Uma nova abordagem para estimativa da resistência de aterramento de turbinas eólicas

• Ano de defesa: 2024
• Autor(a) principal: Leal, Alexandre Giacomelli
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Curitiba; Brasil; Programa de Pós-Graduação em Engenharia Elétrica e Informática Industrial; UTFPR
• Palavras-chave em Português: Turbinas eólicas; Aprendizado do computador; Energia eólica; Correntes elétricas - Aterramento; Simulação (Computadores); Algorítmos - Desenvolvimento; Wind turbines; Machine learning; Wind power; Electric currents - Grounding; Computer simulation; Algorithms - Development; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA; Engenharia Elétrica
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/34162

Monitoring the grounding resistance of wind turbine generators over the lifespan of a wind farm is essential for ensuring the safety of personnel and equipment at such facilities. However, conventional methods used for measuring grounding resistance pose challenges in onshore wind farms’ routine maintenance due to their complexity and associated high costs. In this study, a systematic review was conducted to assess the benefits and limitations of various ground resistance measurement methods and to identify gaps in current research. The investigation highlighted the practicality and safety advantages of the Clamp-on Ground Meter Method (CGM) over other approaches. Nevertheless, the direct application of CGM measurements to assess turbine ground resistance is hindered by significant methodological errors. To address this issue, a novel solution leveraging computational simulations and machine learning techniques was proposed to minimize errors associated with CGM measurements in wind farms. By training a machine learning model using simulated CGM data on the wind farm grounding system, including scenarios with unforeseen higher ground resistance values indicative of corrective maintenance needs, the approach accurately predicts turbine ground resistance values from clampon meter readings during routine inspections. Once trained and validated, the machine learning model can make predictions in future checks throughout the installation’s lifetime, assuming no changes to its grounding project. In a case study utilizing a grounding system from a Brazilian complex, a comparative analysis was conducted between errors in CGM direct application and those of the proposed solution. Additionally, the proposed solution was juxtaposed with the widespread high-frequency method, revealing a 58.2% reduction in estimation errors. This advancement not only enhances security but also simplifies organizational aspects. Moreover, the proposed method substantially reduced measurement execution time by 80% compared to conventional approaches, addressing a critical gap by accurately estimating higher ground resistance values crucial for diagnosing soil or electrode deterioration processes. Furthermore, an experimental validation on a multi-grounded system similar to a wind farm further confirms the credibility of the proposed methodology. By comparing estimates with those obtained through conventional methods, the study demonstrated the effectiveness and reliability of the approach, even in practical environments.