基于数据驱动的配电网停电预测模型

Data-driven predictive model of distribution system blackout

停电事故作为影响配电网供电可靠性重要的因素之一,其预测的准确性将给整个电力系统的可靠性带来积极影响。本文提出了一种基于数据驱动的配电网停电预测模型,能够有效地预测停电事故的发生。该模型首先采用一种基于K-means聚类的停电数据集欠采样方法降低原始数据集的不平衡比;然后在此基础上,提出了一种改进的Adaboost集成学习算法,在每次权值更新时,通过使用已经训练的弱分类下的分类误差进行权重更新,用于对后面的弱分类器进行训练,进而改善分类性能。某地区的实际数据测试结果表明,本文提出的基于数据驱动的配电网停电预测模型能够有效地预测配电网停电事故的发生,相比于传统预测方法具有更好的精确度、召回率、F1值,停电预测性能得到明显提高。

Blackout is one of the most critical influence factors of distribution network reliability,whether to accurately predict in order to take measures,is an import way to improve the reliability of the power system.This paper presents a data-driven predictive model of power system blackout,which could predict the probability of blackout preciously.This paper firstly uses a sampling method based on K-means algorithm to solve the imbalance of the data set.In order to achieve outage prediction in distribution network,the improved integrated learning algorithm-Adaboost is proposed and the performance is improved by the weight update method with considering the error of the weak classifier.The test proves that the model based on the improved Adaboost algorithm has better accuracy,recall,and F1 value as compared with the Adaboost algorithm.The outage prediction performance is largely improved.