Power Demand Prediction in Smart Microgrids Using Interacting Multiple Model Kalman Filtering

Optimized management of energy resources within smart microgrids may require an approximation of near future power demands to institute efficient scheduling of tasks. Since demands are volatile in shorter time spans, localized short-term prediction of demand is non-trivial. Local prediction requires efficiency of calculations to minimize computational resources.To address this concern we present a predictor based on the Interacting Multiple Model Kalman filters. This approach supports effective year round prediction while only storing two model demand profiles. We demonstrate that the approach provides consistency in intra-day demand prediction across an entire year.