An algorithm called the stock predictor algorithm is designed in this paper. It can be used in Distributed Control Systems which have these characteristics: (1) it contains lots of sub-systems. (2)Every sub-system could be controlled using a predicted value. (3) Within a certain sub-system, it will not do much harm to the whole system even if the predicted value differs greatly from the actual value. (4) The sub-system could be non-linear, time-dependent, but it will not change quickly.

The stock predictor algorithm described in this paper used a method similar to the way used by the stock-holders towards the stock predictors in the stock market. It applied several different control algorithms (viewed as stock predictors) to every controlled object (sub-systems) in every circle. Within each sub-system, according to the historical precision of every stock predictor, we use a weighted average value as the preferred predict value of that sub system. Then, when the cycle comes to an end, the stock predictors update their algorithms according to the difference between the predicted value and the actual value.

Based on this algorithm, we applied the stock predictor algorithm on a heating system, and got an optimized control strategy which is proved to be both efficient and economical.