LAYOUT OPTIMIZATION OF BRANCH PIPELINE NETWORK ON CURVED SURFACE USING GENETIC ALGORITHM

摘要

Branch structure is common in oil and gas field gathering system. The rational determination of the pipe network structure and pipelines trend is an important part of the gathering system design. The previous optimization of tree structure is usually studied under two dimensions and the distances between stations, which are represented by straight lines. The study can neither correctly reflect the real situation nor get the optimal solution, because of the dramatic topographic. This paper firstly deals with the shortest route issue on a curved surface. The common way to solve the problem is variational method. However, the method relies on the surface function, which becomes the constraints hard to deal with. The author would obtain the optimal route between two points with the genetic algorithm. Both steady state (SS) and generational (GN) GAs were implemented for the test problem. The performance of a GA generally depends on the selected GA parameters, in particular the crossover and mutation probabilities. Based on the variation of crossover and mutation probability CP and MP in the range of 0.4-1.0 and 0.0005-0.3, the sensitivity of GA to the Gas was therefore established. The best solution has been found by GN GA based on the average evaluation value of the best solutions, which obtained from fifty independent GA runs. According to the case analysis, a 21.73% reduction in total length of the pipeline has been found by using GA. The results presented show that the GA is a robust and stable technique for the solution of route optimization problem. According to further study, the practice of engineering designs are often carried out under two dimensions by using minimum spanning tree algorithm to realize the layout optimization of pipe network. On a curved surface, with the GA above, one can get the trends of each pipe section without changing the connection relationship and the total length of the network is reduced by 12.21%. Yet the connection within a plane cannot guarantee the optimum solution, thus the paper developed the optimization model of branch structure distribution within a curved surface. According to the case analysis, the previous connection relationship between points has been changed, further reducing the total length by 0.97%. In conclusion, the technique in this paper can determine the optimal pipeline laying route, effectively develop the network structure and reduce the total length of pipelines.