(29g) Condition-Based Maintenance Scheduling and Planning for Multi-Component Energy Systems | AIChE

(29g) Condition-Based Maintenance Scheduling and Planning for Multi-Component Energy Systems


Wu, Y. - Presenter, University of Wisconsin-Madison
Maravelias, C., Princeton University
Component degradation in energy and process systems can lead to one of the two undesirable outcomes: catastrophic failure or performance decay. While various optimization methods have been proposed to improve the reliability and availability of these systems [1,2], the negative impact of performance degradation on energy efficiency and production capacity has received limited attention [3,4]. Accordingly, the goal of this work is to develop an efficient optimization framework for maintenance scheduling and planning that targets system performance.

Unlike the conventional preventive maintenance that is performed periodically, condition-based maintenance (CBM) is scheduled based on the actual condition of the system. The CBM strategy usually leads to better system performance compared with other maintenance strategies [5]. However, optimization for CBM scheduling and planning of systems, where the system performance is determined by the combined impact of the health of all components, remains an open challenge. The optimization for CBM scheduling and planning of energy systems is usually more challenging because of time-varying demand, highly nonlinear models that describe component operation, as well as the need to consider long planning horizons.

In this work, we present an optimization framework for the CBM scheduling and planning of multi-component energy systems, considering the non-trivial combined impact of component health on system performance. The optimization framework consists of three stages. At the first stage, we seek a first-principles model that describes the relationships among system performance, component degradation, demand, as well as ambient conditions that might affect system operations. If such a model is not available, we build a data-driven model. Because of computational tractability, in the optimization model, we need to adopt a coarse time discretization while capturing detailed operation profiles. Thus, at the second stage, using the models obtained at the previous stage, we generate surrogate models, which are then used to derive simplified constraints that efficiently consider both the impact of degradation and detailed operation profiles. Each surrogate model describes the impact of component degradation on system performance within a specific time period, considering the demand and ambient condition profiles during this time period. At the last stage, we build a mixed-integer programming model that accounts for the interdependencies among operation, maintenance, and component health. Linear constraints derived from the surrogate models are included.

We show the applicability and performance of the proposed framework through a realistic large-scale case study of a variable flow refrigerant (VRF) system. We show how the proposed methods yield high-quality long-term maintenance schedules.


[1] Basciftci, Beste, et al. "Stochastic optimization of maintenance and operations schedules under unexpected failures." IEEE Transactions on Power Systems 33.6 (2018): 6755-6765.

[2] Besnard, François, and Lina Bertling. "An approach for condition-based maintenance optimization applied to wind turbine blades." IEEE Transactions on Sustainable Energy 1.2 (2010): 77-83.

[3] Weber, Bernd, et al. "Performance reduction of PV systems by dust deposition." Energy Procedia 57 (2014): 99-108.

[4] Kim, Woohyun, and James E. Braun. "Evaluation of the impacts of refrigerant charge on air conditioner and heat pump performance." International journal of refrigeration 35.7 (2012): 1805-1814.

[5] G.P. Sullivan, R. Pugh, A.P. Melendez, W.D. Hunt, Operations & Maintenance Best Practices - A Guide to Achieving Operational Efficiency (Release 3), 2010.