There are a large number of abnormal data in switch machines in industrial practice, which seriously affects the health status prediction of switch machines. The traditional Local Outlier Factor(LOF) algorithm based on local density is suitable for single cluster data environments. However, in multi-cluster environments, it tends to misclassify normal points at the cluster boundaries as outliers. In response to this deficiency, a study was conducted on railway switch machines, selecting 5 000 oil pressure data from start-up to connection for analysis, and a new density based influenced outlierness algorithm (INFLO) was proposed by introducing reverse -nearest neighbor to improve the LOF algorithm. A threshold was set based on the distribution of INFLO, and the number of outliers at the threshold was quantified and identified according to different values. The number of outliers identified was increased from 8.92% to 10.6% compared with traditional Quartile, KNN, and LOF algorithms. Finally, a Long Short-Term Memory (LSTM) network model was designed to input the data after removing outliers into the network. Comparative experiments were conducted with the above methods and the original data to verify the effectiveness of the proposed data cleaning method. The results show that the RMSE curve of the improved method is closer to that of the original data, indicating better prediction performance. The proposed new method can better eliminate outliers in the oil pressure dataset of the switch machine, and improve the performance of fault detection and data prediction, indicating that this method has a better effect on clearing outlier data than other methods.
YANGMao, MENGLingjian, LIDayong, et al. Identification of abnormal data of photovoltaic power based on class 3σ [J]. Renewable Energy Resources, 2018, 36(10): 1443-1448. (in Chinese)
SUNJianshu, LOUYuansheng, CHENYujun. Outlier detection of hydrological time series based on ARIMA-SVR model[J]. Computer & Digital Engineering, 2018, 46(2): 225-230. (in Chinese)
YANGXia, GOULiang, MAQian, et al. Outlier detection of electric power big data based on outlier detection algorithm[J]. Electronic Design Engineering, 2023, 31(24): 61-64. (in Chinese)
[11]
WANGB Y, LUOX Y, ZHANGS M. An improved outlier detection algorithm K-LOF based on density[J].Computing, Performance and Communication Systems, 2017, 2(1): 1-7.
HEXiaorui. Small signal stability assessment based on isolation forest with stochastic subspace identification[J]. Electrotechnical Application, 2023, 42(10): 16-20. (in Chinese)
WANGJingyi, CHENYinjia, YUANYe, et al. Efficient data cleaning framework for k-nearest neighbor learning models[J]. Journal of Frontiers of Computer Science and Technology, 2023, 17(9): 2241-2251. (in Chinese)
ZHANGZhongping, YAOChunchen, SUNGuangxu, et al. Outlier detection algorithm based on mapping distance ratio outlier factor[J]. Computer Integrated Manufacturing Systems, 2024, 30(5): 1719-1732. (in Chinese)
[21]
MORRISONR, LIUX, LINZ. Anomaly detection in wind turbine SCADA data for power curve cleaning[J]. Renewable Energy, 2022, 184: 473-486.
[22]
LINZ, LIUX, COLLUM. Wind power prediction based on high-frequency SCADA data along with isolation forest and deep learning neural networks[J]. International Journal of Electrical Power & Energy Systems, 2020, 118: 105835.
HUANGZaorong, WANGChunbao, WEIJianjun, et al. Abnormal gait detetction based on BlazePose and random forest[J]. Computing Technology and Automation, 2024, 43(2): 62-69. (in Chinese)
CHENHongming, YANGXidi, YULing, et al. Research on anomaly detection of turbine engine rotor unbalance based on optimized generative adversarial model[J]. Journal of Ordnance Equipment Engineering, 2024, 45(6): 104-114. (in Chinese)
WULei, KANGYingwei. Prediction model of SO2 concentration in desulfurization system based on improved particle swarm optimization LSTM[J]. Thermal Power Generation, 2021, 50(12): 66-73. (in Chinese)
LIUYunpeng, HUOXiaoli, LIUZhichao. Optical fiber network abnormal data detection algorithm based on deep learning[J]. Infrared and Laser Engineering, 2021, 50(6): 288-293. (in Chinese)
[33]
BREUNIGM M, KRIEGELH P, NGR T, et al. LOF: Identifying density-based local outliers[C]//the 2000 ACM SIGMOD International Conference on Management of Data, 2000: 93-104.
MENGNina, WANGZhengyang, GAOChenbo, et al. A vector building clustering algorithm based on local outlier factor[J]. Geomatics and Information Science of Wuhan University, 2024, 49 (4): 562-571. (in Chinese)
ZHANGXimei, XIEBin, XUTongtong, et al. Intrusion detection algorithm based on weighted Kmeans clustering with reverse K-nearest neighbor and density peak initialization[J]. Journal of Nanjing University of Science and Technology, 2023, 47 (1): 56-65. (in Chinese)
[38]
YAOQ T, ZHUH W, XIANGL, et al. A novel composed method of cleaning anomy data for improving state prediction of wind turbine[J]. Renewable Energy, 2023, 204: 131-140.
[39]
XIANGL, WANGP, YANGX, et al. Fault detection of wind turbine based on SCADA data analysis using CNN and LSTM with attention mechanism [J]. Measurement, 2021, 175:109094.