一种基于 SVR 和 GRU 的新型电力监控防护系统-《计算机技术与发展》

文章信息/Info

Author(s):: SHEN Xiao-jie* ; LIAO Hua; LI Chuang; PAN Peng; LI Geng-da; Nanning Monitoring Center,EHV Transmission Company,China Southern Power Grid,Nanning 530025,China

Keywords:: power monitoring and protection system; safety identification index system; support vector regression; gated recurrent unit; recursive feature elimination

摘要:: 为满足电力监控防护系统精细化、实时化和智能化的复杂要求,设计了一种基于支持向量回归( SVR) 安全态势识别和门循环单元( GRU)预测策略的新型电力监控防护系统。基于支持向量机的递归特征消除( SVM-RFE) 技术和皮尔森相关系数( Pearson) 构建了安全识别指标体系。基于 SVR 技术,构建了基于 SVR 的安全态势识别模型。相较于 BPNN 模型,SVR 模型的安全态势识别结果在均方差误差 ( RMSE) 和平均绝对百分比误差 ( MAPE) 上分别降低了 43. 60% 和70．23% 。基于 GRU 神经网络,构建了基于 GRU 的安全态势预测模型。相较于 RBF 模型和 SVR 模型,GRU 预测模型的RMSE 分别降低了 19. 23% 和 23. 56% ,MAPE 降低了 48. 33% 和 58. 73% 。最后实现了电力监控防护系统,并通过实验验证了系统可行性。该研究为电力监控防护系统的安全运维提供重要参考,为构建智慧电网提供了技术支撑。

Abstract:: For the complex requirements of refinement,real-time and intelligent power monitoring and protection system,a novel powermonitoring and protection system based 　on Support Vector Regression ( SVR) security situation identification and Gated Recurrent Unit( GRU) prediction strategy was designed. Based on the Support Vector Machine - Recursive Feature Elimination ( SVM - RFE )technology and Pearson correlation coefficient,a security identification index system was constructed. Based 　on SVR technology,a SVR-based security situation recognition model was constructed. Compared with the BPNN model,the security situation recognition results ofthe SVR model were reduced by 43. 60% and 70. 23% in Root Mean Square Error ( RMSE) and Mean Absolute Percentage Error( MAPE) ,respectively. Based on the GRU neural network,a GRU-based security situation prediction model was constructed. Comparedwith the RBF model and the SVR model,the RMSE of the GRU prediction model was reduced by 19. 23% and 23. 56% ,and the MAPEwas reduced by 48. 33% and 58. 73% ,respectively. Finally,the power monitoring and protection system was realized,and the feasibilityof the system was verified by experiments. This research provides an important reference for the safe operation and maintenance of thepower monitoring and protection system,and provides technical support for the construction of a smart grid.