Scholarlink

<p>Accurate prediction of water evaporation is essential for managing water resources sustainably, especially in dry and semi-dry areas where a lot of water is lost through evaporation. This study introduces a new deep learning model called the Perrin Sequence Convolutional Neural Network (PS-CNN) to forecast evaporation under different environmental conditions. The PS-CNN uses the Perrin mathematical sequence in its convolutional layers to better capture time-based patterns and model complex relationships—this is one of the first times this sequence has been used in a deep learning model. The model was trained and tested using a 10-year dataset of climate information from several monitoring sites across India. The dataset includes factors like temperature, humidity, sunlight, wind speed, and surface features. The results show that the PS-CNN performs much better than traditional methods similar the Penman and Dalton equations. It has a Mean Absolute Error (MAE) of 0.17&#xa0;mm/day, Root Mean Square Error (RMSE) of 0.20&#xa0;mm/day, Mean Absolute Percentage Error (MAPE) of 6.7%, and a Coefficient of Determination (R<sup>2</sup>) of 0.93. These results show that the model is very accurate and reliable in predicting complex evaporation patterns over time and space. Using mathematical sequences in neural networks opens up new possibilities for environmental modeling, and the PS-CNN offers an effective and fast way to support smart water management, planning for irrigation, and strategies to build climate resilience.</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

Water evaporation forecasting using a deep learning model based on Perrin sequence CNN and minimization techniques

  • 期刊论文

  • 发表日期：

    2025年12月12日

查看全文
  • Jaydeep Narayan Kale,
  • Sanjay Kumar Sharma

摘要

Accurate prediction of water evaporation is essential for managing water resources sustainably, especially in dry and semi-dry areas where a lot of water is lost through evaporation. This study introduces a new deep learning model called the Perrin Sequence Convolutional Neural Network (PS-CNN) to forecast evaporation under different environmental conditions. The PS-CNN uses the Perrin mathematical sequence in its convolutional layers to better capture time-based patterns and model complex relationships—this is one of the first times this sequence has been used in a deep learning model. The model was trained and tested using a 10-year dataset of climate information from several monitoring sites across India. The dataset includes factors like temperature, humidity, sunlight, wind speed, and surface features. The results show that the PS-CNN performs much better than traditional methods similar the Penman and Dalton equations. It has a Mean Absolute Error (MAE) of 0.17 mm/day, Root Mean Square Error (RMSE) of 0.20 mm/day, Mean Absolute Percentage Error (MAPE) of 6.7%, and a Coefficient of Determination (R2) of 0.93. These results show that the model is very accurate and reliable in predicting complex evaporation patterns over time and space. Using mathematical sequences in neural networks opens up new possibilities for environmental modeling, and the PS-CNN offers an effective and fast way to support smart water management, planning for irrigation, and strategies to build climate resilience.