Processing of smart meters data for peak load estimation of consumers

The paper presents a Mining Data based approach for peak load estimation of household consumers. The approach uses an unsupervised learning method (clustering) along with the polynomial regression. With K-means clustering algorithm, the consumption categories of household consumers were determined. The input data for the consumers' classification in consumption categories (monthly energy consumptions and peak loads) are obtained from processing the typical load profiles provided by Smart Meters. For each consumption category, a polynomial regression model is built for peak load estimation of household consumers equipped with classic meters. The obtained results demonstrate that the methodology can be used with the success in peak load estimation for all household customers from distribution systems when information is very poor (based on the data provided by classic meters).