用于电力系统暂态稳定预测的支持向量机组合分类器及其可信度评价

发布时间：2018-09-01 09:37

【摘要】：目前,利用数据挖掘方法进行电力系统暂态稳定分析的研究,所用数据集普遍存在失稳样本少的样本不均衡问题,且挖掘模型的参数选择困难,缺乏对预测结果可信度进行评价。针对以上问题,文章提出用于暂稳预测的支持向量机(support vector machine,SVM)组合分类器及其可信度评价方法。首先采用改进bootstrap抽样得到多个类别均衡的数据集,利用随机特征子空间技术进一步压缩数据集;然后用压缩后的数据训练得到多个SVM分类器,各SVM的参数在经验范围内随机选取;最后,通过综合多个SVM的概率输出,得到组合分类器的预测结果,并对结果可信度进行评价。通过算例分析表明,改进Bootstrap算法能够明显减少对失稳样本的漏判,所提出的SVM组合分类器具有较高的预测准确度和可信度。
[Abstract]:At present, the data mining method is used to study the transient stability analysis of power system. The data sets used in this paper generally have the problem of sample unbalance with less unstable samples, and it is difficult to select the parameters of the mining model. There is a lack of evaluation of the reliability of the prediction results. In order to solve the above problems, this paper presents a support vector machine (support vector machine,SVM) combined classifier for transient prediction and its reliability evaluation method. First, the improved bootstrap sampling is used to obtain the multi-class equilibrium data sets, then the random feature subspace technique is used to further compress the data sets, and then the compressed data is trained to obtain multiple SVM classifiers. The parameters of each SVM are randomly selected in the range of experience. Finally, by synthesizing the probabilistic outputs of multiple SVM, the prediction results of the combined classifier are obtained, and the reliability of the results is evaluated. The example analysis shows that the improved Bootstrap algorithm can obviously reduce the missing judgment of the unstable samples, and the proposed SVM combined classifier has high prediction accuracy and reliability.
【作者单位】：北京交通大学电气工程学院;中国电力科学研究院;
【基金】：国家自然科学基金项目(NSFC)(51577009) 国家电网公司科技项目(XT71-15-001)~~
【分类号】：TM712

【相似文献】