基于智能算法的矿区土壤修复采样优化

Sampling optimization in mine soil remediation based on intelligent algorithm

针对废弃矿区土壤的修复,同一地区周期性的采样评价是验证修复效果必不可少的手段。而采集及分析样本的过程耗时、耗力且昂贵;如何能在不影响整体评价精度的基础上,最大限度地降低后续采样点的数量,对土壤修复评价成本的降低有直接的好处。以广西某县的一重金属铅(Pb)污染严重的废弃矿区为研究对象,基于45个原始采样点,采用了RBF神经网络拟合了Pb的浓度空间分布,并采用遗传算法对采样点进行了进一步优化。在构建网络结构的过程中,根据种群个数及迭代次数的不同提出了6个优化方案。结果表明,种群个数为40且迭代次数为100的方案F的优化结果生成的分布图与原始分布图最接近,误差均方差为0.501 3,且优化后的样本数量减少到25个,减幅达44%。该结果反映了遗传算法能为空间变异缓慢地区的后续采样方案提供科学建议,明显降低采样点数量及分析成本。

For soil remediation in abandoned mines,periodic sampling and evaluation is important to verify the effectiveness of the remediation. However,it is time-consuming,labor-intensive,and costly to collect and analyze such samples. To reduce maximally the sampling number for subsequent evaluations without lowering the evaluation precision would directly cut the budget of the soil remediation. An abandoned mine with serious Pb pollution,located in Guangxi,was studied. Based on 45 original samples,an RBF neural network fitting method was used to produce a density map of Pb,and then a genetic algorithm was applied to optimize this sample set. When building the network structure of the genetic algorithm,six scenarios were proposed according to different settings of population number and iteration times. The optimization results showed that a scenario with a population of 40 and iteration time of 100 generated a density map that had the highest similarity with the original density map,with error variance of 0. 501 3 and the optimized sampling number of this scenario reduced to 25,which is 44%of the original number. The results indicated that the genetic algorithm reduced the subsequent sampling numbers needed and analyzing cost for an area with slow spatial variation.