| China’s water supply network has been laid for a long time,with high investment costs and complex topology.The problem of leakage in the water supply network has been plagued by water supply companies.At present,Chinese water supply companies mostly rely on the working experience of the staff to manage the leakage of the water supply network.The problems are: subjective delineation of vulnerable areas,and it is difficult to quantitatively determine the vulnerable pipelines and the amount of leakage.Therefore,the establishment of an accurate water supply pipeline leakage risk analysis model and a leakage estimation model has practical significance for the water supply enterprise to achieve leakage management and control.In this paper,the random forest algorithm is used to construct the leakage risk analysis model of the water supply pipe network,and the Fast ICA algorithm is used to construct the leakage estimation model.Taking a coastal city as the research area,the established model is used to analyze the leakage risk of the pipeline network and calculate the leakage amount.The specific research contents are as follows:1.The question whether a coastal city water supply pipeline leakage occurred regarded as binary classification.Based on the collected data of water supply pipeline,the classification accuracy of random forest algorithm,logistic regression algorithm,support vector machine algorithm and BP neural network algorithm is calculated and compared.The calculation results show that the classification accuracy of the random forest algorithm is better than that of the other three types of algorithms,so the random forest algorithm is selected to establish a leakage risk analysis model for the water supply network.During the modeling process,the operating parameters are optimized,and the optimization process is visually displayed.The optimization results show that parameter tuning can effectively improve the accuracy of the model.The output results of leakage risk analysis model are combined with Arc Map software to draw the thematic map of leakage possibility of water supply pipeline for visual display.2.There are many influencing factors for water supply pipelines.Based on the domestic and foreign research results in this article,combined with the basic database of water supply pipelines in research area,the selected pipe diameter,pipe material,pipeline depth,road load,water pressure,and pipe age are random Input variables of the forest model.3.Using Fast ICA Algorithm to Establish a Leakage Estimation Model.Taking a residential area in a coastal city as the research area,build a laboratory example pipe network to simulate the water supply conditions of the residential area.The operating data of the pipeline network using laboratory examples prove that the Fast ICA leakage estimation model can estimate the leakage.In the engineering example,the applicability of the Fast ICA leakage estimation model and the wavelet transform algorithm for the calculation of the leakage is compared.The example proves that the simulation effect of Fast ICA model is better than that of wavelet transform model.The relative error between the leakage calculated by the Fast ICA model and the actual leakage is smaller,and the similarity of the change trend is higher. |