A new data-driven model for urban water demand forecasting

尽管加拿大的水资源相对丰富，但供水商对管理用水需求很感兴趣。 比以往任何时候都要好。从这个意义上说，我们的合作伙伴Econics与地方政府、自来水公司和 市政当局。加拿大城市供水(UW)部门的努力正朝着更加一体化和 需求侧管理方法，以控制和扩大其供水系统(WSS)。这是公认的 需要可用于更有效和可持续地规划和管理UW供应的新工具 系统。一种这样的工具是最先进的和高度准确、精确和可靠的UW需求(UWD)预测 模特们。它们为水资源和公用事业管理者提供支持，以适应短期内的变化-或 UWD的长期预测。数据驱动模型是此类建模的主要类型。在过去两年中 几十年来，已经检验了几个这样的统计模型，主要用于短期的UWD预测。 然而，数据驱动的模型还没有被用于UWD的长期预测，特别是在 不同的气候变化情景(CCS)。应审查后者，并应将其对普遍定期审议的影响 已评估。该项目的主要目标是为UWD的长期预测开发数据驱动的模型，而 考虑特定城市的长期CCS及其对UWD的预期影响。到目前为止，还没有研究 检验了人工神经网络模型在CCS条件下对UWD长期预测的性能。尽管有好的一面 尽管人工神经网络模型的性能不高，但它们有其局限性，特别是对于非平稳数据。几项研究已经 显示了与小波(W-ANN)相结合时有希望的性能结果以及使用 集合神经网络(ENN)，其中ENN是更健壮、一致和可靠的模型。这样做的主要目标是 研究项目是开发W-ENN模型，用于不同CCS下的UWD长期预测。这个 新的、高度准确和可靠的UWD长期预测模型将为Econics提供，反过来，他们的 客户拥有非常有用的模型，将允许最准确、精确和可靠的UWD长期 预测，因此将有助于有效和可持续地规划和管理UW供应系统。

Despite the relative abundance of water in Canada, water suppliers are interested in managing water demand than ever before. In this sense, our partner, Econics, works with local governments, water utilities and municipalities. Efforts in the Canadian urban water (UW) supply sector move towards more integrated and demand side management approaches to control and expand their water supply systems (WSS). It is recognized the need for new tools that can be used to more effectively and sustainably plan and manage UW supply systems. One such tool is state of art and highly accurate, precise and reliable UW demand (UWD) forecasting models. They provide support for water resources and utility managers to adapt to changes in the short-or long-term UWD forecasts. Data-driven models represent the main type of such modeling. During the last two decades, several of such statistical models have been examined for UWD forecasting mainly for short-term. However, data-driven models have not been examined for UWD long-term forecasting, especially under different climate changes scenarios (CCS). The latter should be examined and their impact on UWD should be evaluated. The main goal of the project is to develop data-driven models for long-term UWD forecasting, while considering long-term CCS and their expected impacts on UWD for a given municipality. To date, no studies have examined the performance of ANN models for long-term UWD forecasting under CCS. Despite the good performance of ANN models, they have limitations, particularly with non-stationary data. Several studies have shown promising performance outcomes when combined with Wavelets (W-ANN) as well as the use of ensemble ANNs (ENN) where ENNs are more robust, consistent and reliable models. The main goal of this research project is to develop W-ENN models for the long-term UWD forecasting under different CCS. The new, highly accurate and reliable UWD long-term forecasting models will provide Econics, and in turn their clients with very useful models that will allow the most accurate, precise and reliable UWD long-term forecasting and will consequently help in effectively and sustainably plan and manage UW supply system.