Energy consumption prediction for crude oil pipelines based on integrating mechanism analysis and data mining

Abstract

Accurate energy consumption prediction of crude oil pipeline is the basis for energy management and control optimization of oil transportation enterprises. The energy consumption of crude oil pipeline is affected by many factors, which is difficult to predict accurately by mechanism model. Machine learning model is not suitable for small samples and its results lack physical significance. In this paper, mechanism is integrated into machine learning model. A new physically guided neural network (PGNN) is proposed, which is established based on the physical modeling process of energy consumption prediction. The key physical intermediate variables affecting energy consumption are taken as artificial neurons and added to the loss function. The whale optimization algorithm is used to optimize the parameters of the model. A crude oil pipeline in Northeast China is taken as the prediction object to compare different models. The prediction accuracies of PGNN for electric energy consumption and fuel consumption are 2.54% and 4.36%, which are higher than other models. The prediction results of PGNN are more closely correlated with variables that directly affect energy consumption, which proves that PGNN has better physical consistency. In the case of small samples, PGNN has the least decline in accuracy. This study proves the feasibility of PGNN in energy consumption prediction of crude oil pipeline, and provides a new perspective for energy consumption prediction.