Leveraging Graph Neural Networks to Forecast Electricity Consumption

Eloi Campagne,Yvenn Amara-Ouali,Yannig Goude,Argyris Kalogeratos

2024-08-30

Abstract:Accurate electricity demand forecasting is essential for several reasons, especially as the integration of renewable energy sources and the transition to a decentralized network paradigm introduce greater complexity and uncertainty. The proposed methodology leverages graph-based representations to effectively capture the spatial distribution and relational intricacies inherent in this decentralized network structure. This research work offers a novel approach that extends beyond the conventional Generalized Additive Model framework by considering models like Graph Convolutional Networks or Graph SAGE. These graph-based models enable the incorporation of various levels of interconnectedness and information sharing among nodes, where each node corresponds to the combined load (i.e. consumption) of a subset of consumers (e.g. the regions of a country). More specifically, we introduce a range of methods for inferring graphs tailored to consumption forecasting, along with a framework for evaluating the developed models in terms of both performance and explainability. We conduct experiments on electricity forecasting, in both a synthetic and a real framework considering the French mainland regions, and the performance and merits of our approach are discussed.

Machine Learning,Artificial Intelligence

What problem does this paper attempt to address?

The paper aims to address several key issues in power demand forecasting, particularly the increased complexity and uncertainty introduced during the transition to a decentralized grid structure. Specifically, the paper addresses these issues through the following points: 1. **Improving Forecast Accuracy**: Accurate power demand forecasting is crucial for multiple participants in the energy market. With the increasing share of renewable energy (such as solar and wind), which brings volatility and intermittency, short-term power demand forecasting becomes even more important. 2. **Addressing Uncertainty and Complexity**: Factors such as the COVID-19 pandemic and economic downturn have introduced non-stationarity in power consumption patterns, increasing the difficulty of forecasting. Additionally, the decentralized grid structure makes it challenging for traditional methods to adapt to new challenges. 3. **Utilizing Graph Neural Networks (GNNs)**: By using graph-based representations, the spatial distribution and relational complexity in decentralized grid structures can be effectively captured. The paper proposes a method that goes beyond the traditional generalized additive model framework, utilizing models like Graph Convolutional Networks or Graph SAGE to achieve this goal. 4. **Developing Customized Graph Inference Methods**: A series of graph inference methods for consumption forecasting are introduced, along with a framework for evaluating the performance and interpretability of these models. 5. **Experimental Validation**: Experiments were conducted on synthetic datasets and real datasets from local regions in France to validate the effectiveness and advantages of the proposed methods. In summary, the paper improves power load forecasting by introducing graph neural network technology, addressing the limitations of traditional methods when facing modern grid structures, and enhancing the accuracy and reliability of forecasts.

Leveraging Graph Neural Networks to Forecast Electricity Consumption

Short-term residential load forecasting using Graph Convolutional Recurrent Neural Networks

A Novel Multiscale Transformer Network Framework for Natural Gas Consumption Forecasting

Probabilistic Demand Forecasting with Graph Neural Networks

Explainable Approaches for Forecasting Building Electricity Consumption

Spatio-Temporal Graph Neural Networks for Multi-Site PV Power Forecasting

A novel composite electricity demand forecasting framework by data processing and optimized support vector machine

A Single Scalable LSTM Model for Short-Term Forecasting of Disaggregated Electricity Loads

Forecasting day-ahead electricity prices with spatial dependence

GEANN: Scalable Graph Augmentations for Multi-Horizon Time Series Forecasting

Forecasting of electricity prices with neural networks

Short-Term Electricity Price Forecasting based on Graph Convolution Network and Attention Mechanism

Automated Spatio-Temporal Weather Modeling for Load Forecasting

Graph Convolution Network for Carbon Emission Factor Forecast of Power Grid

Hybrid Forecasting Approach Based on Grnn Neural Network and Svr Machine for Electricity Demand Forecasting

Graph Convolutional Networks based short-term load forecasting: Leveraging spatial information for improved accuracy

Mining latent patterns with multi-scale decomposition for electricity demand and price forecasting using modified deep graph convolutional neural networks

Research on future trends of electricity consumption based on conditional generative adversarial network considering dual‐carbon target

Forecasting with an N-dimensional Langevin Equation and a Neural-Ordinary Differential Equation

Machine-Learning-Based Electric Power Forecasting

Community stochastic domestic electricity forecasting