Prediction and assessment of water quality in pit lake based on improved PSO-ELM

Journal Information

J. Hefei Univ. Tech. (Nat. Sci.), Feb.2025, Vol.48 No.2: 145-150

DOI: 10.3969/j.issn.1003-5060.2025.02.001

Authors

SHI Xiufeng $ ^{1,2} $, WANG Jin $ ^{1,2} $, CHUAI Xin $ ^{3} $, WANG Shaoping $ ^{3} $, LUO Changhai $ ^{3} $, YUE Zhengbo $ ^{1,2} $

(1. School of Resources and Environmental Engineering, Hefei University of Technology, Hefei 230009, China; 2. Anhui Engineering Research Center of Industrial Wastewater Treatment and Resource Recovery, Hefei University of Technology, Hefei 230009, China; 3. Nanshan Mining Co., Ltd., Anhui Masteel Mining Resources Group, Ma'anshan 243031, China)

Abstract and Keywords

Abstract: The tailing water generated by the mining industry has a high concentration of metal ions and sulfate, as well as the risk of acidification. The prediction and assessment of tailing water quality is beneficial to guaranteeing the recycling and sustainable development of tailing water resources. In this paper, the linear raw data are transformed into the input matrix of the model by sliding window processing, and the extreme learning machine (ELM) is improved by using the particle swarm optimization (PSO) algorithm to propose a water quality prediction model based on PSO-ELM. Different network models were used to predict the water quality parameters of a pit lake in a mining area in Ma'anshan City, Anhui Province. The results show that the improved PSO-ELM has higher prediction accuracy compared with back propagation (BP) neural network and traditional ELM, with 82% fitting degree and only 0.04 mean square error, and has faster calculation and convergence speed. The training set data was combined with the predicted data, and the Spearman's rank correlation coefficient method was used to evaluate the stability of water quality. The results show that the pH value and the content of major inorganic salt ions are stable, without obvious change trend, which meet the ecological and production needs.

Keywords: water quality monitoring; sliding window; particle swarm optimization (PSO); extreme learning machine (ELM); water quality assessment

作者信息

石秀峰 $ ^{1,2} $，王进 $ ^{1,2} $，揣新 $ ^{3} $，王绍平 $ ^{3} $，罗长海 $ ^{3} $，岳正波 $ ^{1,2} $

(1. 合肥工业大学资源与环境工程学院，安徽合肥 230009；2. 合肥工业大学安徽省工业废水处理与资源化工程研究中心，安徽合肥 230009；3. 安徽马钢矿业资源集团南山矿业有限公司，安徽马鞍山 243031)

摘要和关键词

摘要: 采矿行业产生的尾矿水具有较高的金属离子和硫酸盐质量浓度,同时具有酸化的风险,对尾矿水水质的预测和评价有利于保障尾矿水资源循环利用和可持续发展。文章将线性原始数据通过滑动窗口处理转化为模型的输入矩阵,利用粒子群优化算法(particle swarm optimization, PSO)对极限学习机(extreme learning machine, ELM)进行改进,提出一种基于PSO-ELM的水质预测模型,以安徽马鞍山某矿区坑湖为对象,使用不同网络模型对水质参数进行预测。结果表明,改进后的PSO-ELM模型较BP(back propagation)神经网络、传统ELM具有更高的预测精度,决定系数达到82%,均方误差仅为0.04,并且具有更快的计算和收敛速度。将训练集数据与预测数据相结合,采用Spearman秩相关系数法评价水质稳定性,结果表明pH值和主要无机盐离子质量浓度较为稳定,无明显变化趋势,满足生态和生产需求。

关键词: 水质监测；滑动窗口；粒子群优化算法(PSO)；极限学习机(ELM)；水质评价

Funding

国家自然科学基金联合基金资助项目（U19A20108）；安徽省重点研究与开发计划资助项目（2022107020015）

合肥工业大学学报(自然科学版)

Navigation