Detalhes bibliográficos
Ano de defesa: |
2021 |
Autor(a) principal: |
Massignan, Julio Augusto Druzina |
Orientador(a): |
Não Informado pela instituição |
Banca de defesa: |
Não Informado pela instituição |
Tipo de documento: |
Tese
|
Tipo de acesso: |
Acesso aberto |
Idioma: |
eng |
Instituição de defesa: |
Biblioteca Digitais de Teses e Dissertações da USP
|
Programa de Pós-Graduação: |
Não Informado pela instituição
|
Departamento: |
Não Informado pela instituição
|
País: |
Não Informado pela instituição
|
Palavras-chave em Português: |
|
Link de acesso: |
https://www.teses.usp.br/teses/disponiveis/18/18154/tde-25012023-094926/
|
Resumo: |
Massive and heterogeneous data sources are becoming incrementally available at power distribution networks, due to enhancements on traditional SCADA monitoring, including advanced metering infrastructures, and installing new sensors such as phasor measurement units. Data management then becomes a crucial process for operation and control of such power networks, for processing such diverse information, performing network assessment and optimizing decisions. Within such a perspective, distributions operators rely on state estimation applications, bridging the information from measured data with detailed physics-based models of power grids. This thesis extends the concepts of distribution system state estimation under a Bayesian Inference perspective, exploring a probabilistic interpretation for the state variables and associated randomness instead of only seeking to calculate a fixed state vector. This work employs this conceptual framework under three distinct and novel applications in the context of electrical distribution networks: dealing with non-Gaussian noise models under a Correntropy Extended Kalman Filter in power system state estimation, both from measurement noise and state behaviour; the proposition of a Bayesian information fusion to merge data gathered from pseudo-measurements, smart meters, SCADA and phasor measurements in distribution networks; the exploration of scalability of the three-phase unbalanced state estimation under a multiarea procedure based on Bayesian spatial fusion. Besides, a high-resolution and detailed model for distribution network is presented in the form of a generic component model based on a two-port admittance matrix formulation, improving the resolution of digital twin models for three-phase, unbalanced and asymmetrical distribution networks from the high voltage substations and primary feeders to low voltage secondary circuits. The proposed thesis also employs an orthogonal formulation and sparsity treatments to overcome numerical conditioning issues, a well-known challenge for classical state estimation formulations, while enhancing computational efficiency to ensure real-time performance. The developed algorithms and frameworks are evaluated on the IEEE test feeders and by the application of the proposed methods on real Brazilian test systems (both at distribution and transmission levels). The results corroborate the crucial task of including temporal characteristics on the state estimation while dealing with more generic noise characteristics under a kernel density concept, while properly tunning the kernel\'s bandwidths under different system\'s transitions. The use of information fusion shows itself as an essential practical resource to deal with different sampling and updating rates of the diverse set of measurements employed in distribution networks, especially when abrupt transitions are present while improving computational performance. Besides, the multiarea decomposition methods, along with sparse orthogonal formulations, are prominent in ensuring scalability and numerical stability of the estimation as a whole, a crucial practical contribution for large-scale distribution networks assessment. |