Research On Risk Distribution Assessment Method Of Wildfire-induced Tripping For Overhead Transmission Lines

Wildfires have become one of the natural disasters that seriously affect the stable operation of power system.During the Spring Festival,burning wasteland,and Qingming Festival,it is prone to wildfires due to the improper usage of fire.Once the wildfire spreads into the transmission corridor,the elevated temperature and ashes in flame would greatly reduce the insulation of air gap,which trip the transmission line to collapse the power grid.Thus,it is of great practical significance to assess the risk of wildfire-induced tripping of transmission lines for differentiated wildfire preventions.First,the location information,as well as 19 kinds of wildfire-related factors,of the wildfire incidents in Guangdong Province happened in 2010-2020 were collected.To evaluate the importance of wildfire-related factors,14 feature selection methods were used.And a heterogeneous selector set was obtained considering the independence.By using the BP neural network,the subset with the optimal features was obtained.After the ensemble feature selection,the accuracy of the BP neural network model was 85.96%,which means that the BP neural network model with the optimal features can evaluate the risk of wildfire occurrence effectively.Secondly,a method based on the flame combustion model and air discharge model was proposed to assess the tripping risk of transmission lines.Two flame combustion models,which are Wang Zhengfei model and McArthur model,in assessing the risk of wildfire-induced tripping in the study area was compared.Considering the difference of fire intensity,the withstand voltage of air gaps between line and ground and between lines was calculated.The tripping risk of transmission line was then evaluated according to the ratio of the withstand voltage under the maximum flame height to the applied voltage of gap.The verification results show that the accuracy and recall of the proposed model is 80.3%and 82.76%,seperately.Thirdly,considering the complexity of wildfire-induced breakdown,a tripping risk assessment method of transmission lines was proposed based on data mining.Aiming at the unbalanced categories of samples,an ensemble dataset based on random undersampling was constructed.And three data mining algorithms including Back Propagation Neural Network(BPNN),Support Vector Machines(SVM)and Random Forest(RF),were used to study the relationship between factors and the tripping risk.The results showed that the BPNN,SVM and RF algorithms perform the best with the ensemble dataset,among which the RF had a precision of 87.12%and a recall of 89.66%.Finally,the application of two wildfire-tripping-risk assessment methods were compared.And an assessment method of wildfire-induced tripping risk distribution was proposed by combining the risk assessment models of wildfire occurrence and line tripping.The case in Guangdong Province showed that this method can realize the visualization of risk distribution.It benefits extraction of line sections with high-risk of wildfire-induced tripping and development of differentiated fire prevention measures.