Design of an intelligent AI-based multi-layer optimization framework for grid-tied solar PV-fuel cell hybrid energy systems
The method proposed in this work focused on a multi-level AI framework capable of enhancing the performance of a grid-tied solar PV-fuel cell hybrid energy system. RL-ENN stands for Reinforcement Learning-Driven Evolutionary Neural Network, while T-STFREP is a generalized acronym for Transformer-Bas...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-12-01
|
| Series: | MethodsX |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S221501612500384X |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The method proposed in this work focused on a multi-level AI framework capable of enhancing the performance of a grid-tied solar PV-fuel cell hybrid energy system. RL-ENN stands for Reinforcement Learning-Driven Evolutionary Neural Network, while T-STFREP is a generalized acronym for Transformer-Based Spatiotemporal Forecasting. FL-DEO stands for Federated Learning-Based Distributed Optimization. GNNHSCO is Graph Neural Networks Power Router. Q-GAN-ESO is Quantum-Inspired Generative Adversarial Networks. RL-ENN helps minimize Net Present Cost (NPC) and Cost of Energy (COE) in real-time with adaptive energy dispatch strategies; T-STFREP verifies accuracy based on forecasting transformer temporal sequence; FL-DEO allows for optimization in decentralized manner with privacy; and finally, GNNHSCO considers hybrid networking as energy graphs to diminish transmission losses whereas finally Q-GAN-ESO looks at synthetic degradation scenarios to allow for optimal management of energy storage. The application under consideration was simulated in MATLAB/Simulink and Python with TensorFlow, with a 30-year evaluation of meteorological data. The proposed model has decreased NPC by 27.5 %, COE by 18.2 %, and battery life by 30.2 %. The results validate its capability when compared against traditional methods such as Genetic Algorithms and Particle Swarm Optimization. With this, we now have a scalable and real-time energy-efficient solution for future smart grid systems. • Integrated Intelligence Stack: Combines RL-ENN, T-STFREP, FL-DEO, GNNHSCO, and Q-GAN-ESO into a unified architecture for real-time control, forecasting, decentralized optimization, network routing, and synthetic scenario generation. • Real-Time, Scalable, and Privacy-Preserving: Enables adaptive energy dispatch, federated optimization without compromising data privacy, and graph-based power routing, making it suitable for large-scale, smart grid deployments. • Proven Long-Term Performance: Achieved significant improvements over traditional methods (GA, PSO) with 27.5 % lower NPC, 18.2 % reduction in COE, and 30.2 % increase in battery life, validated using 30 years of meteorological data. |
|---|---|
| ISSN: | 2215-0161 |