Institutional Repository of Vilnius University
Fotovoltinių elektrinių duomenų analizė taikant mašinų mokymosi metodus /
Abstract
dc:descriptionApplications of Machine Learning Methods for Analysis of Big Data from Photovoltaic Power Stations. In this paper the author is aiming to find the optimal machine learning methods for analysis of Big Data from photovoltaic power stations for fault detection and localisation and prediction of the energy output. The author explains the main concepts behind photovoltaic power stations, presents the characteristics, value and correlations of their generated data, provides examples. Due to the inherent nature of the generated data, manual analysis is impractical, if not impossible, thus the application of machine learning methods is essential. The author overviews, briefly explains and compares three machine learning methods for anomaly detection (HTM Detector, Variational Autoencoders and SR-CNN models), and two methods, suitable for forecasting of electrical energy output (DeepAR and Phased LSTM models). During the comparisons, useful or unsuited features of each method are described. The Variational Autoencoder model is extended to analyse multivariate time series data. In their findings, the author selects the probable most-suited methods for each problem. Practical research determining the best methods for anomaly detection is done and empirical data proving the methods’ performance is provided.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Udris, Matas,
- Contributors dc:contributor
-
- Kybartas, Rimantas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD107116730&prefLang=en_US