Efficient evolutionary algorithms for robust Volt/VAr control in modern power distribution systems

• Ano de defesa: 2021
• Autor(a) principal: Vitor, Tiago Soares
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Conservation voltage reduction; Controle Volt/VAr; Distributed generation; Distribution systems; Evolutionary multi-objective optimization; Geração distribuída; Otimização evolutiva multiobjetivo; Projeto robusto; Redução de tensão conservativa; Robust design; Sistemas de distribuição; Volt/VAr control
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/18/18154/tde-24022021-165858/

Modern power distribution systems have posed new challenges to Volt/VAr control strategies due to the increasing penetration of distributed generation (DG) units. But, with the advent of smart grid technologies, these systems have also provided remarkable opportunities for the development of new optimization and energy saving techniques. In this context, this thesis presents a new specialized evolutionary system for Volt/VAr optimization (SES-VVO) to efficiently solve the multi-objective optimization problem (MOP) when the number of tap-changing operations of voltage regulating devices is considered together with the conservation voltage reduction (CVR) goals. This problem is a challenging task for evolutionary algorithms (EAs) in terms of meeting voltage limits constraints and reaching near optimal solutions. Specialized search mechanisms were developed to improve the overall performance of the day-ahead operation planning and the hourly decision-making approaches, giving more realistic and cost saving solutions for distribution systems\' operators. The SES-VVO also corroborated with the development of an advanced strategy that has potential for real-time operation considering the stochastic behavior of load/DG variables. Further results demonstrated that the tap change costs have a decisive impact on the economics of VVO-based CVR. Additionally, this thesis investigates and proposes a novel robust framework – constituted by parameters and metrics – to quantitatively estimate how robust is a VVO solution when facing the stochastic nature of load/DG uncertainties. The analysis demonstrated that the VVO solutions designed to specifically search for CVR goals, both in energy and cost savings, are sensitive to these uncertainties, leading to violation of voltage standards. So, as the integration of DG is evermore increasing in distribution systems, uncertainty realizations can lead to poor voltage quality in non-robust VVO designs. To tackle this problem, the proposed parameters/metrics were incorporated into the VVO-based CVR via MOP. Thus, a balance between performance and robustness could be achieved through trade-off solutions, followed by a decision-making based on the used-preferred level of robustness. The results showed that at the expense of a small amount of energy and operating cost, greater robustness can be obtained to face the uncertainties. In summary, the proposed methods in this thesis are promising for smart grid applications, being capable of finding robust and cost-effective CVR solutions on a real-time scale.