We consider the problem of target tracking in wireless sensor networks where the nonlinear observed system is assumed to progress respecting to a probabilistic state space model. This proposition improves the use of the variational filtering (VF) by jointly estimating the target position and optimizing the power scheduling, where the sensor observations are corrupted by additive noises and attenuated by path-loss coefficient. In fact, the quantized variational filtering (QVF) has been shown to be adapted to the communication constraints of sensor networks. Its efficiency relies on the fact that the online update of the filtering distribution and its compression are executed simultaneously. We first optimize quantization for reconstructing a single sensors measurement, and developing the optimal number of quantization levels as well as the minimal power transmitted by sensors under distortion constraint. Then we estimate the path-loss coefficient by maximizing the a posteriori distribution and the target position by using the QVF. The simulation results prove that the adaptive power optimization algorithm, outperforms both the QVF algorithm using uniform power level and the VF algorithm based on binary sensors.