Scholarlink

<p>Accurate runoff prediction is essential for water resource management and ecological conservation. This study develops a hybrid model, CPO-PSO-LSSVM, which integrates the global search capability of the Crested Porcupine Optimizer (CPO), the local refinement of Particle Swarm Optimization (PSO), and the regression strength of Least Squares Support Vector Machine (LSSVM). The model was applied to the upper Heihe River Basin, China, using hydro-meteorological records from 1980 to 2009. The results indicate that CPO-PSO-LSSVM outperformed CPO-LSSVM, PSO-LSSVM, and LSSVM. At the Yeniugou station, the proposed model achieved an NSE of 0.917 and an RMSE of 12.936, representing at least an 8% reduction in RMSE compared with other models, while maintaining a correlation coefficient close to 0.96. Similar improvements were observed at other stations. The findings also reveal that the representativeness of meteorological datasets strongly influences prediction accuracy, providing guidance for data source selection. Overall, the proposed framework offers a robust and efficient tool for runoff prediction, with significant implications for regional water resource management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel hybrid CPO-PSO-enhanced LSSVM model for monthly runoff prediction in the upper Heihe River Basin

  • 期刊论文

  • 发表日期：

    2025年9月22日

查看全文
  • XinHao Zhang

摘要

Accurate runoff prediction is essential for water resource management and ecological conservation. This study develops a hybrid model, CPO-PSO-LSSVM, which integrates the global search capability of the Crested Porcupine Optimizer (CPO), the local refinement of Particle Swarm Optimization (PSO), and the regression strength of Least Squares Support Vector Machine (LSSVM). The model was applied to the upper Heihe River Basin, China, using hydro-meteorological records from 1980 to 2009. The results indicate that CPO-PSO-LSSVM outperformed CPO-LSSVM, PSO-LSSVM, and LSSVM. At the Yeniugou station, the proposed model achieved an NSE of 0.917 and an RMSE of 12.936, representing at least an 8% reduction in RMSE compared with other models, while maintaining a correlation coefficient close to 0.96. Similar improvements were observed at other stations. The findings also reveal that the representativeness of meteorological datasets strongly influences prediction accuracy, providing guidance for data source selection. Overall, the proposed framework offers a robust and efficient tool for runoff prediction, with significant implications for regional water resource management.