Scholarlink

<p>The International Maritime Organization has introduced several policies to reduce fuel consumption (FC) and pollutant emissions from ships, emphasizing the need for predictive models that are grounded in real-world operational data. This study proposed a hybrid modeling approach combining principal component analysis (PCA) and artificial neural networks (ANN) to predict the FC and emissions (CO, HC, NO<sub>x</sub>, PM) of ships equipped with marine diesel engines operating in the Volga–Caspian region. PCA was employed to address multicollinearity in input variables, thereby improving ANN stability and accuracy in predicting emission characteristics. Using data collected from 245 vessels, the developed PCA–ANN-based models achieved high predictive accuracy, with FC predictions reaching R<sup>2</sup> = 0.968 and MAPE below 10%. Emission predictions also met standard acceptability thresholds (MAPE &lt; 20%). The proposed mesoscopic-scale models offer practical utility for fleet-level energy management and regulatory compliance. In other words, these findings demonstrate the prospect of known data-driven prediction hybrid model development in supporting the maritime sector’s transition toward sustainable and compliant operations.</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 hybrid model using PCA and ANN for predicting fuel consumption and emissions of mixed river–sea navigation ships

  • 期刊论文

  • 发表日期：

    2026年3月16日

查看全文
  • Doan Nguyen Cong,
  • Khanh Nguyen Duc,
  • Hiep Nguyen Ha,
  • Vinogradov Sergey Vladimirovich,
  • Yen-Lien T. Nguyen

摘要

The International Maritime Organization has introduced several policies to reduce fuel consumption (FC) and pollutant emissions from ships, emphasizing the need for predictive models that are grounded in real-world operational data. This study proposed a hybrid modeling approach combining principal component analysis (PCA) and artificial neural networks (ANN) to predict the FC and emissions (CO, HC, NOx, PM) of ships equipped with marine diesel engines operating in the Volga–Caspian region. PCA was employed to address multicollinearity in input variables, thereby improving ANN stability and accuracy in predicting emission characteristics. Using data collected from 245 vessels, the developed PCA–ANN-based models achieved high predictive accuracy, with FC predictions reaching R2 = 0.968 and MAPE below 10%. Emission predictions also met standard acceptability thresholds (MAPE < 20%). The proposed mesoscopic-scale models offer practical utility for fleet-level energy management and regulatory compliance. In other words, these findings demonstrate the prospect of known data-driven prediction hybrid model development in supporting the maritime sector’s transition toward sustainable and compliant operations.