This study aims at assessing the performance of compressed sensing method for faster phosphorus magnetic resonance spectroscopic imaging ((31)P-MRSI) of human brain. A simulated 2D (31)P-MRSI dataset containing a tumor region and a healthy region was created based on the metabolite peak intensities and ratios of a volunteer dataset acquired at 3T. k-space data was randomly undersampled, and reconstructed using compressed sensing algorithm. This simulation study showed that compressed sensing reconstruction could be applied for faster (31)P-MRSI. Future studies will measure the performance of compressed sensing reconstruction for (31)P-MRSI in volunteers and patients with brain tumors.