Improvement Of PBSO Algorithm Based On Irregular Data And Robust Optimization Of Groundwater Monitoring Network

With the rapid development of computer technology,intelligent computing methods have been more and more widely used.It has been one of the important topics of intelligent computing research to find the optimization methods suitable for all kinds of engineering applications.In different applications,many targeted problems are often involved,such as data input.In this thesis,the application of swarm intelligence algorithm in the layout optimization of groundwater monitoring network is studied,which mainly uses the stochastic multi-objective optimization framework with uncertain input,and makes adaptive changes to it.(1)The traditional optimization of groundwater monitoring network usually considers the selection of monitoring points in the case of discrete data.In order to make the solution process more intelligent,this thesis solves the data in the continuous domain.Therefore,when calculating the relative error,the reference object changes from all the candidate points to the existing monitoring points,and the error calculation also changes from close to the reference value to far away from the reference value.(2)In order to find the optimal solution,this study uses multi-objective PBSO algorithm.In view of the limitations of the original optimal solution guidance strategy of the storm particle swarm optimization algorithm,three schemes are replaced,which are normalization processing,Hungarian matching and Auto Encoder.At the same time,the merging operator is introduced.In practice,due to the limitation of normalization,the combination operator is only used in the optimization algorithm based on Hungarian method and Auto Encoder.The comparison results show that the combination operator gives full play to its advantages.When using Hungarian method strategy,the length of input data needs to be fixed to control the dimension.The experimental results show that the Auto Encoder strategy performs the best in the optimization algorithm with fast convergence speed and lower error value,followed by the Hungarian method.(3)In the process of experimental multi-objective stochastic optimization,in order to better observe the error,the PBSO algorithm based on Hungarian method with fixed input data length is adopted.In order to consider the distribution characteristics of pollution plumes,the resolution based pollution plume filtering is replaced by clustering to obtain a group of representative pollution plumes.Finally,using the above framework method,the layout of 31 monitoring wells in an area of Beijing is optimized.A part of 60000 plumes is selected from the samples,and the plumes obtained from the measurements are used as the test scenario set.The results show that the random implementation of multi scene based on clustering has good robustness.