Fusion Mechanism-data Model and Gene Pool-based Approach for Efficient Multiproduct Pipeline Pump Optimization

Abstract Pipelines typically convey multiproduct products, which are crucial for maintaining downstream energy supplies. The pipeline sequentially transports various sorts of oil products. Rigorously monitoring the optimization of multiproduct pump scheduling with consideration of pipeline working points is a complex optimization problem defined by multiple variables, constraints, objective functions, and intricate nested control equations. The optimization model depending on empirical formulas is computationally costly, inadequately accurate, and displays low solving efficiency. This research mixed empirical formulas with machine learning algorithms to enhance the computational accuracy and speed of the model, thereby constructing a surrogate model for simulating the working points. This facilitates rapid and precise simulation computations of the working points in the optimized multiproduct pipeline model. Then, a genetic algorithm (GA) with integer encoding was implemented, and the gene pool construction based on heuristic strategies was applied to minimize the algorithm’s search of the infeasible domain, thus enhancing the solving efficiency. Eventually, the defined approach was implemented in a real-world multiproduct pipeline. The solution results verify the efficiency of the method proposed in this work. This article presented an approach that can efficiently develop an economical and safe strategy for scheduling pumps of multiproduct pipelines, delivering technical assistance for the low-carbon, eco-friendly, and intelligent management of multiproduct transportation.