Cryogenic Propellant Loading is a crucial part of the control system in fields such as space launches. Traditional scheduling methods mostly rely on static or open-loop optimization, which are unable to cope with dynamic factors such as changes in raw material viscosity, equipment aging, and demand fluctuations. These issues lead to large prediction errors, frequent failure of instructions, and long computation time for global re-optimization. Additionally, these systems lack online parameter identification and local rapid correction mechanisms, resulting in limited robustness and practicality in aerospace engineering scenarios. This paper is targeted at typical application requirements such as aerospace fuel injection and aviation aviation fuel supply. We proposed a scheduling method that combines rolling horizon optimization with adaptive control. We constructed an ARHO-MINLP model aiming to minimize the number of tank switching, inventory costs, and equipment idle penalties. At the same time, it strictly satisfies the core constraints of power and mechanical engineering, such as tank capacity constraints, energy consumption balance, and refueling operation sequence. Using a rolling window of length N, only the first M steps are executed and then the process is closed-looped. The parameters are identified online using filtering techniques and the model is dynamically updated. When the deviation exceeds the threshold, local re-optimization is triggered. This method integrates the rolling horizon optimization and adaptive mechanism, effectively overcoming model mismatch, avoiding global recalculation, and significantly enhancing the real-time performance, robustness and adaptability in high dynamic scenarios such as those in aerospace and aviation.