电网技术2017，Vol.41Issue(8)：2737-2742,6.DOI:10.13335/j.1000-3673.pst.2016.2623
基于深度信念网络的变压器油中溶解气体浓度预测方法
Concentration Prediction of Dissolved Gases in Transformer Oil Based on Deep Belief Networks
摘要
Abstract
Prediction of development trend of gas concentration dissolved in transformer oil can provide important basis for transformer condition assessment. A new prediction model based on deep belief networks is proposed. Seven types of characteristic gas concentration combined with environment temperature and transformer oil temperature are fed to input layer. The model can automatically extract regulation of gas concentration development trend through training a multi-hidden-layer machine learning model based on restricted Boltzmann machine. Correlation between different types of gases and influence of temperatures is activated layer by layer. Irrelevant and redundant information is inhibited by the model. The proposed method has higher prediction accuracy. It overcomes drawbacks of low stability in traditional methods and shortcoming of considering only one characteristic gas. In addition, it avoids manual intervention in calculation process. Finally, case analysis verifies effectiveness and superiority of the proposed model.关键词
变压器/油中溶解气体/深度信念网络/相关性/预测Key words
transformer/dissolved gas in oil/deep belief networks/correlation/predict分类
信息技术与安全科学引用本文复制引用
代杰杰,宋辉,杨祎,陈玉峰,盛戈皞,江秀臣..基于深度信念网络的变压器油中溶解气体浓度预测方法[J].电网技术,2017,41(8):2737-2742,6.基金项目
国家自然科学基金项目(51477100) （51477100）
国家863 高技术基金项目(2015AA050204) （2015AA050204）
国家电网公司科技项目(520626150032).Project Supported by National Natural Science Foundation of China (NSFC) (51477100) （520626150032）
Project Supported by the National High Technology Research and Development of China (863 Programme) (2015AA050204) （863 Programme）
Project Supported by Science and Technology Foundation of State Grid Corporation of China (520626150032). （520626150032）
