Optimierte Informationsverarbeitung in Methoden zur stochastischen Simulation und zur Abschätzung von Parameterwerten: Unsichere zeitabhängige Strömungs- und Transportvorgänge im Untergrund

随机模拟和参数值估计方法中的优化信息处理：地下不确定的时间相关流动和传输过程

• 批准号: 46547152
• 负责人: Professor Dr.-Ing. Wolfgang Nowak
• 金额: --
• 依托单位: Lehrstuhl für Hydromechanik und Hydrosystemmodellierung
• 依托单位国家: 德国
• 项目类别: Research Fellowships
• 财政年份: 2007
• 资助国家: 德国
• 起止时间: 2006-12-31 至 2008-12-31
• 项目状态: 已结题
• 来源: https://gepris.dfg.de/gepris/projekt/46547152?language=en
• 关键词: Optimierte; Informationsverarbeitung; Methoden; zur; stochastischen

Subsurface water quantity and quality is usually predicted using numerical models. Their parameters have unknown spatial distributions which can be estimated from site data using geostatistical inversion techniques. Site data often include time series from well tests or tracer experiments which convey redundant information due to their temporal auto-correlation, leading to excessive computational costs in parameter estimation. Much quicker calculations are possible using temporal moments of time series instead, which is an efficient method to summarize information. Unfortunately, temporal moments still do not constitute independent information, leading to accuracy problems and still allowing further improvement. In this project, temporal moments will be replaced by generalized functionals of time series, including ratios of moments. This is expected to result in a set of independent information units, yielding a basis for rapid and accurate information processing in parameter estimation and stochastic simulation of largescale time-dependent systems. For the first time, concepts from optimal design will be used to optimize information processing and to identify optimal sets of functionals. Also, a method to reconstruct time series from these functionals will be developed. Measured data from a real site will be used to test the developed methods, with applications to parameter estimation and stochastic modeling.

地下水量和水质的预测通常采用数值模型。它们的参数具有未知的空间分布，可以使用地质统计反演技术从现场数据中估计出来。现场数据通常包括来自试井或示踪试验的时间序列，由于其时间自相关性，这些数据传递了冗余信息，导致参数估计的计算成本过高。使用时间序列的时间矩可以更快地计算，这是一种有效的信息汇总方法。不幸的是，时间矩仍然不构成独立的信息，导致精度问题，仍然允许进一步改进。在这个项目中，时间矩将被时间序列的广义函数所取代，包括矩的比率。预计这将产生一组独立的信息单元，为大规模时相关系统的参数估计和随机模拟中的快速和准确的信息处理提供基础。这是第一次，从最优设计的概念将被用于优化信息处理和识别最优的功能集。此外，还将开发一种从这些函数重建时间序列的方法。来自真实站点的测量数据将用于测试开发的方法，并应用于参数估计和随机建模。