Load profiles identification based on autoencoders and Kohonen Maps

Load profiles are a crucial tool for power system planning and operation, and also in several operations of electricity markets. This article proposes a new methodology for the determination of load profiles based on a two-step approach. The first phase employs a neural network autoencoder to reduce the dimensionality of the input vectors. The second phase is a clustering process based on the Kohonen Self-Organizing Maps, to identify cohesive consumers´ classes. The implemented approach produces classes based on load diagrams and, simultaneously, a class identification based on consumers´ billing data.