A Data-Driven Two-Stage Dynamometer Diagram Inference Approach for Variable Operational Conditions in Pumping Machines

In recent years, the soft measurement and state estimation of key parameters of industrial control systems and critical infrastructure have become crucial. The acquisition of dynamometer diagrams for pumping machines holds great significance in monitoring the operational status of drilling industrial control systems. To achieve the accurate inference of dynamometer diagrams from electrical parameters under variable operational conditions, this paper proposes a data-driven two-stage approach. Specifically, without prior knowledge, this paper proposes an electrical parameter derivation diagram approach based on extreme gradient boosting (XGBoost), which can accurately online suspended load and displacement estimation, thereby enabling efficient dynamometer diagram deduction. Furthermore, facing the variable operational conditions in actual engineering, this paper proposes a two-stage strategy. From a fine-grained perspective, the XGBoost algorithm is constructed separately for the operational condition categories and dynamometer diagram elements. This paper verifies the proposed dynamometer diagram approach under variable operational conditions through the real-world data provided by Daqing Oilfield Research Institute. The experimental result demonstrates the advantages of the proposed approach.