Multi-factor Coupling Approach For Power Grid Accident Early Warning

The power grid constitutes a crucial infrastructure for the nation,and the stable operation of the electric power system is essential for maintaining societal order.With the intensification of global warming and environmental issues,natural disasters occur frequently,resulting in increasingly evident impacts on the power grid.Modern power grids are vast and complex,with many electrical equipments directly exposed to the external environment,making them more susceptible to weather and natural elements,potentially causing accidents.Improving the grid’s early warning ability against accidents has become particularly important,requiring a shift from reactive or post-incident strategies to proactive,pre-incident warnings.Every component of the power system is interconnected,and a problem in any link can impact the safe and stable operation of the entire system.The occurrence of power grid accidents is often caused by a combination of various factors,such as equipment failure,operational errors,meteorological conditions,and external environments.This paper investigates the coupling relationship between various disaster-causing factors and power grid accidents,proposing an early warning method for power grid accidents considering multi-element coupling.Based on raw meteorological data,power grid equipment data,and geographical environmental data,we use Conditional Deep Convolutional Generative Adversarial Networks to expand the sample data.Secondly,a feature extraction model is constructed using Deep Convolutional Autoencoders to extract low-dimensional abstract features of disaster-causing factors.Finally,a Deep Denoising Autoencoder network is established as the power grid accident early warning model.The abstract features of the three types of disaster-causing factors are concatenated to form a new feature vector,which serves as the input for the Deep Denoising Autoencoder network.The power grid failure prediction and early warning are realized through unsupervised training and supervised fine-tuning.Firstly,conditional deep convolutional generative adversarial networks are established for meteorological factors,equipment factors,and environmental factors to expand the samples.This resolves issues of inadequate sample quantity and imbalance in the number of samples under different accident categories,ensuring the generalization ability of the model in feature extraction and accident prediction,laying the groundwork for model training in the following sections.Secondly,a deep convolutional autoencoder network is established to extract the features of disaster-causing factors.The convolutional layers used in the deep convolutional autoencoder network can preserve the spatial semantic information of the input signals.The multi-combinations of its internal network parameters can effectively express the internal coupling relationship of various factors.Three feature extraction subnetworks are established for the three types of disaster-causing factors,effectively extracting abstract features.Then,we concatenate the abstract features of disaster-causing factors extracted by the three deep convolutional autoencoder networks.These are used as inputs to the deep denoising autoencoder network.We train the deep denoising autoencoder network in an unsupervised manner,and then add a softmax classifier to the end layer of the encoding part for supervised fine-tuning.This establishes the coupling association between abstract features and power grid accidents,enabling the prediction and early warning of power grid accidents.Finally,in the simulation section,this paper uses the Whale Optimization Algorithm to find the optimal hyperparameters for each network and validates the proposed methods using actual grid accident cases on the Python software platform.The results from the case studies indicate that the power grid accident warning method proposed in this paper can effectively extract abstract features from various factors,providing more accurate recognition and prediction of power grid accidents,thereby comprehensively establishing the coupling relationship between disaster-causing factors and power grid accidents.