Fault Prediction Method For Power Grid Based On Improved Stacked Denoising Autoencoder Network

As an important national public security infrastructure,the power grid is the lifeline for ensuring normal social order and is closely related to the security and stability of modern society.Faults are one of the main forms of power grid disasters.When natural disasters and accidents occur,the power grid will have different degrees of faults,resulting in power outages,seriously affecting social and economic development and the normal life of the people.The focus of power grid disaster prevention and mitigation is to reduce the frequency and probability of faults.Therefore,the historical fault data can be effectively used to analyze grid faults,and timely and effective prediction of grid faults can be realized to determine the corresponding emergency response scheme,which can minimize the impact of faults and significantly improve the level of grid disaster prevention and mitigation.The main cause of power grid faults has been changed from the level of electrical equipment manufacturing process and the level of on-site operation and maintenance to natural weather factors such as lightning,mountain fire,wind and ice disaster.The disaster prevention and mitigation of power grids should also focus on meteorological disasters.Aiming at the correlation characteristics and laws between meteorological and grid faults,this paper proposes fault prediction method for power grid based on improved stacked denoising autoencoder network.Based on meteorological historical data and grid operation and maintenance data,the unbalanced degree of the original data set is reduced by using the synthetic minority sample oversampling algorithm.The subjective and obj ective weighting method obtains the initial combined weight of each hazard factor.The autoencoder network completes the extraction of meteorological information features and the expression of meteorological information and grid fault mapping through deep learning,and establishes the mapping relationship between meteorological information and grid faults.First of all,this paper combines the advantages of subjective and objective weighting methods,and uses the combined weighting to obtain the initial weight of the meteorological hazard factor.Firstly,the subjective weighting of the hazard factors is based on the G1 method,and the order relationship of each hazard factor is determined.The order of importance is sorted and the corresponding experts are used for quantitative analysis to obtain the subjective weight with order preservation.Then,the entropy weight method is used to calculate the objective weight,the initial matrix of the factors of each sample is constructed,and the corresponding entropy and entropy weight are calculated based on the initial matrix,and the objective weight based on the actual sample data can be obtained.Finally,the subjective and objective weights are combined according to the principle of minimum relative information entropy,and the comprehensive weights of the advantages of integrated subjective and objective weights are obtained.The combined weight obtained by subjective and objective weighting will be used as the initial weight of the meteorological hazard factor,laying the foundation for the training of the following model.Secondly,this paper proposes the expression method of grid hazard factors based on stacked denoising autoencoder network.According to the discrete characteristics and coupling characteristics of grid meteorological hazard factors,this paper uses the initial weighting and deep self-coding network to express the grid hazard factors.Initial weighting maximizes the characteristics of meteorological factors.The stacked denoising autoencoder network can perform feature extraction by layer-by-layer learning of input data through a multi-layered coding network,and can express complex coupling relationships within the hazard factor.Finally,using the characteristics of the association between meteorological information and grid faults,this paper proposes a grid fault prediction method.Based on the above-mentioned grid hazard factor expression method,considering the influence of the grid's own factors and external uncontrollable factors,this paper makes full use of the characteristics of stacked denoising autoencoder network deep self-learning.Establish an association mapping relationship between multi-disaster factors focusing on meteorological factors and grid faults to realize grid fault prediction.Taking the actual grid meteorological historical data as an example,this paper simulates the grid fault prediction method proposed in the paper on the MATLAB platform.The analysis of the example shows that the proposed grid fault prediction method based on stacked denoising autoencoder network has better grid fault prediction capability,and can accurately and comprehensively establish the association mapping between meteorological information and grid fault.It can accurately predict the grid faults and lay the foundation for improving the early warning capability of grid faults under complex meteorological conditions.