The nonlinear cumulants analysis is an important method in photon correlation spectroscopy (PCS) particle sizing technique. Using unconstrained nonlinear optimization algorithm, optimum fitting result is achieved in the nonlinear fitting process with random original value, and the implementation of the algorithm is presented. Autocorrelation functions of scattered light signal of normal polystyrene latex particles of 50 nm, 100 nm, 200 nm and 500 nm are fitted using this method, and particle sizes are inversed. Comparied with other curve fitting algorithm, this method has little dependence on original value, and the inversed particle sizes are more close to the nominal value with less measurement error.