The prediction of the efficiency of oil well pumping systems plays an important role in optimizing the energy efficiency parameters of these systems. Currently, the prediction of oil well pumping system efficiency relies primarily on mechanistic models, but these models are often overly complex in predicting efficiency. Some researchers have attempted to use deep learning to predict system efficiency, but due to insufficient consideration of influencing factors and the causal relationships between these factors and system efficiency, they often include irrelevant variables as influencing factors, leading to less accurate prediction models. In this paper, a hybrid model (MDS–SSA–GNN) is proposed for the prediction of pumping well system efficiency. The model consists of six parts: Pearson's product moment correlation coefficient (PPMCC), multidimensional scaling (MDS) transform, maximum–minimum normalization, sparrow optimization algorithm (SSA), graph neural network (GNN), and maximum–minimum inverse normalization. First, the size of the correlation coefficient between each influencing factor and the system efficiency is quantitatively calculated by using PPMCC. Second, the main influencing factors are downscaled by using MDS and normalized based on the principle of maximum–minimum normalization. Third, the GNN algorithm is used for the prediction of the pumping unit system efficiency, and the SSA algorithm is used for the optimization of the initial values of the network parameters. Finally, the prediction results are obtained by the maximum–minimum antinormalization. To validate the model's accuracy, this study randomly selected 100 actual oil wells for comparative analysis and analyzed the impact of structural parameters of the hybrid algorithm on the prediction accuracy of system efficiency. The analysis results demonstrate that the proposed model can effectively predict system efficiency and has a certain role in improving the accuracy of oil well pumping system efficiency predictions.
