首页|改进灰狼算法优化GBDT在PM2.5预测中的应用

改进灰狼算法优化GBDT在PM2.5预测中的应用

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
针对灰狼算法易陷入局部最优解和全局搜索能力不足的问题，通过霍尔顿序列(Halton Sequence)搜索算法初始化狼群位置，避免灰狼算法陷入局部最优解和重复运算;引入莱维飞行和随机游动策略对灰狼算法的寻优过程进行优化，以增加算法的全局搜索能力;利用粒子群算法模拟灰狼种群得出的最佳适应度以用于惩罚项改进灰狼算法中的头狼更新策略。使用改进算法优化的梯度提升树(Gradient Boosting Decision Trees，GBDT)模型对北京市大气污染物监测数据中PM2。5质量浓度进行预测，采用3种评估函数对各模型以及混合模型预测效果得分进行评估。结果显示，本文改进的灰狼算法对梯度提升树的优化效果优于其他算法，均方根误差ERMS为6。65 μg/m3，平均绝对值误差EMA为3。20μg/m3，拟合优度(R2)为99％，比传统灰狼算法优化结果的均方根误差减少了 19。19 µg/m3，平均绝对值误差降低了10。03 µg/m3，拟合优度增加了 9百分点;与霍尔顿序列和莱维飞行改进的(Levy Flight-Halton Sequence，LHGWO)相比，改进的灰狼算法预测得分的均方根误差降低了 10。39μg/m3，平均绝对值误差减小了 6。71 µg/m3，拟合优度提高了 5百分点。研究表明了预测模型优化的有效性，为未来城市改善空气质量提供了科学依据和技术支持。
Application of improved Gray Wolf Algorithm to optimize GBDT in PM2.5 prediction
Focusing on the issue that the global search capability of the Grey Wolf Optimization(GWO)algorithm is insufficient and that it tends to converge to locally optimal solutions,this study proposes several improvements to enhance its performance.Firstly,the Halton sequence search algorithm is utilized to initialize the position of the wolf group,aiming to prevent the GWO algorithm from falling into local optimal solutions and repeating operations.Secondly,the Levy flight and random walk strategies are introduced to optimize the optimization process of the GWO algorithm,thereby increasing its global search ability.Additionally,in the GWO algorithm with penalty items,the optimal fitness is determined by simulating the gray wolf population using the Particle Swarm Optimization(PSO)methodology for optimizing the leader wolf update strategy.To evaluate the effectiveness of the proposed improvements,traditional examples are optimized using the modified GWO algorithm.The results are compared with those obtained using the conventional GWO algorithm and another improved GWO algorithm.The findings demonstrate that the proposed modified GWO algorithm in this study outperforms the other algorithms in terms of convergence accuracy and speed.Furthermore,the improved algorithm is employed to optimize the Gradient Boosting Decision Tree(GBDT)model for predicting PM2 5 in Beijing's air pollutant monitoring data.Three evaluation functions are used to assess the prediction performance of each model and the mixed model.The results indicate that the improved GWO algorithm in this study achieves superior optimization for GBDT,with a root mean square error of 6.65 μg/m3,a mean absolute error of 3.20μg/m3,and prediction accuracy(R2)of 99％;Compared to the original GWO algorithm,the root mean square error of the prediction score is reduced by 19.19 μg/m3,the average absolute value error is reduced by 10.03 μg/m3,and the goodness of fit is improved by ten percent;Moreover,compared to the Holden Sequence and Levy Flight improved LHGWO-GBDT,the root mean square error of the prediction score is reduced by 10.39 μg/m3,the average absolute value error is reduced by 6.71 μg/m3,and the goodness of fit is increased by five percentage points.Our study confirms the effectiveness of the Optimized prediction model and also provides the scientific basis and technical support for the improvement of urban air quality in the future.

environmentalologyPM2.5 mass concentration predictionimproved Grey Wolf Optimization(GWO)algorithmGradient Boosting Decision Tree(GBDT)algorithmLevy flightHalton SequenceParticle Swarm Optimization(PSO)algorithm

江雨燕、傅杰、甘如美江、孙雨辰、王付宇

展开 >

安徽工业大学管理科学与工程学院,安徽马鞍山 243002

复杂系统多学科管理与控制安徽普通高校重点实验室,安徽马鞍山 243002

安徽工业大学电气与信息工程学院,安徽马鞍山 243002

湖北大学商学院,武汉 430061

展开 >

环境学 PM2.5质量浓度预测 改进灰狼算法(GWO) 梯度提升树算法(GBDT) 莱维(Levy)飞行 霍尔顿序列(Halton Sequence) 粒子群算法(PSO)

国家自然科学基金项目国家自然科学基金项目安徽省教育厅重点实验室开放课题项目安徽省高校人文社科研究重大项目

7227400171872002CS2022-ZD02SK2020ZD16

2024

10.13637/j.issn.1009-6094.2023.0815

安全与环境学报
北京理工大学 中国环境科学学会 中国职业安全健康协会

安全与环境学报

CSTPCD北大核心
影响因子：0.943
ISSN：1009-6094
年,卷(期)：2024.24(4)
  • 25