To improving the accuracy of Brillouin Optical Time Domain Reflectometer (BOTDR) in long-distance temperature sensing, the detection optical power optimization and wavelet threshold denoising for long-distance BOTDR temperature sensing are proposed. By analyzing the SNRE at the end of the timing curve under the intrinsic Brillouin frequency shift, the detection optical power optimization of BOTDR is achieved, while combining wavelet threshold denoising algorithm to improve the temperature measurement accuracy. The experimental results demonstrate that the range of Brillouin frequency shift fluctuation is decreased from 22MHz to 10MHz in the temperature region by the wavelet threshold denoising. Moreover, the temperature fluctuation of end temperature region is between ±5.3℃. In repeated measurement, the Root Mean Square Error (RMSE) of temperature fluctuation is less than 1.68℃ and the RMSE of localization is less than 1.78m. The detection optimal power and Brillouin frequency shift accuracy at the end of the fiber were explored under different spatial resolutions. This research results indicate that the accuracy of long-range BOTDR systems can be improved by wavelet threshold noise reduction combined with detection optical power optimization.