Millimeter wave massive multiple input multiple output (mmWave mMIMO) with lens an-tenna array known as beamspace (BS) system can reduce dimension of mmWave mMIMO system and required number of radio frequency (RF) chains using different beam selec-tion algorithms without performance loss. However, the beam selection requires precise information of BS channel and number of supported users cannot be greater than num-ber of RF chains. In this work, we use pilot-based channel estimation technique with a compressive sensing (CS) tool to estimate the BS channel with few sample and low pilot overhead by applying orthogonal matching pursuit (OMP) algorithm. It is simple to im-plement and computationally efficient. And also to support greater number of users than number of RF chain in the estimated BS channel we integrate non-orthogonal multiple accesses (NOMA) with BS-MIMO system called BS-NOMA. Minimum mean square error (MMSE) and maximum ratio transmission (MRT) precoders are consider in addition to zero-forcing (ZF) to improve the performance of BS-NOMA with estimated BS channel at low power regime and also to mitigate the existing inter-beam interference. Further to maximize achievable sum rate of BS-NOMA with estimated BS channel, dynamic power allocation is apply which solves joint power optimization problem including intra and inter beam power. Simulation result shows that under the same BS channel estimation spectral efficiency of BS-NOMA able to achieve 7.0866 - 23.028 and 8.534 - 20.29 gain than BS-MIMO at UL-SNR = 0dB and = 20dB respectively. Similarly, energy efficiency of BS-NOMA able to achieve 0.8562 - 3.125 and 1.034 - 3.262 gain than BS-MIMO at UP-SNR = 0dB and = 20dB respectively. Key Words: BS-NOMA, Channel Estimation, CS, mmWave, OMP, Power Allocation