The power consumption of the base stations is the major part of the total consumption of mobile communications, which implies that it is of great significance to reduce the radio transmit power consumption of base stations under the constraint of quality of service. The dynamic of the traffic leads to the adaptive adjustment of the antenna parameters of the base stations or even turning off some base stations serving low user traffic. We propose a penalty method to convert the power consumption optimization problem which constrained by the coverage condition into a simple form with only lower and upper bound condition. We also transform the discrete-valued coverage index into a continuous one and utilize the sub-gradient to conduct the gradient descent algorithm. Moreover, an elite scheme is adopted to preserve the best solution of the optimization problem. The proposed method is applicable to various optimization conditions, either to adjust the transmit power or to adjust the transmit power and the antenna tilt jointly. Experiment results show that our proposed algorithm performs well in BS ON/OFF switching networks. Besides, the effect of jointly adjusting the antenna tilt and the transmit power is better than that of adjusting the transmit power alone.