Condition evaluation of in-service concrete electric poles based on improved multi-granularity cascading forest

Abstract

Abstract The collapse accident of in service concrete poles has seriously affected the safe operation of the distribution network. Accurate assessment and reinforcement of concrete poles in danger can greatly reduce the occurrence of concrete pole breaking accidents. In this paper, an improved multi granularity cascade forest model is proposed to predict the safety of concrete poles. Firstly, 12 features are selected. In order to reduce the influence of redundant features on prediction accuracy, the importance of random forest features is used to select features; After sorting the features according to Pearson correlation coefficient, the improved multi granularity scanning strategy (IMGSS) is used to scan the features to preserve the correlation between features. Finally, the weak classifier in the original deep forest is improved, and all of them are replaced by extreme random trees. Bayesian method is used to optimize the hyperparameters in the model. The experimental results show that the prediction accuracy of the improved multi granularity cascade forest is as high as 92.62%, which is higher than the traditional machine learning algorithm and can effectively evaluate the safety state of concrete poles.