Guaranteed Load Prediction for Distribution Network Considering Extreme Disasters

Ziheng Hu,Chang Ye,Shaorong Wang,Changshu Tan,Jie Tian,Yan Li

DOI: https://doi.org/10.1109/ispec53008.2021.9735791

2021-01-01

Abstract:The concept of resilience is always used to describe the ability of the distribution network to withstand disasters. To evaluate the resilience of the distribution network, it is of great practical significance to predict the load level that can ensure power supply after the occurrence of extreme disasters. This paper first defines the concept of guaranteed load level under extreme disasters in the distribution network and uses it as the quantitative resilience index of the distribution network. On this basis, this paper proposes a method of predicting the guaranteed load level of the distribution network under extreme disasters. In the proposed method, after a comprehensive analysis of various factors affecting the guaranteed load level of the distribution network after the typhoon disaster, the random forest method is used to calculate the feature importance degree of various influencing factors. Then the key influencing factors are selected according to the importance of the characteristics. Then, the paper built a random forest regression model for the guaranteed load level of the distribution network under typhoon disaster, and the model parameters were obtained by grid search and cross-validation to get the optimal prediction model. Finally, the proposed method is tested with the distribution network data affected by typhoons at home and abroad, and the prediction results of the other seven regression models are compared and analyzed. The comparative analysis results show that the proposed method has better applicability and higher accuracy and is more consistent with the actual load level of the distribution network after extreme disasters.