Hierarchical modeling of systems with similar components: A framework for adaptive monitoring and control
Hierarchical modeling of systems with similar components: A framework for adaptive monitoring and control
复制标题
具有相似组件的系统的分层建模：自适应监视和控制的框架
DOI：
10.1016/j.ress.2016.04.016
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
J. Z. Kolter
中科院分区：
文献类型：
--
作者：
Milad Memarzadeh;M. Pozzi;J. Z. Kolter
系统管理包括根据可用的观察结果选择维护措施：当系统由已知相似的组件组成时，在其中一个组件上收集的数据也与其他组件的管理相关。这是典型的风力发电场的情况，风力发电场由类似的涡轮机组成。风电场的优化管理是一项重要的任务，因为涡轮机的运行和维护成本很高：在这种情况下，我们最近提出了一种在系统级进行规划和学习的方法，称为PLUS，建立在部分可观察马尔可夫决策过程（POMDP）框架上，将转移和排放概率视为随机变量，因此适合于包括模型不确定性。PLUS将组件建模为独立或相同。在本文中，我们扩展了配方，允许组件之间的相似性较弱。所提出的方法，称为多重不确定POMDP（MU-POMDP），模型的组件POMDP，并假设相应的参数作为相依随机变量。通过这个框架，我们可以校准每个组件的特定降解和排放模型，同时在系统级进行观察。我们比较了所提出的MU-POMDP与PLUS的性能，并讨论了它的潜力和计算复杂度。
System management includes the selection of maintenance actions depending on the available observations: when a system is made up by components known to be similar, data collected on one is also relevant for the management of others. This is typically the case of wind farms, which are made up by similar turbines. Optimal management of wind farms is an important task due to high cost of turbines׳ operation and maintenance: in this context, we recently proposed a method for planning and learning at system-level, called PLUS, built upon the Partially Observable Markov Decision Process (POMDP) framework, which treats transition and emission probabilities as random variables, and is therefore suitable for including model uncertainty. PLUS models the components as independent or identical. In this paper, we extend that formulation, allowing for a weaker similarity among components. The proposed approach, called Multiple Uncertain POMDP (MU-POMDP), models the components as POMDPs, and assumes the corresponding parameters as dependent random variables. Through this framework, we can calibrate specific degradation and emission models for each component while, at the same time, process observations at system-level. We compare the performance of the proposed MU-POMDP with PLUS, and discuss its potential and computational complexity.