Interactions linear-Based Prediction of Kinetic Energy in Flywheel Energy Storage for Enhanced Grid Stabilization

Flywheel energy storage systems have emerged as a promising technology for grid stabilization due to their ability to rapidly store and release kinetic energy. Accurate prediction of the kinetic energy stored in the flywheel allows for better control of charging and discharging cycles, maximizing energy utilization, and minimizing losses. This prediction is influenced by multiple interdependent factors, including rotational speed, mass distribution, and external load conditions. This study evaluates three machine learning approaches, including ensemble bagged trees, exponential gaussian process regression, and an interactions linear regression model. In this paper, an interactions-based linear regression model is proposed to incorporate interaction terms, capturing not only the individual effects of these variables but also their combined influence on the kinetic energy. This approach enhances the predictive accuracy of kinetic energy estimation and enables more efficient grid stabilization strategies. The interactions linear regression model demonstrates superior performance by achieving the lowest root mean squared error compared to other methods.