摘要
水务日需水量涉及多类型复杂因素,深度挖掘分析时易因参数搜索陷入局部最优,导致泛化不足或过拟合等,影响了需水量预测值的准确性。为此,提出基于改进混沌闪电搜索算法的水务日需水量预测方法。通过改进的灰色关联分析法量化评估多个影响因素与日需水量的关联程度,筛选出主导因素并构建时间序列数据集。基于最小二乘支持向量机构建变结构回归模型,将原始输入空间的非线性拟合问题转化为高维特征空间的线性拟合问题,通过引入核函数与偏置参数优化回归方程,提升模型对非线性关系的捕捉能力。引入改进混沌闪电搜索算法,结合混沌搜索的广泛遍历性与闪电搜索的快速收敛性,动态调整放电体种群能量与路径,实现模型参数的全局最优搜索。实验结果表明:该方法在预测精度方面表现出较高的性能,能够准确预测出城市水务的日需水量。
Daily water demand of water management involves multiple types of complex factors,and during deep mining and analysis,it is easy to fall into local optima due to parameter search,resulting in insufficient generalization or overfitting,which affects the accuracy of water demand prediction.Therefore,a method for predicting daily water demand in water management based on an improved chaotic lightning search algorithm was proposed.Quantitatively evaluating the correlation degree between multiple influencing factors and daily water demand through an improved grey relational analysis method,and selecting the dominant factors,a time series dataset was constructed.Based on the least squares support vector mechanism,a variable structure regression model was constructed to transform the nonlinear fitting problem of the original input space into a linear fitting problem of high-dimensional feature space.By introducing kernel functions and bias parameters to optimize the regression equation,the model's ability to capture nonlinear relationships was improved.Introducing an improved chaotic lightning search algorithm,combining the extensive traversal of chaotic search with the fast convergence of lightning search,dynamically adjusting the energy and path of the discharge body population,a global optimal search of model parameters was achieved.The experimental results showed that this method exhibits high performance in prediction accuracy and can accurately predict the daily water demand of urban water management.
作者
包洋洋
王春燕
曾祥纪
魏浩
BAO Yangyang;WANG Chunyan;ZENG Xiangji;WEI Hao(Technology Research and Development Center of Road&Bridge International Co.,Ltd.,Nanjing 210019,China;Road&Bridge Construction Co.,Ltd.,Beijing 100027,China;Research and Development Center on Accelerated Construction,China Communications Construction Group,Nanjing 210019,China)
出处
《国外电子测量技术》
2025年第5期232-237,共6页
Foreign Electronic Measurement Technology
关键词
混沌闪电搜索算法
日需水量预测
改进灰色关联分析
支持向量机
回归模型
chaotic lightning search algorithm
predicting daily water demand
improved grey relational analysis
support vector mechanism
regression model
作者简介
包洋洋,硕士,高级工程师。E-mail:baoyangyang2025@163.com。
