A three-dimensional flow field intelligent optimization method based on a surrogate model is proposed to address the development chal-lenges faced by complex offshore fault block reservoirs, such as low water flooding control, rapid formation pressure drop, and develop-ment of local flow-through channels. Through the strategy of synchro-nously constructing constraint proxy models such as comprehensive water cut and formation pressure in the main proxy model construc-tion process, multi constraint collaborative control optimization has been achieved. By coupling the proxy model construction and optimi-zation process, the number of iterations in reservoir numerical simula-tion has been greatly reduced. Furthermore, combined with parallel solving strategy, significant improvement in computational efficiency has been achieved. The application example shows that the new meth-od can achieve second level calculation of intelligent flow field opti-mization schemes for million grid reservoir numerical simulation models under various constraint strategies such as comprehensive wa-ter cut and formation pressure. Through multi well joint debugging of oil wells and injection wells, the degree of water drive control is in-creased by more than 10%, the average formation pressure coefficient is restored to 0.9, the comprehensive water cut is reduced by 4 per-centage points, and the block recovery rate is increased by more than 2%. The new method has achieved three-dimensional intelligent con-trol of reservoir flow field, solving the problems of multi-objective collaborative optimization such as improving water drive control de-gree, restoring formation pressure, and suppressing local flow chan-nels. The computational efficiency can meet the needs of oilfield site, and it has shown good application prospects in complex offshore fault block reservoirs.