基于改进SOM聚类模型的配电网短期负荷用电行为分析

Analysis of Short-term Load Behavior of Distribution Network Based on SOM Clustering Model

建立贝叶斯正则化的SOM聚类模型对配电网短期负荷用电行为分析,并利用UCI数据集对其有效性进行了验证,对电网运行过程的实际负荷进行了测试。从宁夏某电网采集负荷样本进行验证,测试结果表明,权重值超过0.6,编号10对应的峰时耗电率对于不同日负荷走势相似性具有显著影响。根据这5项负荷特性指标便可以获得具有较高准确率的一日负荷曲线变化曲线并精确反映出特定点的信息。以SOM神经网络总共进行了5次聚类,虽然每次聚类激活神经元存在差异,但都采用相同的分类过程,最终形成了平均准确率测试结果。

In this paper,the SOM clustering model of Bayesian regularization is established to analyze the short-term load behavior of distribution network.After that,its effectiveness is verified by using UCI data set,and the actual load of power network operation process is tested.Load samples are collected and verified from a power grid in Ningxia Province.The test results showed that the weight value exceeds 0.6,and the peaking power consumption rate corresponding to No.10 has a significant impact on the similarity of load trends in different days.According to these five load characteristic indexes,the change curve of daily load curve with high accuracy can be obtained and it accurately reflects the information of specific points.SOM neural network is used for clustering for a total of 5 times.Although there are differences in the activation of neurons in each cluster,the same classification process is used to form the average accuracy test results.