Identification of centrifugal pump flow based on feature recognition

Abstract

The flow rate of water pumps is a crucial parameter for diagnosis, optimisation and renovation of transmission and distribution systems. To solve the problems of limited measurement conditions, inadequate consideration of operational wear and tear, and large deviations in the fitted flow curves when obtaining flow rates by direct measurement or characteristic curve fitting, this paper proposes a method for identification of centrifugal pump flow rates based on feature recognition. The method combines the unknown structural parameters of a centrifugal pump with three simple and easy-to-measure operating data: pump head, frequency and power, to obtain the pump characteristics and construct a model for identification of centrifugal pump flow. At the same time, the influence of measurement errors and wear and tear on the model’s accuracy is investigated. The results show that the maximum relative error of the pump operating flow obtained by the method is 2.27%, a slight deviation from the actual measured flow. The maximum relative error of flow rate is 2.7% when a maximum measurement error of ± 0.1%, 0.5% and ± (3%+5) for the required operational data, respectively, indicating that the method has a high resistance to interference. The method considers the different degrees of operating wear and tear of the pumps, saves on equipment input and enables fast and accurate identification of the flow rate of the pumps under different operating conditions.