Solução do problema estimação de estado usando ferramentas de análise de dados

• Autor(a) principal: Souza, David Silva Marques de
• Data de Publicação: 2022
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UTFPR (da Universidade Tecnológica Federal do Paraná (RIUT))
• Texto Completo: http://repositorio.utfpr.edu.br/jspui/handle/1/32427
• topic: Aprendizado do computador; Otimização matemática; Inteligência computacional; Machine learning; Mathematical optimization; Computational intelligence; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA

The determination of the operating point in steady-state is an essential activity carried out to guarantee that the Electric Power System operates within the established operating limits. To determine the operating point of an Electric Power System, it is necessary to calculate the modules and angles of the nodal voltages. Among the methodologies used for this purpose is the state estimator. Although the state estimator has been widely used and published in the specialized literature, there are particular situations in which the state estimator cannot satisfactorily determine the operating point of a Power System. This work verifies the performance of computational intelligence techniques for this analysis with the aid of Python and AMPL programming languages. A database with multiple measurement scenarios of the adopted power system, the IEEE 14 Bus, was created to implement them. His data was used to train three machine learning methods: the Support Vector Machine, the Decision Tree, and the Multilayer Perceptron Neural Network. The models elaborated were analysedin terms of their generalization capacity and the accuracy of the results obtained with those obtained by the state estimation. The performance of computational intelligence methods wascompared from a statistical approach using statistical methods. The models obtained are effective in determining the operating point of the electrical power system, with results very similar to the conventional methodology.