Scholarlink

Water scarcity is one of the main challenges that Gulf Cooperation Council (GCC) countries have been facing for multiple decades. The challenge grows bigger as the region experiences rapid social and economic transformation, leading to an increase in water demands. Out of the various solutions that were implemented to face the challenge, demand and supply side management were heavily investigated. The balance between water supply and demand requires efficient water management system techniques, which are highly dependent on the use of accurate forecasting methodologies and tools. Accurate demand forecasting is a necessary input to many water processes, such as determining the required water reserve precisely as well as developing optimum operational plans for pumping stations and water production plants. There is no single global optimum method used to forecast water demand. It is more of a case-by-case approach depending on the network complexity, operational limitations, available data, forecast horizon, intuitiveness of the tool, and accepted percentage of deviation between actual and forecasted demand. The purpose of this paper is to achieve the global sustainability goal by addressing the gap that is currently present between water supply and demand. This is done by presenting an innovative and accurate methodology to forecast water demand using machine learning (ML) for the short-term. The results for a water utility in the United Arab Emirates (UAE) showed that the mean absolute percentage error (MAPE) between the actual and forecasted demand was reduced from 5.42% for the conventional forecasting method to 3.25% for the proposed ML forecasting method. Similarly, the root mean square error (RMSE) was reduced from 11.14 million imperial gallons per day (MIGD) for the conventional forecasting method to 7.6 (MIGD) for the proposed ML forecasting method. Additionally, the total difference per year between the actual and forecasted demand was reduced from 2683 million imperial gallons (MIG) for the conventional forecasting method to 898 (MIG) for the proposed ML forecasting method. This shows that by having an accurate demand forecast, the gap between actual and forecasted demand can be reduced, which will improve the overall efficiency and performance of the water management system.

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

Reliable Short Term Water Demand Forecast Using Machine Learning—Part 1

  • 章节

  • 发表日期：

    2025年

查看全文
  • Waleed Eldamaty,
  • Mohammed Abdallah,
  • Khalid Al Zaabi

摘要

Water scarcity is one of the main challenges that Gulf Cooperation Council (GCC) countries have been facing for multiple decades. The challenge grows bigger as the region experiences rapid social and economic transformation, leading to an increase in water demands. Out of the various solutions that were implemented to face the challenge, demand and supply side management were heavily investigated. The balance between water supply and demand requires efficient water management system techniques, which are highly dependent on the use of accurate forecasting methodologies and tools. Accurate demand forecasting is a necessary input to many water processes, such as determining the required water reserve precisely as well as developing optimum operational plans for pumping stations and water production plants. There is no single global optimum method used to forecast water demand. It is more of a case-by-case approach depending on the network complexity, operational limitations, available data, forecast horizon, intuitiveness of the tool, and accepted percentage of deviation between actual and forecasted demand. The purpose of this paper is to achieve the global sustainability goal by addressing the gap that is currently present between water supply and demand. This is done by presenting an innovative and accurate methodology to forecast water demand using machine learning (ML) for the short-term. The results for a water utility in the United Arab Emirates (UAE) showed that the mean absolute percentage error (MAPE) between the actual and forecasted demand was reduced from 5.42% for the conventional forecasting method to 3.25% for the proposed ML forecasting method. Similarly, the root mean square error (RMSE) was reduced from 11.14 million imperial gallons per day (MIGD) for the conventional forecasting method to 7.6 (MIGD) for the proposed ML forecasting method. Additionally, the total difference per year between the actual and forecasted demand was reduced from 2683 million imperial gallons (MIG) for the conventional forecasting method to 898 (MIG) for the proposed ML forecasting method. This shows that by having an accurate demand forecast, the gap between actual and forecasted demand can be reduced, which will improve the overall efficiency and performance of the water management system.