The maintenance measure of fully restoring components to their initial operational state results in the inability to provide accurate maintenance plans. In real-world systems, there are more maintenance details that require a more complex description of the component repair process. A condition-based maintenance method was proposed for key components of subsea Christmas trees, utilizing a multi-level maintenance measures based on Wiener processes. A continuous degradation model of components was established based on historical operating data. A maintenance classification method was proposed by combining historical maintenance data with the K-nearest neighbors classification method. The system's maintenance plan over its operational cycle was obtained through multi-objective optimization algorithms. The results demonstrate that considering a multi-level maintenance measures can provide more accurate maintenance plans for multi-component systems, effectively balancing the trade-off between system reliability, availability, and maintenance costs.