Cognitive radio is expected to be implemented in smart grids since it presents high reliability, high accuracy and low transmission time by utilizing licensed bands opportunistically. Shadowing environment affects the performance of channel availability detection of local spectrum sensing since it occurs occasionally. Therefore, the cooperative spectrum sensing is encouraged to be used for addressing shadowing issues. The principle cooperative spectrum sensing techniques suffer from unreliable local information from secondary users (SUs) who are encountered by the shadowing effect. Then, several alternative methods, adaptive majority rule and improved weight algorithm (IMA) is proposed by taking the SUs reliability into account. However, the unreliable SUs are still considered according to the algorithm. Therefore, in this paper, we propose an adaptive multi-criteria thresholding (AMT) to determine the channel availability according to the SUs reliability. The main contribution of AMT is three-fold. First, the new reliable weight calculation is proposed by utilizing analytic hierarchy process (AHP) under three major criteria. Second, AMT is flexible to the number of SUs since it adapts the decision weight on the optimal number of SUs according to the reliable SUs. Third, the shadowing issue is addressed by taking only reliable SUs into account.