A carregar...
Publicação
Data-driven approach to power line inspection prioritization based on criticality assessment
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.84 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
O fornecimento contí nuo de eletricidade e um pilar fundamental da
sociedade. Interrupço es, mesmo que breves, podem ter impactos significativos a
ní vel operacional, econo mico e social. Com o aumento da complexidade das redes
ele tricas, torna-se cada vez mais crucial garantir a resilie ncia e fiabilidade das
infraestruturas energe ticas, nomeadamente das linhas ae reas de alta e me dia
tensa o. Ém Portugal, a ÉDP Labelec tem um papel de destaque neste domí nio,
inspecionando anualmente mais de 23000 km de linhas de alta e me dia tensa o. No
entanto, o crescente volume de dados resultante das inspeço es levanta novos
desafios na identificaça o e priorizaça o das linhas ele tricas. Como os preditores
disponí veis neste estudo esta o agregados a ní vel municipal (e na o por linha
individual), a priorizaça o sera feita agrupando as linhas por municí pio, em vez de
analisa -las individualmente.
Éste Trabalho Final de Mestrado apresenta um modelo de priorizaça o
baseado em dados de inspeço es de linhas ae reas de alta e me dia tensa o,
desenvolvido no a mbito de um esta gio curricular no departamento de Inspeço es de
Ativos da ÉDP Labelec. Recorrendo a dados de anomalias, o estudo combina modelos
estatí sticos cla ssicos como Regressa o Linear, A rvore de Regressa o e Regressa o
Logí stica com te cnicas de aprendizagem automa tica, como Random Forest, Gradient
Boosting Machines e Redes Neuronais, para estimar um í ndice de criticidade por
municí pio. Éste í ndice reflete na o so a freque ncia e gravidade das anomalias,
ajustadas ao comprimento das linhas, como tambe m integra fatores contextuais, tais
como o consumo de eletricidade, densidade populacional, a rea abrangida pela Rede
Natura 2000 e exposiça o costeira.
O ranking de criticidade resultante, por municí pio, permite um planeamento
mais estrate gico das inspeço es, ajudando a ÉDP Labelec a concentrar os seus
recursos nas zonas de maior risco, melhorando a eficie ncia operacional e reduzindo
custos. Ao promover inspeço es mais eficazes e orientadas, esta abordagem contribui
para uma gesta o mais sustenta vel e robusta da infraestrutura energe tica.
The uninterrupted supply of electricity is essential to the functioning of modern society, with even brief outages carrying substantial operational, economic and social impacts. As electrical networks grow in complexity, maintaining the resilience and reliability of electrical infrastructure, particularly Overhead Power Lines (OPLs), has become increasingly important. In Portugal, ÉDP Labelec plays a leading role in this domain, inspecting over 23000 km of high and medium voltage lines annually. However, the increasing volume of inspection data poses new challenges in identifying and prioritizing the most critical power lines. Since available predictors in this study are aggregated at the municipal level (not per individual line), prioritization must be performed by grouping power lines into municipalities rather than analyzing them individually. This Master’s Final Work proposes a data-driven prioritization model for OPL inspections, developed during a curricular internship at ÉDP Labelec’s Asset Inspections department. Using anomaly data, the study combines classical statistical models such as Linear Regression, Regression Tree and Logistic Regression with advanced machine learning techniques, including Random Forest, Gradient Boosting Machines and Neural Networks, to estimate a criticality score for each municipality. This score reflects both the frequency and severity of anomalies, adjusted for line length, but also incorporates contextual factors such as electricity consumption, population density, Rede Natura 2000 area coverage and coastal exposure. The resulting municipality-level risk ranking supports strategic inspection planning, enabling ÉDP Labelec to target high-risk areas, improve operational efficiency and reduce costs. By supporting more efficient and targeted inspections, the proposed approach contributes to a more sustainable and robust energy infrastructure.
The uninterrupted supply of electricity is essential to the functioning of modern society, with even brief outages carrying substantial operational, economic and social impacts. As electrical networks grow in complexity, maintaining the resilience and reliability of electrical infrastructure, particularly Overhead Power Lines (OPLs), has become increasingly important. In Portugal, ÉDP Labelec plays a leading role in this domain, inspecting over 23000 km of high and medium voltage lines annually. However, the increasing volume of inspection data poses new challenges in identifying and prioritizing the most critical power lines. Since available predictors in this study are aggregated at the municipal level (not per individual line), prioritization must be performed by grouping power lines into municipalities rather than analyzing them individually. This Master’s Final Work proposes a data-driven prioritization model for OPL inspections, developed during a curricular internship at ÉDP Labelec’s Asset Inspections department. Using anomaly data, the study combines classical statistical models such as Linear Regression, Regression Tree and Logistic Regression with advanced machine learning techniques, including Random Forest, Gradient Boosting Machines and Neural Networks, to estimate a criticality score for each municipality. This score reflects both the frequency and severity of anomalies, adjusted for line length, but also incorporates contextual factors such as electricity consumption, population density, Rede Natura 2000 area coverage and coastal exposure. The resulting municipality-level risk ranking supports strategic inspection planning, enabling ÉDP Labelec to target high-risk areas, improve operational efficiency and reduce costs. By supporting more efficient and targeted inspections, the proposed approach contributes to a more sustainable and robust energy infrastructure.
Descrição
Palavras-chave
Planeamento de Inspeções Avaliação de Criticidade Aprendizagem Automática; Modelação Éstatística Inspection Planning Criticality Assessment Machine Learning Statistical Modelin
Contexto Educativo
Citação
Cruz, Catarina Regueira (2025). “Data-driven approach to power line inspection prioritization based on criticality assessment”. Dissertação de Mestrado. Universidade de Lisboa. Instituto Superior de Economia e Gestão
Editora
Instituto Superior de Economia e Gestão