On-Meter Graph Machine Learning: A Case Study of PV Power Forecasting for Grid Edge Intelligence

Jian Huang; Zixiang Ming; Yongli Zhu; Linna Xu

cs.LGApr 10, 2026

On-Meter Graph Machine Learning: A Case Study of PV Power Forecasting for Grid Edge Intelligence

Authors: Jian Huang, Zixiang Ming, Yongli Zhu, Linna Xu

Organizations: Sun Yat-sen University, Guangzhou, China

Abstract

This paper presents a detailed study of how graph neural networks can be used on edge intelligent meters in a microgrid to forecast photovoltaic power generation. The problem background and the adopted technologies are introduced, including ONNX and ONNX Runtime. The hardware and software specifications of the smart meter are also briefly described. Then, the paper focuses on the training and deployment of two graph machine learning models, GCN and GraphSAGE, with particular emphasis on developing and deploying a customized ONNX operator for GCN. Finally, a case study is conducted using real datasets from a village microgrid. The performance of the two models is compared on both the PC and the smart meter, exhibiting successful deployments and executions on the smart meter.

Figures & tables

Fig. 1: Illustration of ONNX and ONNX Runtime

Fig. 5: The real smart meter used in this paper

Fig. 6: Flowchart regarding how to register the customized operator in ONNX

Fig. 7: The architecture of the adopted GCN model

Fig. 8: The ONNX’s computation graph for the exported GCN model, grouped into functional blocks: Data Input, 2-layer GCN (with MyGcnOp), Flattening/Reshaping, and the MLP prediction head.

Fig. 9: The process for loading models with customized GCN operators in ONNX Runtime

Fig. 10: The process for deploying the GraphSAGE model

PV ID	Data Length	On-PC Time (sec)	MAPE	On-Meter Time (sec)	MAPE
1	2209	0.627	9.01%	1.199	9.01%
2	2209	0.627	8.42%	1.199	8.42%
3	2209	0.627	8.99%	1.199	8.99%

TABLE I: Results Comparison for the GCN Model

PV ID	Data Length	On-PC Time (sec)	MAPE	On-Meter Time (sec)	MAPE
1	2209	0.118	10.29%	2.970	10.29%
2	2209	0.118	9.29%	2.970	9.29%
3	2209	0.118	10.33%	2.970	10.33%

TABLE II: Results Comparison for the GraphSAGE Model

Fig. 11: Comparison of on-meter forecasting results: GCN vs. GraphSAGE

Explore similar work

GraphToolbox: A Configurable Python Framework for Graph Neural Network Forecasting
Sep 21, 2026Eloi Campagne, Yvenn Amara-Ouali, Yannig Goude +1Time Series Forecasting Electricity Load Forecasting
CB · LMO, CELESTE · EDF R\&D +1
Scalable Heterogeneous Graph Foundation Models for Data-Driven Optimal Power Flow in Smart Grids
May 22, 2026Massimiliano Lupo Pasini, Yijiang Li, Kibaek Kim +1Neural Surrogate Modeling Heterogeneous GNNs
Computational Sciences and Engineering Division, Oak Ridge National Laboratory, 1 Bethel Valley Road, Oak Ridge, TN 37831, USA
When Does Advection-Aware Graph Nowcasting Help? A Controlled Study of Distributed Solar Ramp Forecasting with a Self-Supervised Cloud-Motion Estimator
Sep 8, 2026Phillip JiangSolar Power Forecasting Spatiotemporal GNNs
Appsofa LLC

Sep 21, 2026J/K move · Enter open · S save

GraphToolbox: A Configurable Python Framework for Graph Neural Network Forecasting

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

CB · LMO, CELESTE · EDF R\&D+1

Time Series Forecasting Electricity Load Forecasting Spatiotemporal GNNs

Open paper AskPDF arXiv

Electricity forecasting often involves spatially related signals observed over regions, substations, and feeders, and Graph Neural Networks (GNNs) provide a natural way to represent these relations. Building a complete GNN forecasting experiment is nonetheless laborious, because graph construction, model selection, training, aggregation, and interpretation sit in incompatible tools. We present GraphToolbox, an open-source Python framework that unifies these stages in one configurationdriven pipeline built on PyTorch Geometric. It offers data-driven graph construction, an adapter that instantiates and trains 51 of the 65 PyTorch Geometric convolutions together with the recurrent cells of PyTorch Geometric Temporal, online expert aggregation, forecasting interpretability, and significance testing on cached forecasts. We evaluate the pipeline in two case studies. On French regional load, the 48 convolutions included in the complete forecasting sweep fall in a band from 1.14% to 1.60% error, online aggregation lowers this to 0.98%, and the graph models improve on classical additive and boosting baselines. On net-load, direct graph models are less accurate than a classical additive model, while forecasting each physical component separately improves them without closing that gap. Both comparisons use the same experimental interface, illustrating the role of GraphToolbox in systematic architectural evaluation.

Abstract

Explore similar work

GraphToolbox: A Configurable Python Framework for Graph Neural Network Forecasting

Scalable Heterogeneous Graph Foundation Models for Data-Driven Optimal Power Flow in Smart Grids

When Does Advection-Aware Graph Nowcasting Help? A Controlled Study of Distributed Solar Ramp Forecasting with a Self-Supervised Cloud-Motion Estimator