Hierarchical Structure Based Energy Consumption Forecasting in Top-Down Approach

The increasing penetration of roof-top Photovoltaic (PV) panels in residential consumers has made electricity usage unpredictable. The challenge ahead for utility operators is to manage the operation and maintenance of a plant efficiently under the given circumstances. Therefore, the utility operators need to estimate the load so that there is no significant demand and supply mismatch. In the residential sector, the power consumption varies in every household, significantly depending upon the socio-demographical features, e.g., income, family size, age of resident etc. It is crucial to understand the relation of these socio-demographics on electricity consumption for its accurate forecasting. Besides, the smart meters in residential sector has made it possible for the household consumption to be known. A good estimation of load consumption at different levels is very important for efficient load management. Therefore, this paper focuses on developing a hierarchical structure by mining the electricity metered data to form customer clusters based on energy consumption patterns. To improve the forecasting performance, every cluster is associated with its distinct socio-demographical features. Based on the analytical results, it is evident that the hierarchical structure including socio-demographical features has the prospect to forecast load using a top-down approach, with an improvement in its mean absolute percent error by 9.04%, as compared to state-of-the-art bottom-up approaches.