Switching state prediction for residential loads with weather data for smart automated demand response

Impact of renewable energy based generation is very much visible on the framework of a demand response system. Dependency of small, isolated power networks (micro grids) on the main utility grid is also reduced with the increase in renewable energy injection in the power system. Mitigation of power imbalance by load management is essential for islanded systems having intermittent generation like photovoltaic systems. Since such systems can sufficiently form an island at domestic level, understanding the pattern of residential load switching becomes important for maintaining power balance. Switching state of any home appliance is fully dependent on the behaviour of the occupants. Human behaviour is thus the controlling parameter. Weather is one of the element on which it depends and accordingly important features of weather data have been selected for the prediction of loads' switching state. This work discusses the prediction of On/Off states for specific domestic loads by both time series prediction and classification techniques with weather data as input features. Estimation accuracy of switching was low for few loads which were generally critical loads with automatic power cut controller however non-critical loads showed correct prediction of switching states. This load-learning can be applied for implementing smart automated demand response.