Full Library

Analyzing thermostat data together with outdoor weather to predict indoor temperature can help smart thermostats optimize the operation of a residential building's HVAC system and make buildings more comfortable. Using weather forecasts, these smart thermostats can become proactive in responding to actual and future outdoor conditions. Using the ecobee thermostat data from sixteen Canadian and US houses, the prediction accuracy of the generalized regression neural network (GRNN) algorithm and the resilient back propagation neural network (ANN) algorithm were evaluated. Solar radiation was added to the data available from the smart thermostat to account for the effect of passive heating. GRNN proved to be a better predictor with a lower MSE and a maximum of 7.5°C deviation from actual measurements.

openalex: W4236330255