A Batch-Driven Scheduling Algorithm of Complex Crude Oil Pipeline Networks: Minute-Level Response Achieved by Strongly-Constrained Decoupling

The generation of scheduling schemes for crude oil pipeline networks is characterized as a spatiotemporal optimization problem subject to strong constraints involving multiple sources, users, products, and batches. Conventional mathematical programming methods, such as mixed-integer nonlinear programming (MINLP), are challenged by high-dimensional equation systems, and the pursuit of global optimality hinders the rapid generation of diverse feasible solutions, while intelligent algorithms such as ant colony optimization are found to be poorly adaptable to variations in pipeline network structures and constraints, resulting in limited portability. A batch-driven scheduling algorithm of complex crude oil pipeline networks is proposed to improve portability and computational efficiency. First, oil terminals/oil fields, pipelines, refineries, and pump stations are defined as four topological invariants through node attribute encoding and adjacency matrices, enabling any network to be mapped into a unified and scalable matrix structure. Second, all oil products are abstracted as discrete “packets”, transforming the continuous flow problem into a batch-based packet dispatching task. Virtual time steps are introduced to integrate movement, sending, and storage operations into a unified processing sequence, thereby achieving the decoupling of continuous product flow from complex operational constraints. Within this framework, a hybrid search strategy combining depth-first search and trie storage is employed to generate feasible paths from sources to refineries, and a stepwise enumeration method is used to exhaustively produce feasible schemes compliant with operational rules, incorporating a dynamic pruning strategy to reduce the actual search space by eliminating constraint-violating branches. Proposed schemes are subsequently validated and expanded through verification of hydraulic and temporal constraints. Through case studies, stable solving times under 10 min are demonstrated across varied network structures. In a complex network involving 13 pipelines, four refineries, two terminals, and 26 products, 122,886 feasible schemes are generated within 300 s─a 144-fold speed improvement over MINLP. Feasibility is verified via Monte Carlo simulation with 10 4 iterations, showing 99.2% constraint satisfaction. Stability is confirmed by a coefficient of variation of 2.9% over 30 trials. The algorithm can be extended to intelligent scheduling in crude oil and new energy pipeline systems, facilitating a transition from experience-driven to algorithm-driven decision-making.