Fuzzy particle swarm optimization (FPSO) integrating with a self-adaptive penalty function was proposed to solve constrained optimization problem. In order to improve the global optimal performance of FPSO, fuzzy logic controller is employed to adaptively control inertia weight, and random velocity is selected on condition that the algorithm falls into the local convergence. To obtain the global feasible optimal solution, the proposed self-adaptive penalty function adjusts penalty parameter according to the object function value and the ratio of feasible solution number to total solution number. It has great optimization ability for both inequality constraint functions and equality constraint functions. Simulation results show that FPSO effectively overcomes the problems of slow and premature convergence in unconstrained optimization problems, and has a better ability to find the global optimum solution than the standard PSO. PSO-based approach, combined with the presented new penalty function, is applied to solve the constrained problems successfully. Moreover, simulation results also demonstrate that FPSO-proposed penalty function has the best performance, compared with three other kinds of penalty functions.