This paper investigates beamforming and power allocation problems in the dynamic time division duplex (TDD) cellular networks. Based on the dynamic TDD coexistence scheme, the network comprises uplink and downlink networks serving machine-type devices (MTDs) and human-type devices (HTDs), respectively. The design goal is to minimize the total system power consumption by optimizing the transmit/receive beamforming and uplink power under MTD and HTD quality-of-service (QoS) constraints. The resulting optimization problem is challenging to solve because the variables to be designed are tightly coupled in the constraints. By using the uplink-downlink duality (UDD) and alternating optimization (AO) algorithm, we propose a novel algorithm to overcome the difficulty in this work. Numerical results demonstrate the superiority of the proposed algorithm.