"HYD-P4DT Dataset: A Physics-Informed Dataset for Digital Twin Modeling of Hyderabad Households"

"A physics-informed digital twin is a virtual representation of a physical system in which data generation is constrained by underlying electrical laws rather than purely statistical patterns. In household energy systems, this allows realistic simulation of how appliances respond to variations in grid conditions such as voltage fluctuations. To address this, this work presents HYD-P4DT, a physics-informed digital twin dataset designed for modeling household electrical behavior under Hyderabad conditions. The framework is grounded in in-house smart meter measurements of a 9 W LED load, from which electrical parameters are derived and used to calibrate the simulation. Specifically, grid voltage is modeled as a stochastic process, while appliance usage is represented using a time-dependent Markov model. Furthermore, electrical behavior is computed using a ZIP (Constant Impedance, Current, Power) formulation, ensuring that power consumption varies consistently with voltage. The dataset is generated by aggregating device-level responses to obtain system-level quantities, including voltage, current, active power, power factor, frequency, and cumulative energy. A controlled fault injection mechanism is incorporated to simulate voltage sags and device-level anomalies, with exact labels available for each event. The resulting dataset provides physically consistent multi-variable time-series data under both nominal and disturbed conditions. In contrast, existing datasets primarily capture steady-state operation or provide limited measurements. HYD-P4DT enables analysis of voltage-dependent behavior and fault scenarios within a physically consistent framework, supporting the development and evaluation of models for smart household and grid applications."