基于人工智能算法的地浸采区产量预测  

Production Prediction of In-situ Leaching Mining Area Based on Artificial Intelligence Algorithm

作　　者：贾皓 廖文胜[1,2] 杜志明 王亚安[1,2] 王立民 JIA Hao;LIAO Wensheng;DU Zhiming;WANG Yaan;WANG Limin(Beijing Research Institute of Chemical Engineering and Metallurgy,CNNC,Beijing 101149,China;China Nuclear Mining Science and Technology Corporation,CNNC,Beijing 101149,China)

机构地区：[1]核工业北京化工冶金研究院,北京101149 [2]中核矿业科技集团有限公司,北京101149

出　　处：《铀矿冶》2023年第3期10-16,共7页Uranium Mining and Metallurgy

基　　金：中核集团集中研发项目(〔2021〕A100-3);总装和科工局项目(A90-1)。

摘　　要：在地浸采铀生产后期,金属产量持续下降,其铀浓度也受浸采工艺影响而不断变化,抽液量与浓度的乘积不能准确反映金属产量。通过对某矿床C10采区生产数据进行归纳与分析,采用人工神经网络模型对该采区金属日产量进行预测,并使用多元线性回归模型与之比较。结果表明二者均具有良好的精确度和泛化能力,其中人工神经网络模型效果更优,对金属日产量预测平均误差仅为-0.36%,预测结果离群值较少且误差分布均衡。与简单计算金属日产量相比,使用人工神经网络模型预测采区产量,在精度上显著提高,能够为制定生产计划与增产提供理论指导。At the later stage of production in the uranium mining area,the metal production usually continues to decline,and the metal concentration also changes with the influence of the intensified leaching process.At this time,the production cannot be accurately calculated by the liquid pumping volume and the concentration.Based on the induction and analysis of production parameters,a multiple linear regression model and an artificial neural network model are established for the prediction of daily metal production in a single mining area.The results show that both of them have good accuracy and generalization ability,the effect of artificial neural network model is better,the average error of metal daily production prediction is only-0.36%,the outliers are less and the error distribution is balanced.The accuracy of this method is significantly improved compared with the simple calculation to obtain the daily metal production.The use of artificial neural network model to predict the production of the mining area can provide solutions and theoretical guidance for the adjustment and change of leaching process and production system.

关 键 词：地浸铀矿 产量预测 人工神经网络 多元回归 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TD868[自动化与计算机技术—控制科学与工程]