Ice cover on transmission lines is a significant issue that affects the safe operation of the power system.Accurate calculation of the thickness of wire icing can effectively prevent economic losses caused by ice disa...Ice cover on transmission lines is a significant issue that affects the safe operation of the power system.Accurate calculation of the thickness of wire icing can effectively prevent economic losses caused by ice disasters and reduce the impact of power outages on residents.However,under extreme weather conditions,strong instantaneous wind can cause tension sensors to fail,resulting in significant errors in the calculation of icing thickness in traditional mechanics-based models.In this paper,we propose a dynamic prediction model of wire icing thickness that can adapt to extreme weather environments.The model expands scarce raw data by the Wasserstein Generative Adversarial Network with Gradient Penalty(WGAN-GP)technique,records historical environmental information by a recurrent neural network,and evaluates the ice warning levels by a classifier.At each time point,the model diagnoses whether the current sensor failure is due to icing or strong winds.If it is determined that the wire is covered with ice,the icing thickness will be calculated after the wind-induced tension is removed from the ice-wind coupling tension.Our new model was evaluated using data from the power grid in an area with extreme weather.The results show that the proposed model has significant improvements in accuracy compared with traditional models.展开更多
基金supported by the Science and Technology Project of State Grid Corporation of China(SGXJDK00GYJS2400035).
文摘Ice cover on transmission lines is a significant issue that affects the safe operation of the power system.Accurate calculation of the thickness of wire icing can effectively prevent economic losses caused by ice disasters and reduce the impact of power outages on residents.However,under extreme weather conditions,strong instantaneous wind can cause tension sensors to fail,resulting in significant errors in the calculation of icing thickness in traditional mechanics-based models.In this paper,we propose a dynamic prediction model of wire icing thickness that can adapt to extreme weather environments.The model expands scarce raw data by the Wasserstein Generative Adversarial Network with Gradient Penalty(WGAN-GP)technique,records historical environmental information by a recurrent neural network,and evaluates the ice warning levels by a classifier.At each time point,the model diagnoses whether the current sensor failure is due to icing or strong winds.If it is determined that the wire is covered with ice,the icing thickness will be calculated after the wind-induced tension is removed from the ice-wind coupling tension.Our new model was evaluated using data from the power grid in an area with extreme weather.The results show that the proposed model has significant improvements in accuracy compared with traditional models.
