Abstract
This paper presents a case-based reasoning (CBR) technique, a novel approach for voltage security assessment and optimal load shedding in real-time environment. The CBR extracts the information of voltage security from large sets of possible operating conditions of the power system, which are screened off-line by randomly varying the real and reactive loads at PQ buses. The severity of contingency is then assessed by the number of insecure operating states, which are predicted by the CBR system under that contingency. Security classes are defined by threshold value of maximum loadability margins (MLM), calculated using continuation power flow (CPF) method. For insecure states, the presented technique also gives optimal load shedding plan to bring the system from insecure to secure region. Decision trees (DT) are used for case indexing to reduce the search time to obtain the set of relevant cases, which are in thousands of numbers. In addition, decision trees also provide the information about those case features, which mostly influence the severity of a given contingency of the power system. Unlike other non-deterministic approaches, the proposed method is easy to understand for operators, easy to modify under the change of topology of the power system and has continuous learning capability. The effectiveness of the proposed approach is tested on IEEE test systems. The proposed CBR approach provides a fast and accurate voltage security assessment for unknown load patterns and a load shedding plan for insecure cases to mitigate the voltage collapse. Hence it found to be suitable for on-line applications of energy management systems.
