Neural Network Identification of the Head of a Pumping Unit with an Asynchronous Electric Drive

The article describes the operation of the neural network identifier for the liquid head of centrifugal pumping units. The considered pumping unit consists of a centrifugal pump, suction and pressure pipelines with shut-off valves and an asynchronous frequency-controlled electric drive. Currently, such an electric drive of pumping units is the most common. However, to improve the operating modes of such units, it is often necessary to maintain the set-up parameters using feedback systems including pressure sensors. The work presents the experience of receiving a pressure signal using electrical current and voltage signals of an asynchronous motor. For this, a pre-trained neural network was used. The results of numerical experiments using the developed simulation model of a liquid pumping station are presented. As a result of the research, the error of the neural network identifier of the head in static and dynamic modes of operation was determined.