Virtual power plants (VPPs), by aggregating and collaboratively controlling flexible demand-side resources, have become an important means to mitigate the supply-demand imbalance in modern power systems. Due to unpredictable factors, resources such as air conditioners and electric vehicles may exhibit power response deviations when executing day-ahead dispatch plans as a result of forecast errors. To address this, a real-time rolling optimization correction method for VPPs with multiple flexible resources is proposed. First, control models for adjustable resources—including electric vehicle clusters, air conditioning loads, and battery storage—are established. Combined with real-time updated weather conditions and electric vehicle connection information, this approach accurately quantifies the potential power deviation of the VPP during the execution of the day-ahead dispatch plan. Second, by introducing the price-quantity relationship curves of various flexible resources, the necessary price incentives for further tapping into their intra-day flexibility are analyzed. Finally, a real-time multi-resource coordinated optimization correction model is constructed to determine the optimal allocation strategy for power deviation correction in VPPs, thereby economically eliminating the power deviations caused by day-ahead forecast errors in real time. Case study results show that, compared with traditional correction strategies relying solely on battery storage as a backup resource, the proposed method is more economical and fully exploits the regulation potential of multiple flexible resources, enabling efficient and economic operation of VPPs.