An intelligent relational pattern matching system for electricity demand prediction

The forecast of electricity consumption is a key element to develop successful policies for electricity demand management. A significant variable affecting the demand of electric energy in commercial and industrial building is the outdoor temperature. In this paper, an intelligent relational pattern matching system is proposed to forecast electricity demand using smart meter data and outdoor temperature profiles. In order to identify the relationship map between the patterns of power consumption and temperature, a learning rule based approach is developed to incrementally learn the correlation between both pattern bases. One-week-ahead forecast is executed by a similarity search method using the found relationship map. The effectiveness of the proposed system and prediction method is verified using real smart meter data from a commercial building, and weather forecasts from the Australian Bureau of Meteorology.