Energy Consumption Patterns and Load Forecasting with Profiled CNN-LSTM Networks

By virtue of the steady societal shift to use smart technologies built on increasingly popular grid framework, we have noticed an increase in need analyze household electricity consumption at individual level. In order work efficiently, these rely load forecasting optimize operations that are related energy (such as appliance scheduling). This paper proposes a novel method utilizes clustering step prior group together days exhibit similar patterns. Following that, attempt classify new into pre-generated clusters by making available context information (day week, month, predicted weather). Finally, using historical data (with regard consumption) alongside meteorological and temporal variables, train CNN-LSTM model per-cluster basis specializes based profiles present within each cluster. leads improvements performance (upwards 10% mean absolute percentage error scores) provides us with added benefit being able easily highlight extract allows identify which external variables effect any household.