Learning-based state estimation: Comparison performance study of DL with ML
Traditional grid monitoring methods are well-established in high voltage networks. However, replicating these techniques in low voltage networks, which represent the largest share, is both technically and economically too costly due to the number of devices. The growth of renewable electricity distr...
| Autor: | |
|---|---|
| Formato: | tesis de maestría |
| Fecha de publicación: | 2024 |
| País: | España |
| Recursos: | Universitat Politècnica de Catalunya (UPC) |
| Repositorio: | UPCommons. Portal del coneixement obert de la UPC |
| Idioma: | inglés |
| OAI Identifier: | oai:upcommons.upc.edu:2117/405351 |
| Acesso em linha: | https://hdl.handle.net/2117/405351 |
| Access Level: | acceso abierto |
| Palavra-chave: | Deep learning Aprenentatge profund Àrees temàtiques de la UPC::Economia i organització d'empreses |
| Resumo: | Traditional grid monitoring methods are well-established in high voltage networks. However, replicating these techniques in low voltage networks, which represent the largest share, is both technically and economically too costly due to the number of devices. The growth of renewable electricity distributed generation affects the stability of the grid due to its intermittency. Recent efforts have centred on to machine learning approaches, specifically artificial neural networks, as an alternative to traditional state estimation techniques. These learning-based models are praised for their ability to handle complexity and non-linearity in power grids, providing accurate estimations with fewer measurements. This thesis presents a comparative study between deep learning machine learning models in the state estimation task for distribution power grids. The objective of the study is to evaluate the performance of the models in predicting voltage states of low voltage networks with renewable energy generation. The study uses a couple of deep learning models, a feed-forward neural network and a multi-layer perceptron regression, and various machine learning models, including random forest, light GBM, and multi-output linear regression. The models are tested on two synthetic networks that represent rural and suburban distribution power grids, comparing their accuracy and efficiency. The primary research objective is to assess the theoretical superiority of deep learning models over machine learning models in terms of accuracy and computational efficiency for the state estimation task. It is intended that this study will provide valuable information on the performance of machine learning and deep learning models for power grid state estimation. |
|---|