Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automated Deep Learning for Load Forecasting

Julie Keisler, Sandra Claudel, Gilles Cabriel, Margaux Brégère

TL;DR

The paper addresses the challenge of accurately forecasting electricity load in grids with high renewable penetration by introducing EnergyDragon, an Automated Deep Learning framework that extends the DRAGON NAS/HPO approach for load forecasting. EnergyDragon embeds a feature-selection mechanism within training, uses a two-DAG meta-architecture to map 2D inputs to 1D outputs, and optimizes architectures and hyperparameters via an asynchronous steady-state evolutionary search. In experiments on the French national load, EnergyDragon outperforms a industry GAM baseline and several AutoDL approaches, with the best variant (ED SSEA CNN/MLP) delivering substantial forecast accuracy gains. The work demonstrates that automated, interpretable-ready DNNs can yield robust, high-performance load forecasts, potentially improving resource scheduling and renewable integration, while highlighting the need for re-calibration and interpretability for industrial adoption.

Abstract

Accurate forecasting of electricity consumption is essential to ensure the performance and stability of the grid, especially as the use of renewable energy increases. Forecasting electricity is challenging because it depends on many external factors, such as weather and calendar variables. While regression-based models are currently effective, the emergence of new explanatory variables and the need to refine the temporality of the signals to be forecasted is encouraging the exploration of novel methodologies, in particular deep learning models. However, Deep Neural Networks (DNNs) struggle with this task due to the lack of data points and the different types of explanatory variables (e.g. integer, float, or categorical). In this paper, we explain why and how we used Automated Deep Learning (AutoDL) to find performing DNNs for load forecasting. We ended up creating an AutoDL framework called EnergyDragon by extending the DRAGON package and applying it to load forecasting. EnergyDragon automatically selects the features embedded in the DNN training in an innovative way and optimizes the architecture and the hyperparameters of the networks. We demonstrate on the French load signal that EnergyDragon can find original DNNs that outperform state-of-the-art load forecasting methods as well as other AutoDL approaches.

Automated Deep Learning for Load Forecasting

TL;DR

The paper addresses the challenge of accurately forecasting electricity load in grids with high renewable penetration by introducing EnergyDragon, an Automated Deep Learning framework that extends the DRAGON NAS/HPO approach for load forecasting. EnergyDragon embeds a feature-selection mechanism within training, uses a two-DAG meta-architecture to map 2D inputs to 1D outputs, and optimizes architectures and hyperparameters via an asynchronous steady-state evolutionary search. In experiments on the French national load, EnergyDragon outperforms a industry GAM baseline and several AutoDL approaches, with the best variant (ED SSEA CNN/MLP) delivering substantial forecast accuracy gains. The work demonstrates that automated, interpretable-ready DNNs can yield robust, high-performance load forecasts, potentially improving resource scheduling and renewable integration, while highlighting the need for re-calibration and interpretability for industrial adoption.

Abstract

Accurate forecasting of electricity consumption is essential to ensure the performance and stability of the grid, especially as the use of renewable energy increases. Forecasting electricity is challenging because it depends on many external factors, such as weather and calendar variables. While regression-based models are currently effective, the emergence of new explanatory variables and the need to refine the temporality of the signals to be forecasted is encouraging the exploration of novel methodologies, in particular deep learning models. However, Deep Neural Networks (DNNs) struggle with this task due to the lack of data points and the different types of explanatory variables (e.g. integer, float, or categorical). In this paper, we explain why and how we used Automated Deep Learning (AutoDL) to find performing DNNs for load forecasting. We ended up creating an AutoDL framework called EnergyDragon by extending the DRAGON package and applying it to load forecasting. EnergyDragon automatically selects the features embedded in the DNN training in an innovative way and optimizes the architecture and the hyperparameters of the networks. We demonstrate on the French load signal that EnergyDragon can find original DNNs that outperform state-of-the-art load forecasting methods as well as other AutoDL approaches.
Paper Structure (18 sections, 8 equations, 3 figures, 1 table)

This paper contains 18 sections, 8 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Deep Learning and AutoDL for load forecasting
  3. EnergyDragon
  4. Search Space
  5. Objective function
  6. Meta-Architecture
  7. Search Algorithm
  8. Experiments
  9. Dataset
  10. Baseline
  11. Generalized Additive Model
  12. Deep Neural Networks
  13. AutoPytorch
  14. DARTS
  15. EnergyDragon
  16. ...and 3 more sections

Figures (3)

  • Figure 1: DNN encoding as a directed acyclic graph (DAG), as proposed by keisler23.
  • Figure 2: Daily meta-model for load datasets.
  • Figure 3: Load power forecasting for the last week of November 2019. The ground truth is displayed in dotted line, the GAM forecast is drawn with a blue line whereas the forecast from the best version of EnergyDragon (ED SSEA CNN/MLP) is drawn in yellow.