Quantum mechanics

Random forest

Engineering

Reliability engineering

Transmission tower

Power network

Power transmission

Distribution (mathematics)

Typhoon

Power (physics)

Data mining

Computer science

Grid

Physics

Mathematics

Geography

Electric power system

Environmental science

Mathematical analysis

Artificial intelligence

Geometry

Tower

Structural engineering

Meteorology

Artificial neural network

International Journal of Electrical Power & Energy Systems

• We develop an optimal data-driven tower damage prediction model to predict the damage spatial arrangement of 10 kV towers under typhoon disaster. The model has a cell accuracy of 1 km × 1 km. • Metrological, power grid, and geographical information are comprehensively considered when predicting the damage spatial arrangement of 10 kV towers. Meanwhile, sample balancing techniques and correlation analysis are used for data processing and variable selection. • The optimal algorithm Gradient Boosting Regression (GBR) is selected by comprehensively considering the subjective and objective factors. It is able to find a right balance between the prediction accuracy and the timeliness. This paper presents a data-driven tower damage prediction model to predict the damage spatial arrangement of 10 kV towers under typhoon disaster. The 10 kV tower belongs to distribution network in China. Compared with high voltage level in transmission network, the 10 kV power towers are more vulnerable to typhoon disasters due to their lower design criterion and large amount. The data-driven model proposed in this paper can effectively predict the tower damage situation. It takes meteorological, power grid and geographic information into account and can be divided into two steps. The first step is to predict the damage probability of each tower by using data-driven method such as AdaBoost, Gradient Boosting Regression, K Nearest Neighbor Regressor, Random Forest and Support Vector Regression algorithms. In this step, the data space is constructed by data processing and variable correlation analysis. Then, we use the processed meteorological, power grid and geographic information as input and the damage probability as output for model comparison. The second step is to select the optimal model based on comprehensive index weighting. Through a comprehensive comparison of the efficiency and accuracy of the five models in various actual scenarios, the optimal model is Gradient Boosting Regression, which outperforms the other adverse algorithms and produces the prediction damage consistent with actual data. © 2017 Elsevier Inc. All rights reserved.

Damage prediction of 10 kV power towers in distribution network under typhoon disaster based on data-driven model