To address uncertainty challenges in highly renewable microgrids with large-scale electric vehicle (EV) integration, in this study, we developed a multi-source data-driven cooperative scheduling framework for wind–photovoltaic–storage–EV systems. The model integrates heterogeneous real-time monitoring data (grid status, renewable generation, and EV charging) through a Kalman-filter-based fusion architecture with dynamic anomaly detection. Renewable output and load uncertainties are predicted using a long short-term memory (LSTM)-DLinear hybrid model combining DLinear’s decomposition efficiency with LSTM’s residual correction. The scheduling optimization employs the Red-billed Blue Magpie Optimization Algorithm to solve a four-dimensional economic objective (TEB = C_procure + RBE + B_carbon + R_lifecycle) using gene-encoded 96-period decision variables. Key innovations include a price-responsive EV charging mechanism with adjustable power boundaries (P^min_ev(t)=α(t)⋅Prated) and temporal energy constraints. Implemented through a closed-loop predict-then-optimize framework with day-ahead planning and real-time correction layers, the solution demonstrates an 18.9% reduction in renewable curtailment and a 42% economic benefit improvement in high-EV penetration scenarios. This research validates the critical role of multi-source sensor data fusion in enhancing grid flexibility and schedulability, providing an effective real-time optimization approach for highly renewable microgrids.