Indicator Diagram for Power Prediction of Pumping Unit Based on Machine Learning

This paper studies the problem of using machine learning to predict the indicator diagram of pumping unit according to the power of pumping unit in a period of time. In the past, it was troublesome to install a machine to collect indicator diagram data or indirectly collect the size of pumping units. This paper introduces several methods of indirectly predicting indicator diagram based on power. They are two nearest neighbor methods and three DNN methods. Finally, through experimental comparison, it is proved that the DNN network model with embedding layers used to distinguish different pumping unit specifications has better prediction index diagram ability, and its prediction index diagram is basically consistent with the original index diagram, which can replace the traditional method.