Current machine learning models for predicting geological conditions during earth pressure balance(EPB)shield tunneling predominantly rely on accurate geological conditions as model label inputs.This study introduces ...Current machine learning models for predicting geological conditions during earth pressure balance(EPB)shield tunneling predominantly rely on accurate geological conditions as model label inputs.This study introduces an innovative approach for the real-time prediction of geological conditions in EPB shield tunneling by utilizing an unsupervised incremental learning model that integrates deep temporal clustering(DTC)with elastic weight consolidation(EWC).The model was trained and tested using data from an EPB shield tunneling project in Nanjing,China.Results demonstrate that the DTC model outperforms nine comparison models by clustering the entire dataset into four distinct groups representing various geological conditions without requiring labeled data.Additionally,integrating EWC into the DTC model significantly enhances its continuous learning capabilities,enabling automatic parameter updates with incoming data and facilitating the real-time recognition of geological conditions.Feature importance was evaluated using the feature elimination method and the Shapley additive explanations(SHAP)method,underscoring the critical roles of earth chamber pressure and cutterhead rotation speed in predicting geological conditions.The proposed EWC-DTC model demonstrates practical utility for EPB shield tunneling in complex environments.展开更多
Photovoltaic(PV)systems are being increasingly implemented in the grid,and their intermittent output fluctuations threaten the stability of the grid,thereby requiring effective power ramp control(PRRC)strategies.In th...Photovoltaic(PV)systems are being increasingly implemented in the grid,and their intermittent output fluctuations threaten the stability of the grid,thereby requiring effective power ramp control(PRRC)strategies.In this study,we proposed a power fluctuation identification method to optimize the PRRC strategy.The K-means++cluster based on DTW used in this method,which clusters the historical PV power generation data into power curves corresponding to a specific weather type(sunny,cloudy,and rainy)in a time zone.Subsequently,wavelet decomposition is applied to discretize the power curves with extreme RR overrun to accurately identify the extreme fluctuation time zones.We conducted an analysis using minute-level data from a 100 kW PV plant in Arizona,which demonstrates that the proposed method can effectively identify high-risk periods.Weather patterns within the time zones were quantitatively identified using a weather probability model.A hardware-in-the-loop experimental platform was employed to validate two days of actual power data in Arizona,demonstrating the weather zoning accuracy of the method and the reasonableness of the control.The proposed methodology contributes significantly to PRRC strategy selection and parameter optimization(e.g.,ESS capacity storage allocation and APC power reserveΔP)in different time zones and weather conditions.展开更多
The original temporal clustering analysis (OTCA) is an effective technique for obtaining brain activation maps when the timing and location of the activation are completely unknown, but its deficiency of sensitivity i...The original temporal clustering analysis (OTCA) is an effective technique for obtaining brain activation maps when the timing and location of the activation are completely unknown, but its deficiency of sensitivity is exposed in processing brain activation signal which is relatively weak. The time slice analysis method based on OTCA is proposed considering the weakness of the functional magnetic resonance imaging (fMRI) signal of the rat model. By dividing the stimulation period into several time slices and analyzing each slice to detect the activated pixels respectively after the background removal, the sensitivity is significantly improved. The inhibitory response in the hypothalamus after glucose loading is detected successfully with this method in the experiment on rat. Combined with the OTCA method, the time slice analysis method based on OTCA is effective on detecting when, where and which type of response will happen after stimulation, even if the fMRI signal is weak.展开更多
Based on the observation by a Regional Air Quality Monitoring Network including 16 monitoring stations, temporal and spatial variations of ozone （O3）, NO2 and total oxidant （Ox） were analyzed by both linear regres...Based on the observation by a Regional Air Quality Monitoring Network including 16 monitoring stations, temporal and spatial variations of ozone （O3）, NO2 and total oxidant （Ox） were analyzed by both linear regression and cluster analysis. A fast increase of regional O3 concentrations of 0.86 ppbWyr was found for the annual averaged values from 2006 to 2011 in Guangdong, China. Such fast O3 increase is accompanied by a correspondingly fast NOx reduction as indicated by a fast NO2 reduction rate of 0,61 ppbV/yr. Based on a cluster analysis, the monitoring stations were classified into two major categories - rural stations （non-urban） and suburban/urban stations. The 03 concentrations at rural stations were relatively conserved while those at suburban/urban stations showed a fast increase rate of 2.0 ppbV/yr accompanied by a NO2 reduction rate of 1.2 ppbV/yr. Moreover, a rapid increase of the averaged O3 concentrations in springtime （13%/yr referred to 2006 level） was observed, which may result from the increase of solar duration, reduction of precipitation in Guangdong and transport from Eastern Central China. Application of smog production algorithm showed that the photochemical O3 production is mainly volatile organic compounds （VOC）-controlled. However, the photochemical O3 production is sensitive to both NOx and VOC for O3 pollution episode. Accordingly, it is expected that a combined NOx and VOC reduction will be helpful for the reduction of the O3 pollution episodes in Pearl River Delta while stringent VOC emission control is in general required for the regional O3 pollution control.展开更多
Simulated scenario-based test on Highly Automated Vehicles(HAVs)has been widely-received to ensure HAVs’safety as a solution to the problem brought on by mileage-based road testing.An appropriate scenario library is ...Simulated scenario-based test on Highly Automated Vehicles(HAVs)has been widely-received to ensure HAVs’safety as a solution to the problem brought on by mileage-based road testing.An appropriate scenario library is an important guarantee for the reliability of test results and test efficiency.Most scenario datasets opened to public contain redundant scenarios,making the required testing resource intractable.Testers need to design condensed scenario libraries based on existing datasets to expedite testing.The difficulty lies in how to measure the similarity of scenarios and the reliability of test results.In response to these problems,a framework for establishing and validation of the condensed scenario library based on double-C(i.e.,Conciseness and Consistence)principle is proposed.A Deep Temporal Clustering(DTC)algorithm,which reduces the dimensionality of scenario data using an autoencoder,is developed and combined with screening criteria to ensure the conciseness of the scenario library.In the case of the HighD dataset,all typical scenarios are found and the number of test-worthy scenarios is reduced by 59%.The OnSite Autonomous Driving Algorithm Challenge was hosted to verify the consistency of the scenario library in terms of test results,based on which 312 Planning and Control(PNC)algorithms were collected.A condensed scenario library is established based on the OnSite scenario library.There is no discernible change in the scenario library's difficulty level,with no significant difference in SUTs’scores or the score difference between SUTs.It is proved that the condensation method ensures conciseness and consistency of the testing scenario library.展开更多
基金financially supported by the National Natural Science Foundation of China(Grant Nos.52378392,52408356)Foal Eagle Program Youth Top-notch Talent Project of Fujian Province,China(Grant No.00387088).
文摘Current machine learning models for predicting geological conditions during earth pressure balance(EPB)shield tunneling predominantly rely on accurate geological conditions as model label inputs.This study introduces an innovative approach for the real-time prediction of geological conditions in EPB shield tunneling by utilizing an unsupervised incremental learning model that integrates deep temporal clustering(DTC)with elastic weight consolidation(EWC).The model was trained and tested using data from an EPB shield tunneling project in Nanjing,China.Results demonstrate that the DTC model outperforms nine comparison models by clustering the entire dataset into four distinct groups representing various geological conditions without requiring labeled data.Additionally,integrating EWC into the DTC model significantly enhances its continuous learning capabilities,enabling automatic parameter updates with incoming data and facilitating the real-time recognition of geological conditions.Feature importance was evaluated using the feature elimination method and the Shapley additive explanations(SHAP)method,underscoring the critical roles of earth chamber pressure and cutterhead rotation speed in predicting geological conditions.The proposed EWC-DTC model demonstrates practical utility for EPB shield tunneling in complex environments.
基金supported by the Natural Science Research Project of Jiangsu Higher Education Institutions(23KJB470019)the Natural Science Foundation of Jiangsu Province under Grant BK20240594.
文摘Photovoltaic(PV)systems are being increasingly implemented in the grid,and their intermittent output fluctuations threaten the stability of the grid,thereby requiring effective power ramp control(PRRC)strategies.In this study,we proposed a power fluctuation identification method to optimize the PRRC strategy.The K-means++cluster based on DTW used in this method,which clusters the historical PV power generation data into power curves corresponding to a specific weather type(sunny,cloudy,and rainy)in a time zone.Subsequently,wavelet decomposition is applied to discretize the power curves with extreme RR overrun to accurately identify the extreme fluctuation time zones.We conducted an analysis using minute-level data from a 100 kW PV plant in Arizona,which demonstrates that the proposed method can effectively identify high-risk periods.Weather patterns within the time zones were quantitatively identified using a weather probability model.A hardware-in-the-loop experimental platform was employed to validate two days of actual power data in Arizona,demonstrating the weather zoning accuracy of the method and the reasonableness of the control.The proposed methodology contributes significantly to PRRC strategy selection and parameter optimization(e.g.,ESS capacity storage allocation and APC power reserveΔP)in different time zones and weather conditions.
基金the National Natural Science Foundation of China (30370432)
文摘The original temporal clustering analysis (OTCA) is an effective technique for obtaining brain activation maps when the timing and location of the activation are completely unknown, but its deficiency of sensitivity is exposed in processing brain activation signal which is relatively weak. The time slice analysis method based on OTCA is proposed considering the weakness of the functional magnetic resonance imaging (fMRI) signal of the rat model. By dividing the stimulation period into several time slices and analyzing each slice to detect the activated pixels respectively after the background removal, the sensitivity is significantly improved. The inhibitory response in the hypothalamus after glucose loading is detected successfully with this method in the experiment on rat. Combined with the OTCA method, the time slice analysis method based on OTCA is effective on detecting when, where and which type of response will happen after stimulation, even if the fMRI signal is weak.
基金supported by the National Natural Science Foundation of China(No.21190052,41121004)the Strategic Priority Research Program of the Chinese Academy of Sciences(No.XDB05010500)+1 种基金the National Public Benefit Special Fund for Environmental Protection Research(No.201009001-4)the Special Fund of State Key Joint Laboratory of Environment Simulation and Pollution Control(No.13Z02ESPCP)
文摘Based on the observation by a Regional Air Quality Monitoring Network including 16 monitoring stations, temporal and spatial variations of ozone （O3）, NO2 and total oxidant （Ox） were analyzed by both linear regression and cluster analysis. A fast increase of regional O3 concentrations of 0.86 ppbWyr was found for the annual averaged values from 2006 to 2011 in Guangdong, China. Such fast O3 increase is accompanied by a correspondingly fast NOx reduction as indicated by a fast NO2 reduction rate of 0,61 ppbV/yr. Based on a cluster analysis, the monitoring stations were classified into two major categories - rural stations （non-urban） and suburban/urban stations. The 03 concentrations at rural stations were relatively conserved while those at suburban/urban stations showed a fast increase rate of 2.0 ppbV/yr accompanied by a NO2 reduction rate of 1.2 ppbV/yr. Moreover, a rapid increase of the averaged O3 concentrations in springtime （13%/yr referred to 2006 level） was observed, which may result from the increase of solar duration, reduction of precipitation in Guangdong and transport from Eastern Central China. Application of smog production algorithm showed that the photochemical O3 production is mainly volatile organic compounds （VOC）-controlled. However, the photochemical O3 production is sensitive to both NOx and VOC for O3 pollution episode. Accordingly, it is expected that a combined NOx and VOC reduction will be helpful for the reduction of the O3 pollution episodes in Pearl River Delta while stringent VOC emission control is in general required for the regional O3 pollution control.
文摘Simulated scenario-based test on Highly Automated Vehicles(HAVs)has been widely-received to ensure HAVs’safety as a solution to the problem brought on by mileage-based road testing.An appropriate scenario library is an important guarantee for the reliability of test results and test efficiency.Most scenario datasets opened to public contain redundant scenarios,making the required testing resource intractable.Testers need to design condensed scenario libraries based on existing datasets to expedite testing.The difficulty lies in how to measure the similarity of scenarios and the reliability of test results.In response to these problems,a framework for establishing and validation of the condensed scenario library based on double-C(i.e.,Conciseness and Consistence)principle is proposed.A Deep Temporal Clustering(DTC)algorithm,which reduces the dimensionality of scenario data using an autoencoder,is developed and combined with screening criteria to ensure the conciseness of the scenario library.In the case of the HighD dataset,all typical scenarios are found and the number of test-worthy scenarios is reduced by 59%.The OnSite Autonomous Driving Algorithm Challenge was hosted to verify the consistency of the scenario library in terms of test results,based on which 312 Planning and Control(PNC)algorithms were collected.A condensed scenario library is established based on the OnSite scenario library.There is no discernible change in the scenario library's difficulty level,with no significant difference in SUTs’scores or the score difference between SUTs.It is proved that the condensation method ensures conciseness and consistency of the testing scenario library.
