- 作者: 黃世杰; 黃慶連
- 作者服務機構: 高雄工學院電機工程研究所; 成功大學電機工程研究所
- 中文摘要: 本文應用遺傳演算法加強型之神經網路評估電力系統穩態安全度。此方法乃將多層感知器與遺傳演算法相結合,借助遺傳演算法之尋優能力以協助找出神經網路之最佳起始點,因而減少神經網路學習停滯之機率,並大幅提昇整體計算效率。本文將所提之方法應用於實際電力公司運轉資料,執行結果十分滿意,證實此計算方法兼具可行性及實用性。
- 英文摘要: The security assessment task can be categorized as a classification problem. The combination ofpower system topologies, operating states, system constraints and disturbances determine the security statusof power systems. Such patterns can be utilized by neural networks (NN) to capture some underlyingrelationships between the pre-contingency system patterns and the post-contingency security status. Oncethey are adequately trained, the networks can classify unseen patterns with minimum computation time,thus paving the way for on-line assessment. A new method of embedding genetic algorithms into neural networks to solve power system staticsecurity assessment problems is proposed in this paper. In this proposed approach, genetic algorithmsare applied to determine the initial weights of neural networks, thereby guiding the neural network toa near-optimal initialization. These well-initialized networks are then trained with back error propagationalgorithms. Using this proposed approach, the local minimum phenomenon, which may cause the learningprocess to stagnate, can be avoided. Overall learning performance is, thus, significantly improved. Theproposed method has been tested using utility data provided by the Taiwan Power Company (Taipower)Results show the potential of the proposed method for applications.
- 中文關鍵字: neural networks; genetic algorithms; security assesssment
- 英文關鍵字: --