Generating Load Profiles Using Smart Metering Time Series

In this work we present a practice-oriented approach for generating load profiles as a means to forecast energy demand by using smart metering time series. The general idea is to apply fuzzy clustering on historic consumption time series. The segmentation yielded helps electricity companies to identify customers with similar consumption behavior. This knowledge can be used to plan available energy capacities in advance. What makes this approach special is that this approach segments consumption time series by time in addition to identifying customer groups. This is done not only to accommodate for customers potentially behaving completely different on working days than on local holidays for example, but also to build the resulting load profiles in a way the electricity companies can adapt with minimal adjustments. We also evaluate our approach using two real world smart metering datasets and discuss potential improvements.