Integrated machine learning forecasting and grey wolf optimization for optimal operation of virtual power plants in smart distribution networks.

The increasing penetration of distributed energy resources (DERs) necessitates intelligent coordination strategies that simultaneously enhance the technical, economic, and environmental performance of modern distribution networks. This paper proposes an integrated Virtual Power Plant (VPP) framework that combines optimal distributed generation (DG) planning, machine learning-based load forecasting, and battery energy storage system (BESS) scheduling within a unified energy management architecture. The proposed framework integrates the Grey Wolf Optimizer (GWO) for optimal DG placement and sizing with a hybrid Artificial Neural Network-Support Vector Machine (ANN-SVM) forecasting model to enable data-driven operational scheduling under varying load conditions. The methodology is validated using the IEEE 69-bus radial distribution system through comprehensive steady-state and time-series simulations. The optimization results demonstrate that the proposed planning strategy reduces active power losses by 61.7% while improving the minimum bus voltage from 0.909 p.u. to 0.966 p.u., thereby enhancing network voltage stability. The hybrid ANN-SVM forecasting model provides accurate hourly load predictions that support reliable energy scheduling and coordinated DER operation. Time-series simulations further verify that the coordinated dispatch of renewable energy resources and BESS maintains secure voltage profiles and stable system operation during peak-demand periods. Renewable generation supplies approximately 4.2 MWh/day, while the BESS performs effective energy arbitrage that increases renewable energy utilization and reduces dependence on grid electricity. Comprehensive techno-economic evaluation confirms the practicality and scalability of the proposed VPP framework, achieving daily operating cost savings of $1,741.32, a discounted payback period of 3.8 years, and an internal rate of return (IRR) of 30%. Furthermore, increased renewable energy penetration leads to significant reductions in CO₂ emissions, demonstrating the environmental benefits of the proposed approach. The obtained results consistently demonstrate that the coordinated integration of metaheuristic optimization, hybrid machine learning forecasting, and intelligent energy storage management provides a robust, reproducible, and scalable solution for improving the reliability, operational efficiency, economic viability, and sustainability of future smart distribution networks.