Abstract：

 The establishment of pump model is the basis for studying the energy saving optimization control of the central air conditioning water system. At present, most of the water system optimization control models are established by curve fitting. This method has problems such as limited accuracy, difficulty in online adjustment, and poor versatility. Presents a fuzzy modeling method and combines with online subtractive clustering to realize online update and correction of the fuzzy model. The method classifies the running data of the pump according to different operating conditions, and gradually expands and improves the fuzzy rules in the model using these data samples. Fits the pump model by the actual operating data of the pump under multiple conditions, and compares and verifies the proposed algorithm. The results show that the algorithm has strong self-learning and self-adjusting ability, with good versatility and adaptability.