Verified Interval-Based Predictive and Variable-Structure Control for Solid Oxide Fuel Cell Systems: VerIPC-SOFC

区间方法为参数辨识提供了强有力的工具，近年来，具有实时能力的鲁棒控制和状态估计方法的发展受到了极大的关注。这种控制和估计程序尤其基于可变结构和模型预测技术以及基于敏感性的方法。罗斯托克大学机电一体化教席的申请者已经在模拟和实验室实验中成功地展示了这种新方法和实施在选定的应用中的有效性。与实验室环境相比，区间方法的使用尚未在现实生活中的工业应用中得到广泛推广。这一事实的最重要原因是，如果要将区间方法用于离线控制综合和在线控制实现，则需要大量依赖于应用的实现。因此，该项目的前两个资助期侧重于通过区间技术实施用于识别、控制和状态估计的一般例程。在这里，重点是高温燃料电池在非稳定相中的安全运行。基于SOFC非线性动力学的低维面向控制的模型，推导了保证参数辨识的技术，并随后用实验数据进行了验证。由于区间方法的应用导致了对不可测量量的容限，它能够识别系统参数化过程中的结构模糊，并将这些信息提供给保证稳定的控制设计。为了在不确定的情况下实施鲁棒控制技术，并尽快检测对整个系统寿命至关重要的运行条件和退化影响，开发了用于不确定动态系统的实时控制的区间算法。到目前为止，人们的重点是单输入单输出系统的区间控制技术，现在应将其推广到具有多个输入和输出变量的任务，基于这种多变量控制技术，希望从能量的角度优化系统的整体效率。为此，需要将气体预热器和废气催化转化器等外围系统部件集成到动态系统模型以及基于区间的控制设计中。所有建模和控制程序应在模拟中进行数值验证，并在可用的SOFC试验台上进行实验验证。为了向广大受众提供已开发的软件例程，并突出说明如何将区间方法有效地应用于实际控制应用，计划以通用形式发布已开发的源代码。

Interval methods provide powerful tools for parameter identification, where the development of approaches for real-time capable robust control and state estimation has gained much attention in recent years. Such control and estimation procedures are especially based on variable-structure and model-predictive techniques as well as on sensitivity-based approaches. The applicants at the Chair or Mechatronics at the University of Rostock have already succeeded in demonstrating the efficiency of such novel methods and implementations for selected applications in both simulation and laboratory experiments.In contrast to a laboratory environment, the use of interval methods is not yet widely spread in real-life industrial applications. The most important reason for this fact is the large amount of application-dependent implementations that are necessary if interval methods are to be used for the offline control synthesis and for the online control implementation. Therefore, the first two funding periods of this project were focused on the implementation of general routines for identification, control, and state estimation by means of interval techniques. Here, the focus was on the safe operation of high-temperature fuel cells in non-stationary phases. Based on low-dimensional control-oriented models for the nonlinear dynamics of SOFCs, techniques were derived for a guaranteed parameter identification and subsequently validated with experimental data. This identification, leading to tolerance bounds for the non-measurable quantities due to the application of interval methods, is capable of identifying structural ambiguities in the system's parameterization and of providing this information to a guaranteed stabilizing control design.To implement robust control techniques despite this uncertainty and to detect operating conditions and degradation effects as soon as possible, which are crucial for the overall system lifetime, interval algorithms are developed for the real-time control of uncertain dynamic systems. So far, the focus was on interval-based control techniques for single-input single-output systems, which shall now be generalized to tasks in which multiple input and output variables are of interest.Based on such multi-variable control techniques, it is desired to optimize the overall system efficiency from an energetic point of view. For that purpose, peripheral system components such as the gas preheaters and catalytic converters for the exhaust gas need to be integrated into the dynamic system model as well as in the interval-based control design. All modeling and control procedures shall be verified numerically in simulations and validated experimentally on an available SOFC test rig. To make to developed software routines available to a broad audience and to highlight the way how interval methods can be applied efficiently to real-life control applications, it is planned to publish the developed source codes in a generalized form.