Skip to content

Home

Jump to bottom
MeepOwned13 edited this page Sep 26, 2023 · 9 revisions

Welcome to the Hungarian Elecricity Load Forecasting wiki!

Goal

Forecast the electricity load of Hungary for the next couple hours using Deep Learning techniques and time-series data. Datasets like these generally have autoregressive features and are usually influenced by external variables such as weather, holiday identifiers and time identifiers. The goal is to test multiple Neural Networks and time-series forecasting techniques.

Data

OMSZ (Országos Meteorológiai Szolgálat) shares current and past weather data in their database: link to OMSZ data used
More details on the OMSZ dataset

MAVIR (Magyar Villamosenergia-ipari Átviteli Rendszerirányító Zrt.) shares up-to-date electricity load data in their database: link to MAVIR data used

Used the above mentioned datasets to make a dataset which uses country-wide averages and has proper fields for time-series prediction: wiki page of dataset used for EDA

Steps

  1. Gather data from websites mentioned in the "Data" section
  2. Perform EDA (Exploratory Data Analysis)
  3. Choose at least 3 different Neural Network types and compare their performance
  4. Test out 2 approaches to forecasting
  5. Evaluate and illustrate performance via metrics and graphs
  6. Choose final model and approach that provide the best results
Clone this wiki locally