Scholarlink

<p>To manage water resources, reservoirs, and basins, the accurate prediction of evaporation is one of the most significant and necessary processes for the sustainability and preservation of water reserves. On the other hand, climate change can affect the evaporation parameter, so the prediction of evaporation from the reservoirs of dams is vital for managing water resources. development of deep models and their integration to optimization algorithms for modeling and climate variable prediction has drawn a lot of attention in recent years .The present study integrated Multi-Verse Optimizer (MVO), Black Hole (BH), and Marine Predator (MPA) meta-heuristic algorithms with a Long-Term Memory (LSTM) deep neural network model to develop models to forecast the reservoir’s evaporation rate at Dez Dam, Khuzestan Province, Iran. The data series has been prepared in two formats: satellite data and modified data from 1981 to 2018 for 37 years. This study also used the meteorological parameters of surface pressure, maximum, minimum, and average temperatures, the speed of wind, relative humidity, and evaporation. Different scenarios have been investigated to examine the effectiveness of various input parameters on the prediction accuracy of the LSTM, LSTM-MPA, LSTM-MVO, and LSTM-BH models. Predicted results of the models showed that the integrated LSTM-MPA-Satellite model with evaluation criteria of RMSE, MAE, KGE, and WI of 64.3540, 48.7079, 0.8164, and 0.8266, respectively, outperforms other meta-heuristic models. The LSTM-BH-Correct is also introduced as the second-best model with almost similar results of 64.4992, 48.8199, 0.8146, and 0.8262. temperature, Relative humidity, and wind speed are the main variables influencing evaporation. Additionally, the candlestick chart shows that the variables have a more regular state when using meta-heuristic techniques as opposed to using the LSTM model alone, indicating a better fit to the data and improved model optimization. Based on the results, the investigated meta-heuristic integrated models were able to improve the LSTM model results and can be relied upon to predict evaporation in the Dez Dam tank using restricted data. The suggested models are a useful tool for predicting evaporation from the dam reservoir and, as a result, for influencing management decisions that aim to lower evaporation in the research area and other challenging areas of the globe.</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

Reservoir evaporation prediction with integrated development of deep neural network models and meta-heuristic algorithms (Case study: Dez Dam)

  • 期刊论文

  • 发表日期：

    2025年1月27日

查看全文
  • Reza Farzad,
  • Ahmad Sharafati,
  • Farshad Ahmadi,
  • Seyed Abbas Hosseini

摘要

To manage water resources, reservoirs, and basins, the accurate prediction of evaporation is one of the most significant and necessary processes for the sustainability and preservation of water reserves. On the other hand, climate change can affect the evaporation parameter, so the prediction of evaporation from the reservoirs of dams is vital for managing water resources. development of deep models and their integration to optimization algorithms for modeling and climate variable prediction has drawn a lot of attention in recent years .The present study integrated Multi-Verse Optimizer (MVO), Black Hole (BH), and Marine Predator (MPA) meta-heuristic algorithms with a Long-Term Memory (LSTM) deep neural network model to develop models to forecast the reservoir’s evaporation rate at Dez Dam, Khuzestan Province, Iran. The data series has been prepared in two formats: satellite data and modified data from 1981 to 2018 for 37 years. This study also used the meteorological parameters of surface pressure, maximum, minimum, and average temperatures, the speed of wind, relative humidity, and evaporation. Different scenarios have been investigated to examine the effectiveness of various input parameters on the prediction accuracy of the LSTM, LSTM-MPA, LSTM-MVO, and LSTM-BH models. Predicted results of the models showed that the integrated LSTM-MPA-Satellite model with evaluation criteria of RMSE, MAE, KGE, and WI of 64.3540, 48.7079, 0.8164, and 0.8266, respectively, outperforms other meta-heuristic models. The LSTM-BH-Correct is also introduced as the second-best model with almost similar results of 64.4992, 48.8199, 0.8146, and 0.8262. temperature, Relative humidity, and wind speed are the main variables influencing evaporation. Additionally, the candlestick chart shows that the variables have a more regular state when using meta-heuristic techniques as opposed to using the LSTM model alone, indicating a better fit to the data and improved model optimization. Based on the results, the investigated meta-heuristic integrated models were able to improve the LSTM model results and can be relied upon to predict evaporation in the Dez Dam tank using restricted data. The suggested models are a useful tool for predicting evaporation from the dam reservoir and, as a result, for influencing management decisions that aim to lower evaporation in the research area and other challenging areas of the globe.