A method was established to quantitatively describe nonlinear dynamic behavior of heterogeneous flow structure in gas-solid fluidization by carrying out multiple resolution and time series reconstruction of experimental signals. Optimum sampling frequency was proposed by analyzing the dependence of Kolmogorov entropy of time series on sampling frequencies. Original signals could be resolved with respect first to scale, and then to time and amplitude into a periodic function and three irregular changes (micro-scale fluctuations, and meso-scale irregular amplitude and time fluctuations). Statistical analysis indicated that the three kinds of irregular components are of random characteristics, with a complicated evolution of their probability density distributions with changing operating conditions. A new distribution function, called Compound Weibull distribution with seven parameters, was developed to describe the irregular fluctuations of flow structure, leading to a nonlinear dynamic model for simulating dynamic behaviors of heterogeneous flow structure in gas-solid fluidization, which has been coded into a software package.