A Novel Hybrid Method for Indirect Measurement Dynamometer Card Using Measured Motor Power in Sucker Rod Pumping System

The dynamometer card (DC) is crucial for the remote monitoring of the sucker rod pumping system (SRPS). However, the DC sensors have poor real-time performance and high maintenance costs leading to difficulties in obtaining datasets of the downhole working states from different oil wells. Moreover, the motor power curve is easy to obtain in real-time and can reflect the changes in pumping unit load. Thus, this paper proposed a novel hybrid model for indirect measurement DC based on measured motor power. Firstly, based on the mechanical model analysis of SRPS, the dataset is constructed by transforming the motor power and geometric parameters of the SRPS into the polished rod torque, the first-derivative of polished rod torque, the second-derivative polished rod torque, and the torque factor. These four parameters are merged as the inputs of the data-driven model. Subsequently, a data-driven model implements the particle swarm (PSO) algorithm to optimize the weights and thresholds of the Back Propagation (BP) neural network model, which was trained to predict DCs of three common working states. Finally, the proposed method is verified experimentally through the measured DC and motor power data. Both experimental and prediction results demonstrate the effectiveness of the proposed method for generating DC. The mean DC area relative error of the hybrid method proposed in this paper is 2.52%.