Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Any-Quantile Probabilistic Forecasting of Short-Term Electricity Demand

Slawek Smyl, Boris N. Oreshkin, Paweł Pełka, Grzegorz Dudek

TL;DR

It is shown that the general approach proposed can be seamlessly applied to two distinct neural architectures leading to the state-of-the-art distributional forecasting results in the context of short-term electricity demand forecasting task.

Abstract

Power systems operate under uncertainty originating from multiple factors that are impossible to account for deterministically. Distributional forecasting is used to control and mitigate risks associated with this uncertainty. Recent progress in deep learning has helped to significantly improve the accuracy of point forecasts, while accurate distributional forecasting still presents a significant challenge. In this paper, we propose a novel general approach for distributional forecasting capable of predicting arbitrary quantiles. We show that our general approach can be seamlessly applied to two distinct neural architectures leading to the state-of-the-art distributional forecasting results in the context of short-term electricity demand forecasting task. We empirically validate our method on 35 hourly electricity demand time-series for European countries. Our code is available here: https://github.com/boreshkinai/any-quantile.

Any-Quantile Probabilistic Forecasting of Short-Term Electricity Demand

TL;DR

It is shown that the general approach proposed can be seamlessly applied to two distinct neural architectures leading to the state-of-the-art distributional forecasting results in the context of short-term electricity demand forecasting task.

Abstract

Power systems operate under uncertainty originating from multiple factors that are impossible to account for deterministically. Distributional forecasting is used to control and mitigate risks associated with this uncertainty. Recent progress in deep learning has helped to significantly improve the accuracy of point forecasts, while accurate distributional forecasting still presents a significant challenge. In this paper, we propose a novel general approach for distributional forecasting capable of predicting arbitrary quantiles. We show that our general approach can be seamlessly applied to two distinct neural architectures leading to the state-of-the-art distributional forecasting results in the context of short-term electricity demand forecasting task. We empirically validate our method on 35 hourly electricity demand time-series for European countries. Our code is available here: https://github.com/boreshkinai/any-quantile.
Paper Structure (20 sections, 16 equations, 11 figures, 5 tables)

This paper contains 20 sections, 16 equations, 11 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Motivation and Contributions
  4. Any-Quantile Forecasting Framework
  5. Short-term Distributional Forecasting Problem
  6. Any-Quantile Learning
  7. Any-Quantile Forecasting Models
  8. AQ-ESRNN
  9. AQ-NBEATS
  10. Empirical Results
  11. Datasets
  12. Training and Evaluation Setup
  13. AQ-ESRNN Training and Hyperparameter Settings
  14. AQ-NBEATS Training and Hyperparameter Settings
  15. Baseline Models
  16. ...and 5 more sections

Figures (11)

  • Figure 1: Any-quantile forecaster training methodology. The network accepts time-series history and a quantile level. The quantile level is generated randomly for each training sample, according to a pre-specified distribution (e.g. uniform or Beta) and is used both as neural network input and as the supervision signal in the quantile loss.
  • Figure 2: AQ-ESRNN architecture.
  • Figure 3: Beta(0.3, 0.3) distribution used for the training of AQ-ESRNN.
  • Figure 4: Variants of implementing the any-quantile conditioning in N-BEATS
  • Figure 5: Average electricity demand and daily dispersion demand in 35 European countries.
  • ...and 6 more figures