Proposta de um método para bloqueio do esquema regional de alívio de carga frente a inércia rotativa de desligamento dos motores de indução

• Autor(a) principal: Andrade, Kaynan Maresch de
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Manancial - Repositório Digital da UFSM
• Texto Completo: http://repositorio.ufsm.br/handle/1/23738
• topic: ERAC; Motores de indução; Atuação indevida; Aprendizado de máquina; Classificação de padrões; Under frequency load shedding; Induction motors; Undue operation; Machine learning; Patterns classification; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA

The Under Frequency Load Shedding (UFLS) is an important systemic protection method used to prevent the electrical system from collapsing when the available generation is less than the load demanded. Traditionally, UFLS methods measure frequency and voltage via underfrequency and undervoltage relays. In the violation of operating limits, previously established loads are shed until the frequency returns to normal operating values. However, the low rotating inertia of the electrical system requires a fast load shedding in order to meet stability requirements. In some electrical systems it has been identified in the case of opening a transmission line upstream of a substation, the UFLS scheme can operate improperly, due to the influence of the inertia of the induction motors present in the distribution feeders. This work proposes the use of negative sequence voltage, frequency and its rate of change as variables applied in machine learning methods to, thus, allow faster UFLS (high reliability) without unwanted operation (high security). Logistic regression, quadratic discriminant analysis, linear discriminant analysis and linear discriminant analysis as dimensionality reducer are used, all being evaluated through simulations performed in the modified 9-bus test system. The results show that the logistic regression and linear discriminant analysis methods were able to block the improper performance of the UFLS protection with a high precision, while the quadratic discriminant analysis and the linear discriminant analysis as dimensionality reducer did not show satisfactory performance.