نوع مقاله : علمی-پژوهشی
نویسندگان
1 دانشکده مهندسی برق دانشگاه صنعتی سهند تبریز
2 دانشکده مهندسی برق، دانشگاه صنعتی سهند تبریز، تبریز، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Smart grid is a cyber-physical system, a combination of physical devices and computational processes. Enhancing interaction between the cyber and physical layers is crucial for optimizing system operation, management, and security. Motivated by this, in this paper, a framework for solving the optimal scheduling of an AC-microgrid (ACMG) system, is presented. The optimal scheduling of the system is modelled as an optimization problem. Also, the dragonfly is utilized as a powerful optimization technique to solve the proposed optimization problem. On the other side, considering cyber-attacks as a great threat to the system which can cause disruption and outage in smart grids, a deep-learning-based method, long short-term memory (LSTM), along with the concept of prediction interval is utilized to develop a cyber-attack detection model for false data injection attacks on smart meters. In this structure, the optimization is carried out using dragonfly optimization. Also, the LSTM, which is a subset of recurrent neural networks, is designed. With an accuracy of 97%, this model can ensure the cyber-security of the structure. Furthermore, to demonstrate the excellence of the proposed method, it is compared to an Artificial Neural Network (ANN). As the results show, the deep learning LSTM approach outperforms the ANN method in terms of accuracy and cyber-security. The proposed cyber-attack detection model is first trained using historical data and then is used in real-time conditions. For investigating the effectiveness of the proposed approach, the modified IEEE 33-bus test system is utilized. The results significantly show the effectiveness of the proposed methodologies
کلیدواژهها [English]
)