中山大学学报（自然科学版）（中英文）2025，Vol.64Issue(2)：22-32,11.DOI:10.13471/j.cnki.acta.snus.ZR20240217
珠江三角洲地区用水量影响要素及其关联规则
On the main influencing factors of water use and their correlation rules in the Pearl River Delta region
摘要
Abstract
Identifying the factors that influence regional water use and the corresponding regulations is crucial for accurately predicting water demand and optimizing the allocation of water resources.This study collected historical data on water resource exploitation and socio-economic statistics in the Pearl River Delta(PRD)region.Two machine learning models,namely Random Forest(RF)and Artificial Neural Network(ANN),were employed to systematically identify the factors affecting water use and to uncover the associated rules in the PRD region.In addition,Shapley Additive Explanations(SHAP)and Partial Dependence Plots(PDP)were applied to enhance the interpretability of the modeling outcomes.The results indicate that the factors influencing water use,in order of importance,are GDP,population size,cultivated land area,per capita water resources,water consumption for actual irrigation per unit of farmland,and urban per capita domestic water use.The average determination coefficients of ANN and RF models are above 0.94 and 0.92,respectively.Regarding water use factors,population is the dominant influence in the central cities,while cultivated land is the principal factor in the surrounding areas.Water use in the PRD region shows the most significant response to the changes in population size and cultivated land area.This research provides a scientific basis and technical support for the future prediction of water demand and the balanced allocation of water resources in the PRD region.关键词
人工神经网络/随机森林/用水量/SHAP方法/PDPKey words
artificial neural network/random forest/water consumption/SHAP method/PDP分类
水利科学引用本文复制引用
郑炎辉,徐小迪,李俊辉,林树彦,何艳虎..珠江三角洲地区用水量影响要素及其关联规则[J].中山大学学报（自然科学版）（中英文）,2025,64(2):22-32,11.基金项目
国家自然科学基金(52209025,51979043) （52209025,51979043）
2024年省级水资源节约与保护专项项目 （）
水利部粤港澳大湾区水安全保障重点实验室开放基金(WSGBA-KJ202302) （WSGBA-KJ202302）
