Synchronous Generators Risk Based Maintenance

Investigators: Miguel Velez-Reyes, Efrain O'Neill-Carrillo and Agustin A. Irizarry-Rivera

Project Description: Electric power generation depends largely on the operation of large synchronous machines.  These generators represent a costly investment for electric utilities, thus it is of utmost importance that any anomaly in their operation is promptly corrected.  On-line estimation of generator parameters is a desirable feature that could aid in better monitoring of the machine behavior.  Another important capability is to establish an adequate maintenance schedule for the generator that ensures proper operation while taking into consideration the cost of having a large generator.  The proposed research work will provide useful alternate parameter estimation techniques and expand previous risk-based maintenance work for the NSF PSERC. 

            Efficient and robust computation of parameter estimates is an important issue in modeling of power systems for monitoring and control.  To estimate model parameters, input/output data is gathered from the system and parameter estimates are computed on-line or off-line. Physical restrictions on the system often limit quality of the available data, and prevent uniformly good estimates of all model parameters from being obtained.  Low-richness data results in ill-conditioning of the parameter estimation problem.  Regulation techniques to stabilize parameter estimates by incorporating prior information in the estimation problem is an alternative to develop reliable and accurate parameter estimation methods for ill-conditioned problems. We propose to compare subset selection and Tykhonov regularization techniques developed at UPRM for synchronous generator parameter estimation to other techniques. 

            Because of their initial cost, proper maintenance of generators is essential.  The existing state-of-the-art offers at least three basic approaches for making the decisions associated with identifying maintenance activities: condition-based maintenance, reliability centered maintenance, and optimization techniques.  The proposed work in risk-based maintenance will include the first two approaches, as applied to synchronous machines.  The first objective is to identify and describe the monitoring technologies, the data collected, and the failure modes.  This will provide models relating these issues to synchronous generator failure probabilities.  The second objective is to identify and describe the maintenance strategies that mitigate the failure modes.  This work will provide models relating these maintenance strategies to generator failure probabilities.

Students involved: Linda Monge