A hybrid graph-convolutional moving average framework for spatio-temporal voltage prediction in smart microgrids

Accurate voltage forecasting is critical for maintaining the operational stability and reliability of smart microgrids under dynamic load and generation conditions. This paper proposes a novel Hybrid Graph Convolution Moving Average (Hyb-GCMA) spatio-temporal deep learning framework that synergistically integrates Graph Neural Networks (GNNs), Long Short-Term Memory (LSTM) networks, and Seasonal AutoRegressive Integrated Moving Average with eXogenous variables (SARIMAX) for dynamic voltage prediction in an IEEE 14-bus truncated microgrid. The proposed framework very well encompasses spatial correlations between interconnected buses by graph-based learning while sequential learning of temporal dependencies due to time-varying load patterns. To increase the robustness and generality of the model, historical load-flow data under various operating scenarios are preprocessed by normalizing and generating sliding-window sequences. To prevent overfitting, adaptive learning rate strategies and early stopping mechanisms are used to further optimize the model training. The Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) per-bus accuracy metrics, along with \(R^2\) , show that the proposed Hyb-GCMA framework makes a significant improvement over conventional standalone deep learning baselines. The RMSE and MAE obtained are 0.0049 p.u and 0.0047 p.u, in squared and absolute error respectively. \(R^2\) value calculated is 0.999993, which suggests the model explains the proportion of variance best fit. The results show that spatio-temporal deep learning methods are providing great potential for real-time voltage stability assessment, anomaly detection and predictive control of next generation smart microgrid systems.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-17
DOI
https://doi.org/10.1038/s41598-026-80145-5
Primary Topic
Optimal Power Flow Distribution
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A hybrid graph-convolutional moving average framework for spatio-temporal voltage prediction in smart microgrids

Sultan M. Alghamdi, Muhammad Anas, Muhammad Adnan, Muhammad Tariq
5 links
Scientific Reports
Optimal Power Flow Distribution
article

A hybrid graph-convolutional moving average framework for spatio-temporal voltage prediction in smart microgrids

Sultan M. Alghamdi, Muhammad Anas, Muhammad Adnan, Muhammad Tariq
article en

Abstract

Accurate voltage forecasting is critical for maintaining the operational stability and reliability of smart microgrids under dynamic load and generation conditions. This paper proposes a novel Hybrid Graph Convolution Moving Average (Hyb-GCMA) spatio-temporal deep learning framework that synergistically integrates Graph Neural Networks (GNNs), Long Short-Term Memory (LSTM) networks, and Seasonal AutoRegressive Integrated Moving Average with eXogenous variables (SARIMAX) for dynamic voltage prediction in an IEEE 14-bus truncated microgrid. The proposed framework very well encompasses spatial correlations between interconnected buses by graph-based learning while sequential learning of temporal dependencies due to time-varying load patterns. To increase the robustness and generality of the model, historical load-flow data under various operating scenarios are preprocessed by normalizing and generating sliding-window sequences. To prevent overfitting, adaptive learning rate strategies and early stopping mechanisms are used to further optimize the model training. The Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) per-bus accuracy metrics, along with \(R^2\) , show that the proposed Hyb-GCMA framework makes a significant improvement over conventional standalone deep learning baselines. The RMSE and MAE obtained are 0.0049 p.u and 0.0047 p.u, in squared and absolute error respectively. \(R^2\) value calculated is 0.999993, which suggests the model explains the proportion of variance best fit. The results show that spatio-temporal deep learning methods are providing great potential for real-time voltage stability assessment, anomaly detection and predictive control of next generation smart microgrid systems.

Scientific Reports
Balochistan University of Information Technology, Engineering and Management Sciences (PK), King Abdulaziz University (SA), National University of Computer and Emerging Sciences (PK), University of Balochistan (PK), Applied Technologies (United States) (US)
Affordable and clean energy, Industry, innovation and infrastructure
Optimal Power Flow Distribution
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon