AN-BEATSfor Short-Term Electricity Load Forecasting with Adjusted Seasonality Blocks and Optimization of Block Order

EasyChair Preprint 8358

Abstract

For the proper operation of electrical systems, accurate electricity load forecasting is essential. This study focuses on solving the problem that is the optimization of the block order to archive better accuracy of the forecasting model. Furthermore, the seasonality blocks of N-BEATS are adjusted in theory by correctly using the Discrete Fourier Transform. Therefore, AN-BEATS-Adjusted Neural Basic Expand - Analysis Time series model is proposed to forecast short-term power loads based on electricity load history. Experiments show that the proposed model works better than the LSTM model and the order of blocks strongly affects the model's prediction results.

Keyphrases: Electricity load, Forecasting, N-BEATS, seasonal decomposition, time series

@booklet{EasyChair:8358,
  author    = {Anh Tuan Nguyen and Thanh Chau Do Thi and Anh Ngoc Le and Ngoc Anh Nguyen Thi},
  title     = {AN-BEATSfor Short-Term Electricity Load Forecasting with Adjusted Seasonality Blocks and Optimization of Block Order},
  howpublished = {EasyChair Preprint 8358},
  year      = {EasyChair, 2022}}