT clamps at the consumer unit to measure every circuit, and surface the data on a dashboard such as Grafana, Home Assistant Energy, or Emoncms. Immediate results: identifying which appliances consume the most standby power, knowing the home’s peak demand hour, and establishing a baseline for measuring future savings. Budget: THB 5,000–15,000.
Layer 2 — Basic Automation: Add smart plugs for controllable appliances, smart switches for lighting, and an IR blaster or smart thermostat for air conditioning. Write rule-based automations: lights off when room is empty, AC off when leaving home, washing machine scheduled after 22:00. Budget addition: THB 10,000–30,000.
Layer 3 — Predictive Control: Add Weather API integration so the system knows in advance whether tomorrow will be hotter than today and plans pre-cooling accordingly. Add occupancy prediction learning when the home is typically occupied. Add TOU Price Integration to automatically shift the load profile between Peak and Off-Peak windows. Budget addition: THB 15,000–40,000.
Layer 4 — AI Optimization: Deploy algorithms such as Reinforcement Learning to optimise energy decisions by learning from the trade-off between comfort, energy cost, and carbon footprint specific to that household. Non-Intrusive Load Monitoring (NILM) algorithms detect appliance anomalies automatically — identifying equipment that is consuming more energy than expected, signalling a maintenance need before it becomes a fault. Cost ranges from free (Home Assistant ML add-ons) to THB 30,000–100,000 for commercial HEMS platforms.
Questions & answers
What is the difference between a CT clamp and a Smart Meter, and which is better for residential use?
A Smart Meter is installed by the utility and measures total whole-home energy. A CT clamp is a sensor the homeowner installs at the consumer unit to measure individual circuits separately. For HEMS requiring room-level or appliance-level granularity, CT clamps are the better choice at THB 500–2,000 per channel.
How long does a Reinforcement Learning HEMS take to optimise effectively?
The first 30–60 days are typically an Exploration phase where the algorithm tests different decisions. After that, it Exploits accumulated knowledge. Well-designed systems impose comfort constraints during learning — for example, never allowing indoor temperature to exceed 27°C during the day — so the learning phase does not significantly degrade occupant comfort.
What is NILM and how does it help a Smart Longevity home?
Non-Intrusive Load Monitoring (NILM) analyses the electrical current waveform at the consumer unit to identify which appliances are operating, without a separate sensor on each device. For Smart Longevity, NILM detects when an air conditioning compressor starts drawing abnormally high current — signalling that the filter needs cleaning or refrigerant is low — before it becomes a fault requiring expensive repair.
How does a commercial HEMS such as SolarEdge or Loxone compare to Home Assistant?
Commercial HEMS platforms offer better support, seamless integration within their own device ecosystem, and clear warranties, but at higher cost and with ecosystem lock-in. Home Assistant is far more flexible and supports a wide range of devices from different manufacturers, but requires more setup skill and ongoing self-management. The right choice depends on technical confidence and whether ecosystem diversity or simplicity of support is the priority.
