基于随机森林算法的配电网停电研判方案设计

Research and Judgment Scheme Design of Distribution Network Blackout Based on Random Forest Algorithm

当前配电网停电研判通常仅对电网中的单一节点故障时的故障数据进行提取,缺乏对整体配电网数据的实时监测,研判方案数据查全率及准确度较低。为此提出基于随机森林算法的配电网停电研判方案。使用随机森林算法处理配电网所有数据,设计随机森林电网样本训练方法;建立配电网分层模型,代入经过随机森林处理的数据,得出影响停电的相应数据。采用PMS2.0对获得的数据进行处理,监测停电起始时间和终止时间,实现对配电网数据的研判。为了验证设计的研判方案的性能,使用某地的配电网络数据作为研判基础,设计对比实验。结果证明设计的研判方案在数据的查全率和查准率较高,性能更优。

Currently, distribution network outage research and judgment usually only extracts fault data when a single node in the power grid fails, and there is a lack of real-time monitoring of the overall distribution network data. The research and judgment plan data recall rate and accuracy are low. For this reason, a scheme design for power outage research and judgment of distribution network based on random forest algorithm is proposed to process all the data of the distribution network, and the random forest power grid sample training method is designed. The hierarchical topology model of the distribution network is established, and the data processed by the random forest are brought in to obtain the corresponding data affecting the power outage. PMS2.0 is used to process the obtained data, monitor the start time and end time of the power outage, and realize the research and judgment of the distribution network data. In order to verify the performance of the designed research and judgment scheme, we use the distribution network data of a certain place as the basis of research and judgment, and design a comparison experiment. The results prove that the designed research and judgment scheme has higher recall and accuracy of data, better performance.