Comprehensive Benhcmark System for Load Management Applications Using Morphological-based Load Profile Clustering

One of the main concerns in building demand management applications such as peak shaving, energy efficiency, tariff design and demand response is the limited knowledge from different building-type load patterns, their specifications, and features. In this regard, designing a reliable and comprehensive benchmark system for load profiles would be a critical primary task. Such an encyclopedic database plays a critical role in evaluating different demand side management approaches for buildings. In this work we aim to define an appropriate benchmark system to assess efficacy of buildings load management methods by classifying a large yearly dataset of various load types including commercial, educational, industrial, and grocery buildings. For this purpose, first, daily load patterns of each building are considered as individual load samples. Then, a morphological filtering procedure is used in order to segmentize each of these load samples. Next, a set of energy-based features is extracted from segments and is fed to a hierarchical clustering algorithm to partition this enormous dataset into an optimal number of classes. Finally, each building is assigned to different categories based on the total number of load classes it may include.