Predictive ConvFormer-LSTM Digital Twin for Intelligent Cyber-Physical Operation of Grid-Connected Photovoltaic Systems
- 9 April 2026
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Cyber-Physical Systems
- Vol. 4, 238-245
Abstract
In this paper, an intelligent cyber physical data driven Digital Twin (DT) framework is developed for monitoring and performance enhancement of grid-connected photovoltaic (PV) systems. The developed DT framework integrates sensing, modeling, and learning within a unified cyber-physical environment for accurate system state representation and effective control. The proposed framework employs a hybrid ConvFormer-LSTM architecture, which integrates convolutional feature extraction, Transformer-based attention mechanism, and Long Short-Term Memory (LSTM) to jointly capture local transients and long-range temporal dependencies in multivariate time-series signals. The framework enables accurate estimation of voltage and current stresses, as well as non-invasive parameter identification of critical components, using predicted voltage and current signals. Validation on a 2-stage, 3-phase grid-connected photovoltaic (PV) energy conversion system demonstrates that the proposed method achieves a coefficient of determination (R²) of 0.9573 under hybrid dynamic conditions, with low MAPE values such as 1.78. These results confirm that the proposed DT methodology provides a robust and generalizable solution for monitoring and parameter estimation in nonlinear circuit and system environments.Keywords
Funding Information
- Anusandhan National Research Foundation
- Government of India
- Solar and Wind-Based Microgrid: Real-Time Health Monitoring, Control Optimisation, and Performance Assessment (ANRF/ARG/2025/000645/ENS)
This publication has 17 references indexed in Scilit:
- Transformers for Multivariate Time Series Forecasting: Comprehensive Analysis, Challenges, Research Opportunities, and Future ProspectsIEEE Access, 2026
- Framework of Digital Twin Based on State Space Model for Stability Assessment of 2-Stage 3-Phase Grid Connected PV SystemIEEE Transactions on Circuits and Systems I: Regular Papers, 2025
- Anomaly Detection Using Bi-Directional Long Short-Term Memory Networks for Cyber-Physical Electric Vehicle Charging StationsIEEE Transactions on Industrial Cyber-Physical Systems, 2024
- Survey on Foundation Models for Prognostics and Health Management in Industrial Cyber-Physical SystemsIEEE Transactions on Industrial Cyber-Physical Systems, 2024
- Coordinating Systematic Grid-Forming Control of Hybrid Photovoltaic Plants in Weak GridsIEEE Transactions on Industrial Cyber-Physical Systems, 2024
- Artificial Intelligence (AI) Based Black-Box Modeling Algorithm for System Identification in Three-Phase Single-Stage PV Inverter Systems2023
- A Proactive Intrusion Detection and Mitigation System for Grid-Connected Photovoltaic InvertersIEEE Transactions on Industrial Cyber-Physical Systems, 2023
- A Novel Neural-Network Device Modeling Based on Physics-Informed Machine LearningIEEE Transactions on Electron Devices, 2023
- Digital Twin Enabled Domain Adversarial Graph Networks for Bearing Fault DiagnosisIEEE Transactions on Industrial Cyber-Physical Systems, 2023
- 240°-Clamped PWM in Three Phase Grid-Connected PV Converter Application2021