Vibration energy harvesting technology is a hotspot research area in energy harvesting technology because it can convert the vibrational energy in the environment into electrical energy for output and thus provide the distributed energy for microelectromechanical systems. To improve the energy harvesting performance of the vibration energy harvesting system with partial information, we analyzed the probabilistic response of the stochastic system excited by Gaussian white noise under different geometric structures and effectively predicted the corresponding energy harvesting performance. Firstly, we established the coupling moment equation of the vibration energy harvesting system with the cumulant truncation method and then obtained some high-order moments. Then, the probability density function of the stationary response was set in exponential form with unknown parameters by using the maximum entropy principle, and those the unknown parameters will be obtained by solving the minimum value of an objective function, which contains the obtained moment information. Finally, the effects of the physical parameters (including geometric structure parameters and Gaussian white noise) on the dynamic behavior of the vibration energy harvesting system with only partial information have been studied and verified all results by direct numerical simulation.