Identification of Similar Loads for Electric Power Management in Smart Grid

This paper deals with the identification of highly similar loads in Smart Grids. The goal is to apply load characterization techniques to define a unique load signature. In this work is used a platform composed of four fluorescent lamps with the same technical data and from the same manufacturer. Fluorescent lamps can be driven individually or together allowing for 16 different operating configurations, this means, from no one lamp turned on to all the lamps turned on. For each configuration, 150 voltage cycles and 150 cycles of total circuit current are collected at a frequency of 25 kHz. Then, two forms of load signatures are tested, one where the signature is composed of 14 characteristics and another one, based on the entropy of Shannon and Renyi, where the signature can have 5, 15, 45 or 85 characteristics. After defining the signature, it is necessary to go through the classification system to find the lowest error rate of the identification system. In this work, five classifiers were tested. The lowest error rate found was 25.63% and is close to the literature when the scenario is composed of different loads, thus proving the efficiency of the presented method.