To combat the problem of premature convergence observed in many applications of PSO, a novel self-adaptive particle swarm optimization algorithm-SAPSO is proposed in this paper. There exist two states for each particle in the SAPSO algorithm and a metric to measure a particlepsilas activity is defined which is used to choose which state it would reside. In order to balance a particlepsilas exploration and exploitation capability for different evolving phase, a self-adjusted inertia weight which varies dynamically with each particlepsilas evolution degree and the current swarm evolution degree is introduced into SAPSO algorithm. Simulation and comparisons based on several well-studied non-noisy problems and noisy problems demonstrate the effectiveness, efficiency and robustness of the proposed algorithm.