To calculate the behavior of wind turbines in real-time and highly accurate, we developed a method to assimilate an approximation model with measurement data without the finite element analysis of mechanical systems. There is an advantage that the particle filter is easy to implement as data assimilation which identifies uncertain parameters of simulation with measurement data. On the other hand, it is necessary to prepare many particles, and accordingly the calculation cost of the data assimilation is high when using the finite element analysis. The input and output of the analysis were therefore approximated by using neural networks to realize real-time computation. The proposed method was applied to determine an uncertainty parameter included in the input of simulation for obtaining the behavior of wind turbines, and consequently, the behavior was accurately simulated with the resulting parameter.