Scheduling Post-Disaster Power System Repair With Incomplete Failure Information: A Learning-to-Rank Approach
- 10 February 2022
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Systems
- Vol. 37 (6) , 4630-4641
Abstract
This paper proposes a novel repair rule set (RRS) for scheduling the power system infrastructure repair after the occurrence of extreme events. RRS is made up of multiple repair rules, each of them can be applied in arbitrary post-disaster failure scenarios to rank the repair actions by priority. A learning-to-rank technique called AdaRank is used to train the repair rules by combining the weak learners derived from the dynamic repair scheduling model. Then, RRS is constructed by iteratively clustering the training cases and retraining the repair rule for each cluster. Increasing the number of repair rules within RRS allows it to differentiate various types of failure scenarios, thereby improving its performance. Further combined with multi-label K nearest neighbor (ML-KNN) technique, RRS is able to schedule the repair without the full knowledge of real-time failure information, such as the estimated repair time. The results of case studies on IEEE-118 test systems show that the proposed method has a desirable performance compared to the exact mathematical optimization model. Moreover, it reduces the requirement for failure information while significantly improving the computational efficiency.Keywords
Funding Information
- National Natural Science Foundation of China (51725701, 52022016)
This publication has 34 references indexed in Scilit:
- Transmission system restoration with co-optimization of repairs, load pickups, and generation dispatchInternational Journal of Electrical Power & Energy Systems, 2015
- Transmission system repair and restorationMathematical Programming, 2015
- Proactive Recovery of Electric Power Assets for Resiliency EnhancementIEEE Access, 2015
- Stochastic Pre-hurricane Restoration Planning for Electric Power Systems InfrastructureIEEE Transactions on Smart Grid, 2015
- Restoring infrastructure systems: An integrated network design and scheduling (INDS) problemEuropean Journal of Operational Research, 2012
- A genetic algorithm for the unrelated parallel machine scheduling problem with sequence dependent setup timesEuropean Journal of Operational Research, 2011
- Event-based MILP models for resource-constrained project scheduling problemsComputers & Operations Research, 2011
- Kernel density estimation via diffusionThe Annals of Statistics, 2010
- ML-KNN: A lazy learning approach to multi-label learningPattern Recognition, 2007
- An efficient k-means clustering algorithm: analysis and implementationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2002