NBEATS*: Enhanced N-BEATS for Mid-Term Electricity Demand Forecasting

We presents an enhanced N-BEATS model, N-BEATS*, for improved mid-term electricity load forecasting (MTLF). Building on the strengths of the original N-BEATS architecture, which excels in handling complex time series data without requiring preprocessing or domain-specific knowledge, N-BEATS* introduces two key modifications. A novel loss function combining pinball loss based on MAPE with normalized MSE, the new loss function allows for a more balanced approach by capturing both L1 and L2 loss terms. A modified block architecture– the internal structure of the N-BEATS blocks is adjusted by introducing a destandardization component to harmonize the processing of different time series, leading to more efficient and less complex forecasting tasks. Evaluated on real-world monthly electric ity consumption data from 35 European countries, N-BEATS* demonstrates superior performance compared to its predecessor and other established fore casting methods, including statistical, machine learning, and hybrid models. N-BEATS* achieves the lowest MAPE and RMSE, while also exhibiting the lowest dispersion in forecast errors.

This repository provides an implementation of the NBEATS* algorithm introduced in [https://arxiv.org/pdf/2412.02722].

Electricity Price Forecasting Results

Model	MedAPE	MAPE	IQrAPE	RMSE	MPE
ARIMA	3.32	5.65	5.24	463	-2.35
ETS	3.50	5.05	4.80	374	-1.04
k-NNw+ETS	2.71	4.47	3.52	327	-1.25
FNM+ETS	2.64	4.40	3.46	321	-1.26
N-WE+ETS	2.68	4.37	3.36	320	-1.26
GRNN+ETS	2.64	4.38	3.51	324	-1.26
MLP	2.97	5.27	3.84	378	-1.37
ANFIS	3.56	6.18	4.87	488	-2.51
LSTM	3.73	6.11	4.50	431	-3.12
ETS+RD-LSTM	2.74	4.48	3.55	347	-1.11
N-BEATS	2.55	3.78	3.30	309	0.34
N-BEATS*	2.20	3.44	3.29	304	0.56

Citation

If you use this code in any context, please cite the following paper:

@misc{kasprzyk2024enhancednbeatsmidtermelectricity,
      title={Enhanced N-BEATS for Mid-Term Electricity Demand Forecasting}, 
      author={Mateusz Kasprzyk and Paweł Pełka and Boris N. Oreshkin and Grzegorz Dudek},
      year={2024},
      eprint={2412.02722},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2412.02722}, 
}