Key Performance Index Analysis is a good complementing to the evaluation of past system security states, and the study on it has realistic meanings accordingly. It is realized that Cloud Theory and the Weighted Mean Method for Multi-Sensor could qualitatively describe Key Performance Index States. The data gathered from the multi-sensor under normal states will be tested for consistency by using distributing graph method firstly and will come out the fusion of the minimum average mean variance value by using the Weighted Mean Method for Multi-Sensor secondly. Then the above data fusion will be compared with the key performance index data at every arbitrary moment and each amount of deviation by forgoing compare will be gained. At last the five remark sets constructed from Cloud Theory combining with the amount of deviation will realize the qualitative description of Key Performance Index States. The experiment result shows the validity of this method.