Real-Time Cascading Failures Prevention Through MAS Algorithm and Immune System Reinforcement Learning. Issue 5 (16th March 2017)

• Record Type: Journal Article
• Authors: Belkacemi, Rabie; Babalola, Adeniyi Abdulrasheed; Zarrabian, Sina
• Is Part Of: Electric power components and systems. Volume 45:Issue 5(2017)
• Page Start: 505
• Page End: 519
• Publication Date: 2017-03-16
• Subjects: multi-agent system -- cascading failures -- smart grid -- immune system -- heuristics -- reinforcement learning -- experimental -- line contingency -- classical model -- load shedding -- clonal selection principle -- lookup table; Electric machinery -- Periodicals; 621.3104205
• DOI: 10.1080/15325008.2016.1267280
• Languages: English
• ISSNs: 1532-5008

Abstract

Abstract: The current options power engineers have during a pending CFs event have been limited to those that require load shedding. In this work, a novel algorithm is proposed that avoids loss incurrence as it does not use load shedding and involves real-time power redispatch from the generators through an appropriate combination. This method requires that the system possesses MAS capabilities to intelligently effect the power changes. A tool is needed to select the right combination for each system state as the use of lookup table and heuristics would be ineffective for large systems. The innovative use of AIS, as a reinforcement learning tool and not an optimization tool, is proposed for selecting the appropriate combination. The results obtained when lookup table, heuristics, and AIS are used to select the combination for a real-time experimental system are compared. The test system is the generation and transmission side of IEEE 30-bus system; and the CFs prevention algorithm is activated once the system is in an N-1 contingency state. IEEE 118-Bus System was also used to test the efficacy of AIS-MAS algorithm on large systems. The AIS-MAS algorithm proposed showed great promise for preventing CFs.