Optimizing the oilfield working system has become one of the most effective ways to maximize the oilfield profits. In order to improve the efficiency of the production optimization and handle the nonlinear inequality constraints, this paper applied sequential quadratic programming algorithm to this study area. The sequential quadratic programming algorithm will transform the nonlinear optimization problem that seeks the optimal value of the variable into a series of quadratic programming sub-problems that locates the searching directions of the variables. The gradients of the augmented Lagrange function, used in the construction of the QP sub-problems with respect to the variables, are calculated by the adjoint method while the Hessen matrix is updated using BFGS method. By solving the QP sub-problems, the searching directions of the control variables are then obtained which can be used to adjust the initial vari?ables and finally to locate the optimal values of control variables. The new method has not only super linear convergence speed, but also combines the advantages of adjoint method which can efficiently calculate the gradients even in the case of a large number of parameters. In this paper, a two-dimensional three-phase reservoir model is tested to verify the validity of the method. The displacement results of optimized development plans under total injection and production rate equality constraints and inequality constraints show that,the field cumulative water injection amount of the former is less than that of the latter by 18 percent, and the field cumulative oil production amount of the former is higher than the latter by 14%. However, the pressure in the former has fallen dramatically which would easily lead to energy shortage in the late development stage. In order to get reasonable development performance, we should do production optimization under total injection and production equality constraints or add lower limit constraints to the inequality constraints.The theory is also applied to a real oil field and the total injection rate is set with equality constraints. The recovery has increased by 0.92% and the NPV has increased by 35% after the optimization.