An optimized algorithm for optimal power flow based on deep learning

With the increasing requirements for power system transient stability assessment, the research on power system transient stability assessment theory and methods requires not only qualitative conclusions about system transient stability but also quantitative analysis of stability and even development trends. Judging from the research and development process of this direction at home and abroad in recent years, it is mainly based on the construction of quantitative index models to evaluate its transient stability and development trend. Regarding the construction theories and methods of quantitative index models, a lot of results have been achieved so far. The research ideas mainly focus on two categories: uncertainty analysis methods and deterministic analysis methods. Transient stability analysis is one of the important factors that need to be considered. Therefore, this paper proposed an optimized algorithm based on deep learning for preventive control of the transient stability in power systems. The proposed algorithm accurately fits the generator's power and transient stability index through a deep belief network (DBN) by unsupervised pre-training and fine-tuning. The non-linear differential-algebraic equation and complex transient stability are determined using the deep neural network. The proposed algorithm minimizes the control cost under the constraints of the contingency by realizing the data-driven acquisition of the optimal preventive control. It also provides an efficient solution to stability and reliability rules with similar safety into the corresponding control model. Simulation results show that the proposed algorithm effectively improved the accuracy and reduces the complexity as compared with existing algorithms. (C) 2021 The Authors. Published by Elsevier Ltd.