Integrated Distribution-Free Detection and Adaptive Robust Scheduling for V2G Smart Home Microgrids

Residential smart homes equipped with photovoltaic (PV) systems, battery storage, and vehicle-to-grid (V2G) enabled electric vehicles (EVs) depend on a home energy management system (HEMS) to schedule power flows. False data injection (FDI) attacks that subtly corrupt PV or load measurements can increase electricity costs and battery degradation. These attacks often evade conventional alarms. Existing approaches often separate anomaly detection from scheduling or require labeled attacks and Gaussian error models. This can result in missed attacks, costly false alarms, or overly conservative operation. This work introduces a unified framework, the Solar-Focused Theoretical Hybrid Wasserstein-Adaptive (SF-THWA) model, which integrates distribution-free attack detection with adaptive robust scheduling. Both forecast errors and FDI appear as distribution shifts in net demand. SF-THWA (i) derives time-varying robustness parameters for Wasserstein distributionally robust optimization (DRO) scheduling and (ii) detects attacks using a majority-vote ensemble. Experiments using real residential data show that integrating detection and scheduling reduces conditional value-at-risk (CVaR) by 20% compared to scheduling alone. In contrast, bolt-on detection with standard DRO increases costs by 27%, underscoring the need for integration. The hybrid ensemble achieves F1 = 0.934 at 2.5σ with zero false alarms. The breaking-point margin at 2.25σ establishes a quantifiable security envelope. All methods solve within 60 seconds on consumer-grade hardware.