Non-Intrusive Load Disaggregation of Air Conditioning Systems

Non-Intrusive Load Monitoring (NILM) makes it possible to estimate the energy consumption of devices in a household using only one sensor at power entry. HVAC Systems takes a lot of energy from households to work and a miss-use or a faulty operation can cause energy waste. Applying NILM approaches to HVAC systems allows to disaggregate these devices energy consumption and help to understand how they are operating. The scope of this thesis is to create models that can disaggregate the energy of electric boilers and two types of heat pumps (intense heat pumps and extended heat pumps). Since these devices differs a lot in their behaviors, three different models had to be created. The electric boilers model needs a training set to estimate the peak power of functioning. The intense heat pumps model doesn't need a training set but it has parameters that need to be tuned. The extended heat pump model needs a training set to estimate the base consumption and it has parameters that need to be tune to recognise if the heat pump is on or off. All the data used for this thesis was given by Midori Connect.