Sensor-Driven Condition-Based Generator Maintenance Scheduling—Part I: Maintenance Problem

Abstract

Traditionally, generator maintenance scheduling has been implemented using highly conservative maintenance policies based on manufacturing specifications and engineering expertise on the type of generators. However, recent advances in sensor technology, signal processing, and embedded online diagnosis provide more unit-specific information on the degradation characteristics of the generators. In this two-paper study, we propose a new generation maintenance framework that integrates the sensor-driven predictive maintenance technologies with optimal maintenance scheduling models. In Part I, we propose a new mixed-integer optimization model for generation maintenance scheduling, which effectively incorporate the dynamic information of generators' health and maintenance cost provided by the Bayesian prognostic models. In Part II, we propose a framework that extends the maintenance model presented herein, and consider the effects of maintenance on network operation by coordinating generator maintenance schedules with the unit commitment and dispatch decisions. We introduce new reformulations and efficient algorithms for solving large-scale instances of the proposed maintenance scheduling model. Extensive computational studies using real-world degradation data demonstrates the effectiveness of the new framework.