2. Functional blocks and roles

Cloud ( with Ai function )

• Learn indoor temperature profile of various scenerios…
• Take outdoor temperature online for prediction
• Deep Neural Network to do prediction, override commands to tstats…. • Calculate energy saving % based on prediction vs actual…
• Save data for statistics….
• Etc etc…..

Gateway

Quantum tstat ( upgraded firmware )

TRV or actuators

Salus Premium Lite apps ( with Enhanced Intelligence option )

3. Strategy of energy saving calculation

Total Energy consumption is defined by relay ON duration (sec )

Lab measured Energy saving % = Actual ( non-Ai ) – Actual (Ai) Field calculated Energy saving % = Predicted – Actual (Ai )

Predicted energy consumption is estimated by its learnt temperature history. The longer it learns, the more accurate its  prediction of the energy consumption is. 

Objective: While maintaining the comfort by reaching the temperature setpoint, Ai aims at reducing the overall relay ON duration, hence saving energy.

Based on the learnt temperature data, Deep Neural Network in cloud:

1. Predict the temperature rise rate, possible overshoot caused by relay ON time
2. Cut short the ON time in advance
   
3. Stop ON cycle
   
4. Remove unnecessary ON cycle
   
5. Early OFF schedule
   
6. Etc. etc..
   
   

5. Factors affecting energy saving % calculation

1. Prediction relies on historical temperature data
2. Tstat basic tpi or span performance is good or bad
   
3. Energy saving is counted into calculation or not
   
4. Environmental changes (e.g. hot water temp, window open, outdoor temp )
5. Occupancy of the premise
6. Etc etc…. 
   

6. Lab and Field test result of energy saving %

According to WL’s TRV radiator tests at Cincinati lab, UFH tests at CoE Romania,

AVERAGE ENERGY SAVING BY AI = 8% 

Range of energy saving by AI recorded 0 – 30%, depends on conditions.

N.B. Sales claims energy saving % up to 20% plus ?