Pushing back the boundaries to automating the operational hydrological forecasts used in hydropower reservoir management.

水文学的主要挑战之一是在多个时间尺度上对溪流和河流的未来流量进行准确、可靠和有价值的预测。高质量的预报使水资源管理人员能够就其系统的运作作出重要决定。尽量减少洪水事件造成的损失和提高水力发电效率证明水文界在这方面所作的努力是正确的。在过去的20年里，大量的出版物已经解决了这个问题，水文预报正在迅速变得越来越可靠。尽管取得了这一进展，但仍有重大障碍需要克服，特别是考虑到在生成良好的预报（往往被视为最终产品）与在业务中使用预报之间存在的差距。此外，业务水文预报系统通常包括来自经验丰富的预报员的输入，该预报员在预报之前根据他的最佳判断来调整水文模型的初始状态。这种人为干预会在对过去系统性能的分析中产生不一致，从而混淆水文预报的基本统计特性。换句话说，绩效指标被人为干预搞得一团糟，正如俗话所说，“你不能改善你不衡量的东西”。因此，评估水文预报在业务环境中的价值是一项艰巨的任务。因此，预测方面的自动化将允许更深入的分析，并在与油藏系统模拟器配对时允许有针对性的改进。这项研究计划旨在填补这一空白，并最终允许有针对性地改进预测决策过程。主要目标是：+实现水库调度水文预报决策链的自动化;+提高我们对水文预报与水库管理决策支持系统之间相互作用的理解。提出了以下具体目标来推进研究计划：-在分析传统方法的基础上，开发新的方法来改进入库预报方法。- 使用最先进的放水优化算法实现和扩展水电系统模拟器;以及-探索和量化水文预报中包含的价值，以供模拟器中的业务使用。2这项研究计划将深入了解水文预报在水力发电中的价值。考虑到加拿大每年生产超过380 TWh的水电（160亿美元），任何对现状的改善都将带来巨大的经济和环境收益。

One of the main challenges in hydrology is producing accurate, reliable and valuable forecasts on multiple timescales regarding future discharge in streams and rivers. High-quality forecasts enable water resources managers to make important decisions regarding the operation of their systems. Minimizing the damage caused by flood events and increasing hydropower generation efficiency justify the effort deployed by the hydrological community in this regard. Over the past 20 years, a plethora of publications has tackled the issue and hydrological forecasts are quickly getting increasingly reliable. Despite this progress, there are still major roadblocks to overcome, especially when considering the gap that exists between the generation of good forecasts, which are often considered as the end product, and their use in an operational context. Furthermore, operational hydrologic forecasting systems usually include inputs from an experienced forecaster who adjusts the hydrological model's initial states according to his best judgement prior to the forecast. This human intervention generates inconsistencies in the analysis of past system performance and thus obfuscates the underlying statistical properties of hydrological forecasts. In other words, the performance measures are muddled by human intervention and, as the saying goes, “you can't improve what you don't measure”. Consequently, evaluating the value of hydrological forecasts in an operational setting is a daunting task. It follows, then, that automating the forecasting aspect would allow more in-depth analysis and permit targeted improvements when paired with a reservoir system simulator. This research program aims to fill that gap and eventually allow targeted improvements in the forecast-decision process. The main objectives are to:+ Move towards automating the hydrological forecast-decision chain for reservoir operation; and+ Improve our understanding of the interactions between hydrological forecasts and reservoir management decision-support systems.The following specific objectives are proposed to advance the research program:- Develop novel approaches to improve upon inflow forecasting methods based on an analysis of traditional methods in an operational context;- Coordinate the integration of model state initialization within a multi-model forecasting loop;- Implement and expand upon a hydropower system simulator using state-of-the-art water release optimization algorithms; and- Explore and quantify the value contained in hydrological forecasts for operational use within the simulator.This research program will provide insight into the value of hydrological forecasts in hydropower generation. Considering that over 380 TWh of hydroelectricity ($16 billion) are produced annually in Canada, any improvement to the status quo will result in substantial economic and environmental gains.