Adaptive Load Modeling for Online Parameterization of Digital Twins in Power Grids

The integration of renewable energy sources and the electrification of the mobility and heating sectors present significant challenges for system operators, particularly at the distribution grid level. Ensuring grid observability is essential to manage the resulting increase in system dynamics and complexity, enabling robust stability and efficient resource utilization. Accurate monitoring and modeling of grid nodes are critical due to their stochastic, time-varying behavior and substantial impact on grid dynamics. The paper proposes an adaptive modeling approach for digital twins, incorporating distinct node models with real-time parameter estimation of electrical loads based on a dynamic exponential recovery load model. A constrained joint Square-Root Unscented Kalman filter is implemented for real-time estimation of load model parameters, supported by a dedicated validation structure. Integrated into a digital twin of a medium-voltage grid, different test cases evaluate the performance of Kalman filter algorithms in terms of estimation accuracy and replication of grid events such as voltage fluctuations and parameter changes. The results demonstrate that Kalman filters effectively enable real-time adaptation of load model parameters in power systems, although their sensitivity to dynamic disturbances must be considered. The presented digital twin, combined with a Hardware-in-the-Loop real-time simulation environment, provides a foundational tool for enhanced grid state estimation and snapshot-based analysis.