Energy Allocation Optimization of The Virtual Power Plant using Predictive Control with Artificial Neural Networks

Virtual power plants (VPPs) are crucial in modern power systems, allowing for the integration of distributed energy resources like solar and wind into the grid. They enhance grid flexibility, support demand-side management, and increase resilience during outages. Energy allocation in VPPs involves the efficient distribution and management of energy resources to meet demand while optimizing various objectives such as cost minimization, grid stability, and sustainability. This study investigates the operation of a VPP using predictive control with artificial neural networks (ANNs) models to minimize operational costs and optimize energy allocation. The process commences by constructing a predictive model to forecast solar energy production, wind energy generation, and electricity demand for the next 48-time intervals, each lasting 30 seconds. The significant input variables correlated with solar energy generation, wind energy generation, and electricity demand, are achieved through coefficient testing. Performance evaluation in the forecasting of ANNs models employs various error indices. The ANNs models with the lowest error indices are chosen for energy allocation in the VPP. Three energy allocation strategies in the VPP are evaluated, which include buying and selling electricity, not buying electricity, and not selling electricity. Simulation results indicate that predictive control maintains stable energy allocation compared to uncontrolled systems. Consequently, predictive control improves performance across all seasons and allocation strategies, making it an effective approach for energy allocation in virtual power plants.