In this paper we present an initial investigation of the evolving cloud-based method using partial density calculation for fault detection of HVAC system (Heating, Ventilation and Air Conditioning). Moreover, we investigate different approaches of choosing the most influential components when calculating the partial local density which is further used for evolving the model structure. The method is based on the simplified fuzzy rule- based-system AnYa where the antecedent part is presented by data clouds. The effectiveness of the proposed method is tested on a model of HVAC system and furthermore, different types of faults are investigated. The results were compared with the well established fault detection method DPCA (Dynamic Principal Component Analysis).