Searching for Critical Power System Cascading Failures with Graph Convolutional Network
ABSTARCT :
Power system cascading failures become more time variant and complex because of the increasing network interconnection and higher renewable energy penetration. High computational cost is the main obstacle for a more frequent online cascading failure search, which is essential to improve system security. We propose a more efficient search framework with the aid of graph convolutional network (GCN) to identify as many critical cascading failures as possible with limited attempts. The complex mechanism of cascading failures can be well captured by training a graph convolutional network (GCN) offline. Subsequently, the search for critical cascading failures can be significantly accelerated with the aid of the trained GCN model. We further enable the interpretability of the GCN model by a layer-wise relevance propagation (LRP) algorithm. The proposed method is tested on both the IEEE RTS-79 test system and China’s Henan Province power system. The results show that the GCN guided method can not only accelerate the search of critical cascading failures, but also reveal the reasons for predicting the potential cascading failures.
EXISTING SYSTEM :
? Several existing studies also formulate the islanding problem as a mixed-integer linear programming (MILP) problem, distinctly handling power balance in each created island for a feasible solution.
? In these approaches, the decision on the island margins and generator configuration is taken with respect to the balance of the load-generation and the minimization of load shedding, while ensuring that system constraints are not violated.
? Machine learning (ML) is an area of artificial intelligence that utilises existing datasets to make predictions on unseen data.
DISADVANTAGE :
? Although these methods can significantly improve the sampling efficiency compared to random sampling, they still suffer from duplicated simulations of the same cascading paths. To avoid duplicated simulations, some methods model the cascading failures simulation as a Markov chain search problem.
? Because of the complex mechanism of cascading failures, it still remains a challenging problem to effectively search for the cascading failure paths that result in load shedding.
? Two vital problems remain to be addressed to build better machine learning models. One challenge is that the number of model parameters increases significantly with the size of the power system, making the training process time-consuming and prone to overfitting.
PROPOSED SYSTEM :
• Some machine learning models are proposed to capture the complex mechanism of cascading failures.
• The proposed GCN approach uses graphical model to learn this mechanism and theoretically requires far less parameters than other machine learning models (DNN or CNN) .
• To the best of the authors’ knowledge, the paper is the first to implement the interpretability of GCN models in power grid analysis.
• The proposed algorithms can identify the contributing factors that cause the cascading failures, such results could help manage the power system to mitigate the damage.
ADVANTAGE :
• The hyper-parameters of the GCN model are tuned by 5-fold cross-validation , where we tune the hyper-parameters by grid search and choosing the hyperparameters with the best performance on the cross-validation set.
• The hyper-parameters of the GCN model are tuned by 20% of the training data.
• we present the search efficiency on a set of new load profiles of the power system of Henan province. The search performance of method Rand, PFW, LODF, and GCN .
• The above approach is a hybrid search method of machine learning and physical rules, which shows promising performance in case studies.
|