| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 15.38 MB | Adobe PDF |
Authors
Abstract(s)
Todo o setor energético está a passar por um período de mudanças drásticas em várias valências e isso exige
que, cada vez mais, a rede seja fiável, estável e eficiente.
Dado isto, o presente trabalho está intrinsecamente relacionado com as temáticas mencionados. Sendo o
foco principal o estudo da implementação de machine learning em subestações de forma a aumentar a sua
fiabilidade e, consequentemente, a sua estabilidade.
As subestações atualmente contam com um avançado sistema de monitorização baseado no SCADA
(Supervisory control and data acquisition), no entanto esta implementação apenas permite melhorias na
ótica de monitorização e controlo da rede elétrica. A integração de machine learning irá permitir que se façam
previsões de eventuais defeitos na rede.
Primeiramente será feita uma introdução às subestações, a sua importância e respetivos equipamentos que
a integram. Abordando concisamente os principais componentes de uma subestação e como os mesmos
podem ser integrados no sistema de previsão.
Segundamente será feita uma introdução ao machine learning, incluindo o seu funcionamento e as diferentes
tipologias existentes. Concluindo esta parte com uma breve comparação entre os diferentes tipos.
Após ser feita uma breve introdução das subestações e de machine learning separadamente, é feita uma
análise das vantagens que poderão surgir da integração.
Estando feita a introdução teórica, é elaborado um código em python que permite, com recurso a redes
neuronais, fazer previsões e extrair dado das mesmas. São criados diferentes cenários para perceber quais
os melhores parâmetros nesta situação de integração. Os dados desses cenários são apresentados e
analisados.
Com a análise feita, conclui-se com considerações importantes relativamente á utilidade do algoritmo e a
eficiência demostrada.
The entire energy sector is going through a period of drastic change in various areas, and this requires the grid to be increasingly reliable, stable and efficient. Given this, this work is intrinsically related to those themes. The focus is to study the implementation of machine learning in substations to increase their reliability and, consequently, their stability. Substations currently have an advanced monitoring system based on SCADA (Supervisory control and data acquisition), but this implementation only allows for improvements in terms of monitoring and controlling the electrical network. The integration of machine learning will allow predictions to be made of possible faults in the network. First, an introduction will be given to substations, their importance and the equipment they contain. The main components of a substation and how they can be integrated into the forecasting system will be concisely covered. Secondly, there will be an introduction to ML, including how it works and the different types, this part concludes with a brief comparison between the different types. After a brief introduction of substations and ML separately, an analysis is made of the advantages that could arise from integration. Once the theoretical introduction has been made, a python code is created which allows predictions to be made using neural networks and to extract data from them. Different scenarios are created to see which parameters are best in this integration situation. The data from these scenarios is presented and analysed. The analysis concludes with important considerations regarding the usefulness of the algorithm and its demonstrated efficiency.
The entire energy sector is going through a period of drastic change in various areas, and this requires the grid to be increasingly reliable, stable and efficient. Given this, this work is intrinsically related to those themes. The focus is to study the implementation of machine learning in substations to increase their reliability and, consequently, their stability. Substations currently have an advanced monitoring system based on SCADA (Supervisory control and data acquisition), but this implementation only allows for improvements in terms of monitoring and controlling the electrical network. The integration of machine learning will allow predictions to be made of possible faults in the network. First, an introduction will be given to substations, their importance and the equipment they contain. The main components of a substation and how they can be integrated into the forecasting system will be concisely covered. Secondly, there will be an introduction to ML, including how it works and the different types, this part concludes with a brief comparison between the different types. After a brief introduction of substations and ML separately, an analysis is made of the advantages that could arise from integration. Once the theoretical introduction has been made, a python code is created which allows predictions to be made using neural networks and to extract data from them. Different scenarios are created to see which parameters are best in this integration situation. The data from these scenarios is presented and analysed. The analysis concludes with important considerations regarding the usefulness of the algorithm and its demonstrated efficiency.
Description
Keywords
Electrical substations Machine learning Subestação elétrica