Solar energy production forecasting using deep neural networks

OData support
Supervisor:
Nagy Gábor
Department of Telecommunications and Media Informatics

This document describes my work aiming to find solutions that can predict energy values generated by solar plants. The goal is to make these renewable sources more predictable and reliable, therefore cheaper. I’m going to use machine learning algorithms, mainly neural networks, but I’m also going to implement mainstream model to benchmark the results. I briefly describe the basics of data science and neural networks. Than I write about the data provider services, and the supplied file formats, specially GRIB files. I collected data from 78 locations on hourly resolution. I merged and cleaned the datasets, calculated and visualized basic statistics from them. I also enriched the dataset with values extracted by shifting the date on time axis. In the document I describe the process of implementation of the downloader and data restructurer modules in detail. I also mention best practices and data transformation techniques to get the best out of the machine learning models. Than I describe the implemented neural networks and the way I get the best hyper-parameters. At the and I mention improvement possibilities.

Downloads

Please sign in to download the files of this thesis.