A Hybrid Swarm–Immune Biomimetic AI Framework for Multi-Scale Electric Vehicle Diffusion Modeling: Evidence from Maryland Vehicle Sales and Registration Data

Electric vehicle (EV) diffusion is often examined through incentives, infrastructure, and consumer adoption variables. Still, the computable representation of peer-field diffusion, abnormal perturbation, and self-organization in public registration data remains underdeveloped. This study proposes SI-BADF, a Swarm–Immune Biomimetic AI Diffusion Framework for multi-scale EV/PHEV registration modeling. The framework translates swarm intelligence into peer diffusion pressure and artificial immune systems into anomaly recognition, immune memory, and sample-weighted learning. The empirical setting combines Maryland monthly vehicle sales, county vehicle registrations, county EV/PHEV registrations, and ZIP-level EV/PHEV registrations. To reduce inflated fit due to stock autocorrelation, the main prediction targets are the one-, three-, and six-month cumulative EV/PHEV increments. Validation uses rolling-origin evaluation, baseline comparison, ablation tests, bootstrap confidence intervals, and high-influence-county exclusion. County-level EV/PHEV registrations increased from 62,638 in January 2023 to 151,231 in April 2026, while the longer background series increased from 26,157 in July 2020 to 151,231 in April 2026. Results show that SI-BADF improves mechanism transparency and anomaly-screening capacity, while point-forecast superiority remains conditional on target, horizon, and validation design. The contribution is an auditable biomimetic translation pathway for public technology-diffusion systems.