Batch Optimization Combined With AI Ideas For Refined Oil Pipeline Networks

With the expansion of the scale and the increase in the difficulty of scheduling of the refined oil pipelines network,the existing batch optimization algorithms of the refined oil pipeline network reveal two major problems:The initial one is that the optimization calculation is large and takes a long time;the second one is that some complex operating constraints are difficult to meet.Since the rapid development of artificial intelligence(AI)technology has provided new ideas for solving complex optimization problems in recent years,This paper attempts to use AI technology such as heuristic rules,adaptive search,etc.to better solve these two major problems.Based on the logical structure of batch schedule of refined oil pipeline,the problem of is decomposed into initial station input schedule optimization sub-problem and the distribution /injection schedule optimization sub-problem along the pipeline.Aiming at the problem of optimizing the first station input schedule when the injection volume of each product is unknown,an optimization method based on the initial scheme construction and gradual adjustment is proposed,which incorporates historical pipeline operation data and heuristic rules.In order to reduce the amount of mixed oil,there are usually some restrictive requirements on the location of the oil interface when the pipeline is stopped.There are often multiple oil interfaces in pipelines,and the positions of the interfaces move with the oil flow,so it is difficult to describe the requirements mathematically.Distribution /injection schedule MILP model with a shutdown interface constraint of a refined oil pipeline with multi-sources and multi-depots was first proposed,and the model was solved by GUROBI.From the analysis of results,the artificial rules for optimizing the distribution/injection schedule were summarized.Aiming at the problem of batch optimization of refined oil pipeline network,this paper proposes an improved simulated annealing(ISA)algorithm,which incorporates parallel computing,adaptive search,and heuristic rules for distribution/injection schedule optimization.The ISA algorithm shows that the it can effectively deal with the shutdown interface constraint and has high efficiency.