Forecast-based automatic scheduling of a distributed network of thermostats with learned adjustment

Abstract

Heating and cooling systems at various geographical locations are controlled by a central energy management service unit to maintain comfortable indoor temperatures. In some weather conditions, people may intuitively prefer a slightly warmer or cooler indoor temperature. In systems equipped with environmental learning capabilities, an apparent outdoor temperature is determined based on the geographic location itself, the season at the geographic location, the forecasted actual temperature, and one or more seasonal weather factors such as wind velocity or humidity. The apparent temperature and a trained machine learning system are used to select an automated schedule for the geographic location to be directly transmitted to devices at the location. The automated schedule can vary from typical schedules by causing the heating and cooling systems to maintain a temperature that is slightly warmer or cooler than typical indoor temperatures.