In the United States, the buildings sector accounted for about 41% of primary energy consumption in 2010, which was around 44% more than the transportation sector and 36% more than the industrial sector. Real time forecasts for building energy consumption using weather forecasts are crucial for effective building energy management. And the Variable Base Degree Day (VBDD) model has been proven effective in this field. In the VBDD model, two factors mainly determine the accuracy of the energy forecasts, where the first is the computation of optimal base values in the dynamic building energy consumption forecasts model and the second is the accuracy of weather forecasts. In this paper, we are motivated by the field study of forecasting the energy consumption of commercial buildings using local weather forecasts. We advance the state-of-The-Arts by correcting biases in the weather forecasts and interpolating the daily weather forecasts into higher frequency to synchronize with the frequency of the energy consumption forecasts. Based on this framework, we propose an optimal efficient algorithm to compute base values in the VBDD model through the coordinate gradient descent method. Experiment results on multiple real datasets demonstrate the effectiveness of the proposed method through achieving more than 90% accuracy.