Design optimization and long-term control of borehole heat exchanger fields in heterogeneous ground under descriptive and predictive uncertainty

地源热泵系统是低焓地热能利用的标准应用。为了满足更大的能源需求，采用了具有多个钻孔热交换器(BHES)的大型系统，这些系统可以访问大量的浅层地面。由于这些技术运行了几十年，而且地面热传输过程非常缓慢，必须格外小心，以避免地面和系统性能的长期退化，这在短期内很难恢复。关于地面条件的描述性不确定性伴随着能源需求的预测性不确定性，而能源需求往往只是粗略地了解和在时间上高度可变。尽管如此，仍然缺乏量化这些不确定性的有效概念。这个项目提出了一种解决不确定性的新方法，它更进一步：目标是即使在相关条件不安全的情况下也能找到最佳的系统解决方案。随着运行的开始，可以很容易地在BHES出口处的循环热载流中监测地面温度的变化。这提供了对真实地面条件的宝贵洞察，这些情况在规划阶段通常是高度不确定的。这一洞察力可以直接用于调整系统的S运行方案。相比之下，选择一种确定性的、现成的方法来设计和静态控制BHE气田，挖掘全部技术潜力的机会有限。在这个项目中，我们提出了一种新的优化和控制程序，该程序明确地考虑了描述性和预测性的不确定性，旨在减轻未知条件的影响，同时优化具有多个BHES的地源热泵系统的整个生命周期的性能。事实上，在这里，无论是在理论上还是在实践中，都很少以显式的方式考虑关键的参数不确定性。为了有效地进行基于不确定性的模拟，开发了一个通用的基于线源的模拟框架，该框架考虑了分层的地面非均质性、地下水流动和不均匀的地热通量。这是一个优化和控制程序，在操作过程中不断学习，理想地通过对每个BHE的单独控制来适应井场中的热交换。比较了基于模型预测控制、贝叶斯学习以及多模型集成概念的不同过程变量，得出了理想的公式。开发和验证是在虚拟现实框架内进行的，以模拟一个假想的、完全为人所知的“真实案例”，并在应用过程中学习隐藏的特征。理论上的、以计算机为基础的发展将是在实地可靠实施的基础。这将在项目的最后阶段进行，届时将在以高分辨率监测的总部外办事处验证以模型为基础的学习程序。

Ground source heat pump systems are standard applications of low-enthalpy geothermal energy utilization. For supply of greater energy demands, large scale systems with multiple borehole heat exchangers (BHEs) are applied that access substantial volumes of the shallow ground. As these technologies are operated for decades, and ground heat transport processes are very slow, extra caution has to be taken for avoiding long-term degradation of both the ground and of system performance, which can hardly be restored in the short-run. The descriptive uncertainty regarding the conditions in the ground is accompanied by the predictive uncertainty in the energy demands that are often only roughly known and highly variable in time. Still, efficient concepts that quantify these uncertainties are lacking. This project presents a new approach to address uncertainty and it goes one step further: The goal is to find optimal system solutions even when relevant conditions are unsecure. With the start of operation, the temperature evolution in the ground can easily be monitored in the circulating heat carrier fluid at the outlet of the BHEs. This offers a precious insight in the true ground conditions, which are commonly highly uncertain in the planning phase. The insight can be directly exploited for tuning the system´s operation scheme. In contrast, choosing a deterministic, off-the shelf approach for design and static control of BHE fields has limited chance of tapping the full technological potential. In this project, we suggest a novel optimization and control procedure that explicitly accounts for descriptive and predictive uncertainty, and that is intended to mitigate the impact of unknown conditions while optimizing the performance of ground source heat pump systems with multiple BHEs for the full life cycle. In fact, here critical parametric uncertainty is rarely considered in an explicit way, neither in theory nor in practice. For efficient uncertainty-based simulation, a versatile line-source based modeling framework is developed that accounts for layered ground heterogeneity, groundwater flow and nonuniform ground heat flux. This is cast into an optimization and control procedure, which continuously learns during operation and ideally adapts heat exchange in the borehole field by individual control of each BHE. Different procedural variants based on model-predictive control, Bayesian learning as well as multi-model ensemble concepts are compared and ideal formulations are derived. Development and validation are carried out within a virtual reality framework, to simulate a hypothetical perfectly known “true case”, with hidden features that are learnt during the course of application. The theoretical, computer-based developments will be the basis for reliable implementation in the field. This will be carried out in the last phase of the project, where the model-based learning procedure will be validated at a BHE field site monitored at high resolution.