The aim of this paper was test the performance of different features and four clustering algorithms: K-Means, Hierarchical clustering, Density-based spatial clustering and Gaussian Mixture Models applied to load curve characterization with real data of energy consumption in Colombia. Clustering algorithms were compared using a common data mining methodology, where K-Means produced the best performance. The results show that it is possible to implement machine learning algorithms using different types of features, in order to obtain relevant information for the energy sector. Also, this work presents possible applications using load curve characterization.