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

Area of Science:

• Electrical Engineering
• Computer Science
• Environmental Science

Background:

• Increasing distributed energy resources (DERs) require advanced coordination for grid performance.
• Current distribution networks face challenges in integrating DERs efficiently and maintaining stability.
• Intelligent energy management is crucial for enhancing technical, economic, and environmental aspects of modern grids.

Purpose of the Study:

• To propose an integrated Virtual Power Plant (VPP) framework for optimizing DERs, load forecasting, and energy storage.
• To enhance the technical, economic, and environmental performance of distribution networks.
• To provide a scalable and robust solution for future smart grids.

Main Methods:

• Integrated VPP framework combining optimal distributed generation (DG) planning, machine learning-based load forecasting (hybrid ANN-SVM), and battery energy storage system (BESS) scheduling.
• Grey Wolf Optimizer (GWO) for DG placement and sizing.
• Validated on the IEEE 69-bus radial distribution system using steady-state and time-series simulations.

Main Results:

• Reduced active power losses by 61.7% and improved minimum bus voltage to 0.966 p.u.
• Accurate hourly load predictions via hybrid ANN-SVM model supporting reliable scheduling.
• Coordinated dispatch of renewables and BESS maintained voltage stability during peak demand.
• Achieved daily cost savings of $1,741.32, a 3.8-year payback period, and 30% IRR.
• Significant CO₂ emission reductions due to increased renewable energy utilization.

Conclusions:

• The proposed VPP framework effectively integrates metaheuristic optimization, hybrid machine learning, and energy storage management.
• Demonstrated improvements in grid reliability, operational efficiency, economic viability, and sustainability.
• The framework is robust, reproducible, and scalable for future smart distribution networks.