Browsing by Author "Llivisaca Mejia, Mateo David"

Now showing 1 - 2 of 2

Aplicación de aprendizaje por refuerzo en la planificación de la expansión de los sistemas de transmisión
(Universidad de Cuenca, 2025-09-18) Llivisaca Mejia, Mateo David; Astudillo Salinas, Darwin Fabian; Torres Contreras, Santiago Patricio
Transmission Expansion Planning (TEP) is, from a mathematical point of view, a complex optimization problem and, in practice, a key tool for planning the efficient growth of the Electric Power System (EPS). With the constant increase in demand and generation capacity, if the transmission bottleneck is not solved, a proper operation of the system cannot be guaranteed. In this research, Reinforcement Learning (RL) is applied to solve the TEP through three approaches: the use of Q-learning, a learning-guided metaheuristic and deep reinforcement learning. In this framework, both Q-learning and RL-based metaheuristics learn online and use that knowledge locally. On the other hand, deep reinforcement learning methods, such as Deep Q-Network (DQN) and REINFORCE, allow reusable general policies to be learned, although they require a better representation of the state. This is achieved by modeling the power grid as a complex network and evaluating nodes using centrality-based metrics. The results show that the use of RL in the metaheuristic improves the exploration of the solution space. As for the deep learning approach, this first approach offers promising results: although in some cases optimal topologies are not reached, feasible plans are obtained that can serve as initial solutions in classical optimization schemes.
Planeamiento dinámico de la expansión de los sistemas de transmisión utilizando un algoritmo voraz iterativo
(Universidad de Cuenca, 2022-05-16) Llivisaca Mejia, Mateo David; Torres Contreras, Santiago Patricio
The Dynamic Transmission System Expansion Planning (DTSEP) problem provides a more efficient possibility for the investment in new infrastructure in the electrical system. Currently, the use of meta-heuristics leads the solution methods, however, more simple heuristics methods have not been totally explored. In this paper, a heuristic method is proposed to solve the PDEST based on an iterative greedy algorithm using the AC model. The use of high voltage direct current (HVDC) links within the formulation and scenarios without energy re-dispatch, are also considered. The statistical results of the simulations show that the proposed algorithm is robust and competitive against other state-of-art methods. The four analyzed systems are Garver 6-bus, IEEE 24-bus, Garver HVDC 6-bus, and IEEE HVDC 39-bus test systems.