The present disclosure relates to a performance forecasting method applied to an air conditioning equipment and a central server. The air conditioning equipment has a plurality of detecting elements so as to detect a plurality kinds of physical parameters and a performance parameter. The central server stores those parameters as a set of first parameters and a first performance value. When the central server receives other physical parameters and another performance parameter, it calculates the similarity between the other physical parameters and each set of the first parameters according to the law of cosines. The central server selects the set of the first parameters which have the similarity being greater than a selecting threshold, so as to calculating an expected performance value. Accordingly, the central server can forecast a performance decline degree by comparing the another performance parameter and the expected performance value.