"Integrated Microgrid Dataset for AI-Enabled Resilient Operation Under Uncertainty"

"This dataset presents a detailed microgrid operational dataset designed to improve AI research in five key areas of power and energy systems: forecasting, fault detection, economic dispatch, microgrid control, and cyber-physical security. The dataset combines three aspects that are rarely included together in other resources: multi-horizon forecasting targets , a wide range of fault scenarios with duration patterns, and operational control context, including battery degradation tracking. Based on actual NASA POWER meteorological observations and confirmed through physics-based models, the dataset includes eight types of realistic noise modeling, such as cloud effects, temperature derating, panel soiling, and measurement uncertainty. This approach results in a solar-irradiance correlation of 0.93 instead of artificial perfection. Energy balance is maintained at every timestep, with a residual error below 0.01 kW. Baseline demonstrations confirm research usefulness: solar forecasting reaches R\u00b2 = 0.86, fault detection achieves a precision of 0.935, and battery optimization shows an 82.4% cost reduction. The dataset fills important gaps in current resources by illustrating the connections between forecasting uncertainty, equipment failures, and control decisions that influence real-world microgrid operations. All generation code, fault injection framework, and baseline models are available for complete replication, allowing researchers to recreate the dataset or develop geographic and seasonal variations. "