| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 7.53 MB | Adobe PDF |
Autores
Resumo(s)
Os avanços nas redes elétricas inteligentes têm desempenhado um papel crucial na descarbonização,
facilitando a integração de fontes de energias renováveis e otimizando o equilíbrio entre
oferta e demanda de energia. Nesse contexto, um dos principais objetivos das Smart Grids é aprimorar
a interação entre os usuários finais e o sistema elétrico, por meio da implementação de
programas que incentivam a redução do consumo e a adoção de práticas mais sustentáveis. No
entanto, apesar das vantagens significativas resultantes da expansão da infraestrutura energética,
surgem novos desafios relacionados à segurança e à integridade dos dados coletados. Sem
mecanismos robustos de validação, o sistema fica vulnerável a manipulações, o que pode comprometer
a eficácia dos mecanismos de distribuição de benefícios. Para resolver esse problema,
a presente dissertação propõe uma abordagem fuzzy híbrida: um modelo evolutivo orientado a
dados (denominado de Evolving Takagi‐Sugeno Plus) e um sistema fuzzy Mamdani baseado em
conhecimento. O método evolutivo é utilizado para modelar e prever os padrões de comportamento
dos indivíduos em programas de participação ativa de usuários finais de energia. Este é
capaz de evoluir dinamicamente, adaptando seus parâmetros e ajustando sua estrutura automaticamente
a partir das amostras recebidas. Durante a etapa de concepção do modelo, o método
foi comparado com outras técnicas disponíveis na literatura, mostrando resultados competitivos,
especialmente em relação ao tempo de execução. Por outro lado, o sistema Mamdani utiliza o
resíduo obtido entre a saída do modelo evolutivo e os dados reais de flexibilidade, combinados
com informações sobre geração e consumo de energia, para estimar um grau de alerta caso
comportamentos anômalos sejam identificados. Os resultados desta fase indicam que o sistema
proposto detecta tanto fraudes pontuais quanto aquelas que ocorrem ao longo de períodos extensos.
Dessa forma, os métodos combinados demonstram potencial de aplicação em contextos
práticos, auxiliando as entidades gestoras na tomada de decisões por meio de uma metodologia
robusta e altamente interpretável.
Advances in smart grids have played a crucial role in decarbonization, facilitating the integration of renewable energy sources and optimizing the balance between energy supply and demand. In this context, one of the main objectives of smart grids is to improve the interaction between end users and the electricity system by implementing programs that encourage the reduction of consumption and the adoption of more sustainable practices. However, despite the significant advantages resulting from the expansion of the energy infrastructure, new challenges arise related to the security and integrity of the data collected. Without robust validation mechanisms, the system is vulnerable to manipulation, which can compromise the effectiveness of benefit distribution mechanisms. To solve this problem, this dissertation proposes a hybrid fuzzy approach: a data‐driven evolving model (called Evolving Takagi‐Sugeno Plus) and a knowledge‐based fuzzy Mamdani system. The evolving method is used to model and predict the behavior patterns of individuals in active participation programs for energy end users. It is capable of evolving dynamically, adapting its parameters and adjusting its structure automatically based on the samples received. During the model design stage, the method was compared with other techniques available in the literature, showing competitive results, especially in terms of execution time. On the other hand, the Mamdani system uses the residual obtained between the output of the evolving model and the real flexibility data, combined with information on energy generation and consumption, to estimate a degree of alert if anomalous behavior is identified. The results of this phase indicate that the proposed system detects both one‐off frauds and those that occur over long periods. In this way, the combined methods show potential for application in practical contexts, helping management entities to make decisions using a robust and highly interpretable methodology.
Advances in smart grids have played a crucial role in decarbonization, facilitating the integration of renewable energy sources and optimizing the balance between energy supply and demand. In this context, one of the main objectives of smart grids is to improve the interaction between end users and the electricity system by implementing programs that encourage the reduction of consumption and the adoption of more sustainable practices. However, despite the significant advantages resulting from the expansion of the energy infrastructure, new challenges arise related to the security and integrity of the data collected. Without robust validation mechanisms, the system is vulnerable to manipulation, which can compromise the effectiveness of benefit distribution mechanisms. To solve this problem, this dissertation proposes a hybrid fuzzy approach: a data‐driven evolving model (called Evolving Takagi‐Sugeno Plus) and a knowledge‐based fuzzy Mamdani system. The evolving method is used to model and predict the behavior patterns of individuals in active participation programs for energy end users. It is capable of evolving dynamically, adapting its parameters and adjusting its structure automatically based on the samples received. During the model design stage, the method was compared with other techniques available in the literature, showing competitive results, especially in terms of execution time. On the other hand, the Mamdani system uses the residual obtained between the output of the evolving model and the real flexibility data, combined with information on energy generation and consumption, to estimate a degree of alert if anomalous behavior is identified. The results of this phase indicate that the proposed system detects both one‐off frauds and those that occur over long periods. In this way, the combined methods show potential for application in practical contexts, helping management entities to make decisions using a robust and highly interpretable methodology.
Descrição
Palavras-chave
Abordagem Fuzzy Híbrida Detecção de Fraudes Modelos Evolutivos Programas de Participação Ativa Redes Elétricas Inteligentes Sistema Fuzzy Mamdani Hybrid Fuzzy Approach Fraud Detection Evolving Models Active Participation Programs Smart Grids Mamdani Fuzzy System
Contexto Educativo
Citação
Editora
Licença CC
Sem licença CC