Reducción de espacio de búsqueda usando algoritmos de aprendizaje no supervisado aplicado al problema de la expansión del sistema de transmisión de energía eléctrica

Abstract

The current demand for electrical energy requires infrastructure additions to transmission systems that are becoming larger and larger. A transmission system expansion planning (TEP) is responsible for identifying necessary additions to the power system. However, this process is difficult due to the number of variables to deal with. These variables are the result of the number of candidate lines to consider (search space) in an optimization model. In this paper, a candidate line clustering strategy is presented with the objective of reducing this search space. This strategy combines unsupervised learning tools such as clustering and the analysis of the operation of the Electric Power Supply System (ESS) called optimal power flows. This combination is used to classify the candidate lines under 3 criteria: overload, least effort and cost-benefit. These criteria are applied to each study system under analysis: Garver 6 Buses, IEEE 24 Buses and IEEE 188 Buses, to determine the most appropriate in each case. All this in order to form and identify unrepresentative line groups to discard, resulting in a new reduced search space for the systems and achieving a significant improvement in the efficiency of the Transmission System Expansion Planning process.