The Energy Intensive Enterprises (EIEs) have considerable flexibility in energy consumption, which can be utilized to improve renewable integration. However, EIEs’ self-scheduling problem with the nodal voltage constraints and power flow equations is a mixed integer non-linear programming problem, which is hard to solve for existing solvers. Moreover, simplification of this problem such as linearization method will result in inaccuracy of the results so that the nodal voltage constraints will be violated. In order to solve the aforementioned problems, at first this paper applies second-order cone relaxation method to make the original problem convex, which reduces the difficulty to solve the problem. Then, a feasible solution recovery procedure, which is based on AC power flow, is introduced so as to recover an accurate and feasible solution from the relaxed second-order conic problem. Compared with the results from original problem, linearized problem, relaxed second order conic problem and the results from the proposed method, it is observed that the proposed method can eliminate the inaccuracy induced by second-order cone relaxation or linearization approximation. Meanwhile, the proposed method can obtain a feasible and optimized result in a short time compared to the solving time of the original problem by Knitro. Thus the difficulty of solving the self-scheduling problem is reduced. This work is supported by Science and Technology Project of China Southern Power Grid Co., Ltd (No. GDKJXM20162081).