Pipeline Network Optimization Based On Cluster Intelligent Algorithm

Due to petrochemical products and reserves and transportation volume are large,the distance of transport is long in the face of such a vast engineering ground engineering system,the impact of changes in design or operation parameters on the total cost is very large.In this case,we need to optimize the system to get the scheme that can save the total cost,which is of great significance to the saving of funds.The well area optimization block selected in this paper is Wutonggou in North 16 well area,In order to adapt to the characteristics of the block,conducted a gathering of planning the initial design.Under the premise of not changing the main gathering and transportation mode,the optimization theory is used to select the optimal transportation plan.In the research of this planning scheme,firstly,the objective function of total cost of gathering and transportation system,taking the pipe diameter of each station site and pipeline in the system as the design variable's mathematical model,considered some constraints.Due to the large number of wells in the block and the large distance between some wells,so before optimization,this paper first k-means clustering method,the wells within the cluster clustering,the 159 wells in the block are divided into 10 well groups.Genetic algorithms are then used in each well group to solve the mathematical model,getting the total length of the pipelines of the stations and gathering systems at all levels under the sub-group,then calculate the diameter of all the pipelines,then the program runs to get the investment cost within each group,and the sum totals the final total cost.During the operation of the genetic algorithm,Changes in the parameters of the genetic algorithm will affect the final investment costs.Therefore,when the program is running,different genetic parameters are used to study the influence of the parameters on the final cost.The genetic algebra selected 300,400,and 500 generations after the function has converged for research,50,100,150 populations were selected for research.As a result,it has been found that under the same evolutionary generation,the two grouping methods can increase the population size and reduce the total cost.And in the same population size,increasing the number of evolutionary algebra can also reduce the total cost.However,changes in evolutionary algebra and population size have little effect on the total cost,and the fluctuation of the total cost is small.However,the cost of the SOM grouping method is slightly lower than that of the K-means grouping.So in the end we chose the final optimization parameters with a genetic algebra of 500 and a population size of 100,the grouping method selects the SOM grouping method as the final optimization group.In the aspect of crossover probability,three probabilities of 0.7,0.6 and 0.5 were selected to research.It was found that the crossover probability had little effect on the final result,and the result was within 0.1%.Therefore,the crossover probability was not used as a parameter that could affect the result of the genetic algorithm.By investigating the genetic algorithm literature,we believe that the mutation probability is generally taken as 0.01,which will not have a great impact on the genetic results.