The rapidly increasing residential air conditionings (RACs) have been widely considered as good regulation resources for improving the power system flexibility. However, due to the heterogeneity of different consumers and the difficulty of data acquisition in reality, the evaluation of large-scale RACs' regulation capacity is a tricky problem. To address this issue, this paper proposes an evaluation method based on Gaussian Mixture Model (GMM). First, the regulation framework of largescale RACs is developed based on partial observable data. Then, a calculation model of RACs' regulation capacity is proposed under the premise of guaranteeing heterogeneous consumers' comfort requirements. On this basis, we develop the GMM of RACs to evaluate their regulation capacities in the condition of insufficient data acquisition. The expectation maximization algorithm and Bayesian information criterion are utilized to optimize the multidimensional parameters of the GMM. Finally, we verify the proposed methods based on the realistic data in a demonstration project in China. The results show that the proposed method can evaluate the RACs' regulation capacity with more than 98% accuracy by measuring 1% RACs' parameters.