Sterna migration optimization-based uncertainty-aware energy scheduling in hybrid solar and wind power systems using graph diffusion learning

Operational uncertainty in hybrid networks that integrate solar Photovoltaic (PV) and wind power is considerable and primarily caused by the unpredictable changes in the generation of renewables and the demand for power, which has a negative impact on reliability, power quality, and cost-effectiveness alike. In this respect, the research work under discussion puts forward a comprehensive uncertainty reduction framework that integrates optimal energy scheduling with predictive modelling in order to efficiently control the nonlinear interactions of PV generation, wind power, load demand, and Battery Energy Storage Systems (BESS). The proposed framework, called StMA-DGDNN, integrates the Sterna Migration Algorithm and a Decoupled Graph Diffusion Neural Network. In this framework, StMA minimizes Total System Cost (TSC), Annual Energy Losses (AEL), Energy Not Supplied (ENS), and Voltage Deviation (VD), thus optimizing the hybrid PV-WT-BESS dispatch. On the other hand, DGDNN captures spatial-temporal dependencies through graph diffusion and attention mechanisms, and performs uncertainty-aware nodal prediction of PV power, wind power, and load demand. The suggested system has been put into practice and examined in the MATLAB/Simulink setting, and its efficacy was measured against the performance of several current optimization methods like Deep Convolutional Generative Adversarial Networks with Particle Swarm Optimization (DCGAN-PSO), Artificial Ecosystem-based Optimization with Opposition-Based Learning (AEO-OBL), Particle Swarm Optimization with Sine Cosine Optimization (PSO-SCO), Artificial Hummingbird Algorithm (AHA), and Cuckoo Search Algorithm (CSA). The comparative results demonstrate that the proposed StMA-DGDNN achieves the lowest TSC of $1,402,300, reduces AEL to 894.24 kWh, limits ENS to 4.21 MWh, and maintains VD at 0.2724 p.u., outperforming all benchmark methods. Statistical and ablation analyses further confirm the robustness of the approach and highlight that the synergistic integration of optimization and prediction is the key contributor to the observed performance improvements. Overall, the proposed framework provides an effective and reliable solution for uncertainty-aware operation of hybrid PV-WT power networks.