The research of 3D underwater sensors' coverage-control has great significance in rational allocation of underwater space resources, environment-aware & information-acquisition and the network survivability improvement. In the real world situation, the uneven coverage is always needed in the underwater detection region. The current studies focus on the conception of "Event-Driven" without paying enough attention to the diversity of k-coverage requirements of different 3D regions. This work concentrates on the optimizational deployment for Underwater Sensor Networks' (UWSNs) diverse k-coverage requirements of distinguished regions. Based on the model of "buoys, anchors, underwater sensors which are connected and communicated via a wired cable with their respective buoy", we study the low-boundary number of the sensors required by diversity of k-coverage, and develop a novel algorithm: k-ERVFA (k-Equivalent radius virtual force algorithm) as an improvement to the classic virtual-force algorithm (VFA). Compared with the VFA in our simulation, the new algorithm only loses 1.21%~3.89% of 1-coverage rate in the required 1-coverage region, but significantly improves the 2- and 3- coverage rate to 17.42%~44.3% and 28.95%~49.53% respectively in the corresponding required coverage regions. The improvement in k-coverage requirements illustrates the correctness and effectiveness of k-ERVFA.