The rapid integration of Internet of Things(IoT)technologies is reshaping the global energy landscape by deploying smart meters that enable high-resolution consumption monitoring,two-way communication,and advanced met...The rapid integration of Internet of Things(IoT)technologies is reshaping the global energy landscape by deploying smart meters that enable high-resolution consumption monitoring,two-way communication,and advanced metering infrastructure services.However,this digital transformation also exposes power system to evolving threats,ranging from cyber intrusions and electricity theft to device malfunctions,and the unpredictable nature of these anomalies,coupled with the scarcity of labeled fault data,makes realtime detection exceptionally challenging.To address these difficulties,a real-time decision support framework is presented for smart meter anomality detection that leverages rolling time windows and two self-supervised contrastive learning modules.The first module synthesizes diverse negative samples to overcome the lack of labeled anomalies,while the second captures intrinsic temporal patterns for enhanced contextual discrimination.The end-to-end framework continuously updates its model with rolling updated meter data to deliver timely identification of emerging abnormal behaviors in evolving grids.Extensive evaluations on eight publicly available smart meter datasets over seven diverse abnormal patterns testing demonstrate the effectiveness of the proposed full framework,achieving average recall and F1 score of more than 0.85.展开更多
Red tide is an ecological disaster caused by the excessive proliferation of photosynthetic algae in the ocean.The frequent occurrences of red tide have brought serious harms to the marine aquaculture and caused signif...Red tide is an ecological disaster caused by the excessive proliferation of photosynthetic algae in the ocean.The frequent occurrences of red tide have brought serious harms to the marine aquaculture and caused significant economic losses to the marine industry.Red tide prediction can alleviate and even stop the long-term damages to marine ecosystems,which helps maintain the ecological balance of the ocean environment and contributes to the Sustainable Development Goal of“life below water”formulated by the United Nations.Aiming at red tide prediction using remote sensing technology,this study proposed a novel approach of red tide prediction using time-series hyperspectral observations,and examined the proposed method in the Xinghai Bay,China.Three spectral indices,namely the twoband ratio(TBR),the three-band spectral index(TBSI),and the fluorescence baseline height(FLH),were used to reduce the dimensionality of hyperspectral data and extract spectral features.Two machine learning models including the random forest(RF)and the support vector machine(SVM)were employed to predict whether red tide would occur on a target day based on the time-series spectral indices obtained in the previous days.By comparing and analyzing the prediction results of multiple machine learning models trained with different spectral indices and temporal lengths,it is found that both the RF and the SVM models can predict the red tide outbreaks at the accuracies over 0.9 using adequate temporal lengths of input data.When the temporal length of input data is limited,however,it is suggested to use the RF model,which accurately predicts red tide outbreaks using the temporal input of the 2-d TBSI.The proposed method is expected to provide oceanic and maritime agencies with early warnings on red tide outbreaks and ensure the safety of the coastal environment in large spatial scales using optical remote sensing technology.展开更多
It is crucial to predict future mechanical behaviors for the prevention of structural disasters.Especially for underground construction,the structural mechanical behaviors are affected by multiple internal and externa...It is crucial to predict future mechanical behaviors for the prevention of structural disasters.Especially for underground construction,the structural mechanical behaviors are affected by multiple internal and external factors due to the complex conditions.Given that the existing models fail to take into account all the factors and accurate prediction of the multiple time series simultaneously is difficult using these models,this study proposed an improved prediction model through the autoencoder fused long-and short-term time-series network driven by the mass number of monitoring data.Then,the proposed model was formalized on multiple time series of strain monitoring data.Also,the discussion analysis with a classical baseline and an ablation experiment was conducted to verify the effectiveness of the prediction model.As the results indicate,the proposed model shows obvious superiority in predicting the future mechanical behaviors of structures.As a case study,the presented model was applied to the Nanjing Dinghuaimen tunnel to predict the stain variation on a different time scale in the future.展开更多
Recently,the outbreak and spread of larch caterpillar(Dendrolimus superans)pests have emerged as significant contributors to forest degradation in the Changbai Mountains,China.Understanding the spatiotemporal distribu...Recently,the outbreak and spread of larch caterpillar(Dendrolimus superans)pests have emerged as significant contributors to forest degradation in the Changbai Mountains,China.Understanding the spatiotemporal distribution patterns of these pests is crucial for effective management and protection of forest ecosystems.This study proposes a pest monitoring approach based on Sentinel imagery.Through time-series analysis,we extracted pest-sensitive features and developed a random forest classifier that integrated Sentinel-1,Sentinel-2,and field sampling data from 2019–2023 to monitor larch caterpillar pests in the Changbai Mountains National Nature Reserve(CMNNR),Northeast China.Our findings indicated that bands green(B3),near-infrared(B8),short wave infrared(B11 and B12)from Sentinel-2 remote sensing images exhibited notable discriminative capabilities for identifying larch caterpillar pests.Specifically,the Normalized Difference Vegetation Index(NDVI)at the end of the growing season emerged as the most valuable feature for pest extraction.Incorporating Synthetic Aperture Radar(SAR)features along with optical data marginally enhances model performance.Furthermore,our approach unveiled the outbreak of larch caterpillar pests,achieving classification map with overall accuracy exceeding 85%and Kappa coefficient surpassing 0.8 for five study years.The pest outbreak began in 2019 and progressively intensified over time.In September 2019,the affected area spanned 114.23 km^(2).The infested area exhibited a declining trend from 2020 to 2023.This study introduces a novel method for the high-precision identification of larch caterpillar pests,offering technical advancements and theoretical underpinnings to support forest management strategies.展开更多
Fault diagnosis is important for maintaining the safety and effectiveness of chemical process.Considering the multivariate,nonlinear,and dynamic characteristic of chemical process,many time-series-based data-driven fa...Fault diagnosis is important for maintaining the safety and effectiveness of chemical process.Considering the multivariate,nonlinear,and dynamic characteristic of chemical process,many time-series-based data-driven fault diagnosis methods have been developed in recent years.However,the existing methods have the problem of long-term dependency and are difficult to train due to the sequential way of training.To overcome these problems,a novel fault diagnosis method based on time-series and the hierarchical multihead self-attention(HMSAN)is proposed for chemical process.First,a sliding window strategy is adopted to construct the normalized time-series dataset.Second,the HMSAN is developed to extract the time-relevant features from the time-series process data.It improves the basic self-attention model in both width and depth.With the multihead structure,the HMSAN can pay attention to different aspects of the complicated chemical process and obtain the global dynamic features.However,the multiple heads in parallel lead to redundant information,which cannot improve the diagnosis performance.With the hierarchical structure,the redundant information is reduced and the deep local time-related features are further extracted.Besides,a novel many-to-one training strategy is introduced for HMSAN to simplify the training procedure and capture the long-term dependency.Finally,the effectiveness of the proposed method is demonstrated by two chemical cases.The experimental results show that the proposed method achieves a great performance on time-series industrial data and outperforms the state-of-the-art approaches.展开更多
The frequent missing values in radar-derived time-series tracks of aerial targets(RTT-AT)lead to significant challenges in subsequent data-driven tasks.However,the majority of imputation research focuses on random mis...The frequent missing values in radar-derived time-series tracks of aerial targets(RTT-AT)lead to significant challenges in subsequent data-driven tasks.However,the majority of imputation research focuses on random missing(RM)that differs significantly from common missing patterns of RTT-AT.The method for solving the RM may experience performance degradation or failure when applied to RTT-AT imputation.Conventional autoregressive deep learning methods are prone to error accumulation and long-term dependency loss.In this paper,a non-autoregressive imputation model that addresses the issue of missing value imputation for two common missing patterns in RTT-AT is proposed.Our model consists of two probabilistic sparse diagonal masking self-attention(PSDMSA)units and a weight fusion unit.It learns missing values by combining the representations outputted by the two units,aiming to minimize the difference between the missing values and their actual values.The PSDMSA units effectively capture temporal dependencies and attribute correlations between time steps,improving imputation quality.The weight fusion unit automatically updates the weights of the output representations from the two units to obtain a more accurate final representation.The experimental results indicate that,despite varying missing rates in the two missing patterns,our model consistently outperforms other methods in imputation performance and exhibits a low frequency of deviations in estimates for specific missing entries.Compared to the state-of-the-art autoregressive deep learning imputation model Bidirectional Recurrent Imputation for Time Series(BRITS),our proposed model reduces mean absolute error(MAE)by 31%~50%.Additionally,the model attains a training speed that is 4 to 8 times faster when compared to both BRITS and a standard Transformer model when trained on the same dataset.Finally,the findings from the ablation experiments demonstrate that the PSDMSA,the weight fusion unit,cascade network design,and imputation loss enhance imputation performance and confirm the efficacy of our design.展开更多
Accurate mapping and timely monitoring of urban redevelopment are pivotal for urban studies and decisionmakers to foster sustainable urban development.Traditional mapping methods heavily depend on field surveys and su...Accurate mapping and timely monitoring of urban redevelopment are pivotal for urban studies and decisionmakers to foster sustainable urban development.Traditional mapping methods heavily depend on field surveys and subjective questionnaires,yielding less objective,reliable,and timely data.Recent advancements in Geographic Information Systems(GIS)and remote-sensing technologies have improved the identification and mapping of urban redevelopment through quantitative analysis using satellite-based observations.Nonetheless,challenges persist,particularly concerning accuracy and significant temporal delays.This study introduces a novel approach to modeling urban redevelopment,leveraging machine learning algorithms and remote-sensing data.This methodology can facilitate the accurate and timely identification of urban redevelopment activities.The study’s machine learning model can analyze time-series remote-sensing data to identify spatio-temporal and spectral patterns related to urban redevelopment.The model is thoroughly evaluated,and the results indicate that it can accurately capture the time-series patterns of urban redevelopment.This research’s findings are useful for evaluating urban demographic and economic changes,informing policymaking and urban planning,and contributing to sustainable urban development.The model can also serve as a foundation for future research on early-stage urban redevelopment detection and evaluation of the causes and impacts of urban redevelopment.展开更多
Stock price prediction is a typical complex time series prediction problem characterized by dynamics,nonlinearity,and complexity.This paper introduces a generative adversarial network model that incorporates an attent...Stock price prediction is a typical complex time series prediction problem characterized by dynamics,nonlinearity,and complexity.This paper introduces a generative adversarial network model that incorporates an attention mechanism(GAN-LSTM-Attention)to improve the accuracy of stock price prediction.Firstly,the generator of this model combines the Long and Short-Term Memory Network(LSTM),the Attention Mechanism and,the Fully-Connected Layer,focusing on generating the predicted stock price.The discriminator combines the Convolutional Neural Network(CNN)and the Fully-Connected Layer to discriminate between real stock prices and generated stock prices.Secondly,to evaluate the practical application ability and generalization ability of the GAN-LSTM-Attention model,four representative stocks in the United States of America(USA)stock market,namely,Standard&Poor’s 500 Index stock,Apple Incorporatedstock,AdvancedMicroDevices Incorporatedstock,and Google Incorporated stock were selected for prediction experiments,and the prediction performance was comprehensively evaluated by using the three evaluation metrics,namely,mean absolute error(MAE),root mean square error(RMSE),and coefficient of determination(R2).Finally,the specific effects of the attention mechanism,convolutional layer,and fully-connected layer on the prediction performance of the model are systematically analyzed through ablation study.The results of experiment show that the GAN-LSTM-Attention model exhibits excellent performance and robustness in stock price prediction.展开更多
When interpreting results,it is imperative to have some understanding of the degree to which the results are replicable.If the results cannot be replicated with independent data,then interpretations from the results b...When interpreting results,it is imperative to have some understanding of the degree to which the results are replicable.If the results cannot be replicated with independent data,then interpretations from the results become questionable.To minimize the potential for misinterpretations,the current study analyzes six time-series derived from globally sampled U-Pb zircon databases–of which,two are independent igneous databases,one being a quasi-independent igneous database,and three being independent detrital databases.These time-series are then analyzed with standard statistical methods to evaluate replicability.The methods include bandpass filtering to transform the raw time-series into stationary sequences,Student’s t-test,Monte Carlo simulations,periodograms from spectral analysis,correlation studies,and correlograms.Each test is designed to determine the replicability of a specific time-series,as well as the replicability of periodicities found from the time-series.The results show at least three key components to assessing replicability:(a)U-Pb igneous and detrital zircon age distributions are highly replicable,(b)time-series replicability gradually deteriorates with age,and(c)replicability is scale dependent,with low frequency cycles being more replicable than high frequency cycles.From the tests,we conclude that four harmonic cycles are highly replicable and statistically significant,these being periodicities of 810,270,90,and 67.5-myr.展开更多
Preterm birth remains a leading cause of neonatal complications and highlights the need for early and accurate prediction techniques to improve both fetal and maternal health outcomes.This study introduces a hybrid ap...Preterm birth remains a leading cause of neonatal complications and highlights the need for early and accurate prediction techniques to improve both fetal and maternal health outcomes.This study introduces a hybrid approach integrating Long Short-Term Memory(LSTM)networks with the Hybrid Greylag Goose and Particle Swarm Optimization(GGPSO)algorithm to optimize preterm birth classification using Electrohysterogram signals.The dataset consists of 58 samples of 1000-second-long Electrohysterogram recordings,capturing key physiological features such as contraction patterns,entropy,and statistical variations.Statistical analysis and feature selection methods are applied to identify the most relevant predictors and enhance model interpretability.LSTM networks effectively capture temporal patterns in uterine activity,while the GGPSO algorithm finetunes hyperparameters,mitigating overfitting and improving classification accuracy.The proposed GGPSO-optimized LSTM model achieved superior performance with 97.34%accuracy,96.91%sensitivity,97.74%specificity,and 97.23%F-score,significantly outperforming traditional machine learning approaches and demonstrating the effectiveness of hybrid metaheuristic optimization in enhancing deep learning models for clinical applications.By combining deep learning withmetaheuristic optimization,this study contributes to advancing intelligent auto-diagnosis systems,facilitating early detection of pretermbirth risks and timely medical interventions.展开更多
Studying the inner structure of intraplate earthquakes originating in aseismic areas,which are poorly covered by seismic networks or as historical earthquakes is usually the only way to get knowledge about their sourc...Studying the inner structure of intraplate earthquakes originating in aseismic areas,which are poorly covered by seismic networks or as historical earthquakes is usually the only way to get knowledge about their source mechanisms,which is partially essential for a deeper understanding of intraplate geodynamics.The epicentral zones of earthquakes are situated in hard-toreach areas,so,using active seismic methods for such purposes is unreasonable or even impossible because of high cost and logistical difficulties.We propose a novel approach that combines diverse passive seismic methods,which allows us to get sufficient information about geological environment structure for such task solutions.As an example,we investigated the inner structure of platform earthquake epicentral zone originated up north of Russia.We used four passive seismic methods:microseismic sounding method,passive seismic interferometry,HVSR method,and microseismic activity method.We show that passive seismic data,recoded in the same installation and processed by these different methods,can provide sufficient information about structure of studied environment,needed to explain source mechanism.In sum,the hypocenter zone is presented by intersection of vertical faults and a lateral fractured zone in the middle crust.Results were confirmed by comparison with results by active seismic methods.展开更多
The spatiotemporal characterization of plume sedimentation and microorganisms is critical for developing plume ecological monitoring model.To address the limitations of traditional methods in obtaining high-quality se...The spatiotemporal characterization of plume sedimentation and microorganisms is critical for developing plume ecological monitoring model.To address the limitations of traditional methods in obtaining high-quality sediment,a novel sampling system with 6000 m operational capability and three-month endurance was developed.It is equipped with three sediment samplers,a set of formaldehyde preservation solution injection devices.The system is controlled by a low-power,timing-triggered controllers.To investigate low-disturbance rheological mechanisms,gap controlled rheological tests were conducted to optimize the structural design of the sampling and sealing assembly.Stress-controlled shear rheological tests were employed to investigate the mechanisms governing yield stress in sediments under varying temperature conditions and boundary roughness.Additionally,the coupled Eulerian-Lagrangian(CEL)method and sediment rheological constitutive models were employed to simulate tube-soil interaction dynamics and sediment disturbance.The radial heterogeneity of sediment disturbance and friction variation of the sampling tube were revealed.The tube was completely“plugged”at a penetration depth of 261 mm,providing critical data support to the penetration depth parameters.The deep-sea pressure test and South China Sea field trials demonstrated the system’s capability to collect and preserve quantitative time-series sediment samples with high fidelity.展开更多
Since April 2002,the Gravity Recovery and Climate Experiment Satellite(GRACE)has provided monthly total water storage anomalies(TWSAs)on a global scale.However,these TWSAs are discontinuous because some GRACE observat...Since April 2002,the Gravity Recovery and Climate Experiment Satellite(GRACE)has provided monthly total water storage anomalies(TWSAs)on a global scale.However,these TWSAs are discontinuous because some GRACE observation data are missing.This study presents a combined machine learning-based modeling algorithm without hydrological model data.The TWSA time-series data for 11 large regions worldwide were divided into training and test sets.Autoregressive integrated moving average(ARIMA),long short-term memory(LSTM),and an ARIMA-LSTM combined model were used.The model predictions were compared with GRACE observations,and the model accuracy was evaluated using fi ve metrics:the Nash-Sutcliff e effi ciency coeffi cient(NSE),Pearson correlation coeffi cient(CC),root mean square error(RMSE),normalized RMSE(NRMSE),and mean absolute percentage error.The results show that at the basin scale,the mean CC,NSE,and NRMSE for the ARIMA-LSTM model were 0.93,0.83,and 0.12,respectively.At the grid scale,this study compared the spatial distribution and cumulative distribution function curves of the metrics in the Amazon and Volga River basins.The ARIMA-LSTM model had mean CC and NSE values of 0.89 and 0.61 and 0.92 and 0.61 in the Amazon and Volga River basins,respectively,which are superior to those of the ARIMA model(0.86 and 0.48 and 0.88 and 0.46,respectively)and the LSTM model(0.80 and 0.41 and 0.89 and 0.31,respectively).In the ARIMA-LSTM model,the proportions of grid cells with NSE>0.50 for the two basins were 63.3%and 80.8%,while they were 54.3%and 51.3%in the ARIMA model and 53.7%and 43.2%in the LSTM model.The ARIMA-LSTM model significantly improved the NSE values of the predictions while guaranteeing high CC values in the GRACE data reconstruction at both scales,which can aid in fi lling in discontinuous data in temporal gravity fi eld models..展开更多
Studies to enhance the management of electrical energy have gained considerable momentum in recent years. The question of how much energy will be needed in households is a pressing issue as it allows the management pl...Studies to enhance the management of electrical energy have gained considerable momentum in recent years. The question of how much energy will be needed in households is a pressing issue as it allows the management plan of the available resources at the power grids and consumer levels. A non-intrusive inference process can be adopted to predict the amount of energy required by appliances. In this study, an inference process of appliance consumption based on temporal and environmental factors used as a soft sensor is proposed. First, a study of the correlation between the electrical and environmental variables is presented. Then, a resampling process is applied to the initial data set to generate three other subsets of data. All the subsets were evaluated to deduce the adequate granularity for the prediction of the energy demand. Then, a cloud-assisted deep neural network model is designed to forecast short-term energy consumption in a residential area while preserving user privacy. The solution is applied to the consumption data of four appliances elected from a set of real household power data. The experiment results show that the proposed framework is effective for estimating consumption with convincing accuracy.展开更多
Accurate energy demand forecasting is crucial in today’s rapidly electrifying world with decentralized systems and integrated renewables.Traditional models struggle with the dynamic complexities,but AI(artificial int...Accurate energy demand forecasting is crucial in today’s rapidly electrifying world with decentralized systems and integrated renewables.Traditional models struggle with the dynamic complexities,but AI(artificial intelligence),particularly ML(machine learning)and DL(deep learning),offers transformative solutions.This article explores how AI enhances forecasting accuracy,enables real-time adaptability,and supports strategic energy management.It examines the synergy between AI,IoT(Internet of Things)devices,and smart grids in generating predictive and prescriptive insights.Through case studies,we analyze the benefits and challenges of deploying AI in this domain,including data quality,model explainability,and infrastructure needs.Ultimately,AI emerges as a key enabler for the resilient,data-driven energy systems required to meet modern society’s evolving demands and achieve a sustainable future.展开更多
Accurate prediction of nurse demand plays a crucial role in efficiently planning the healthcare workforce,ensuring appropriate staffing levels,and providing high-quality care to patients.The intricacy and variety of c...Accurate prediction of nurse demand plays a crucial role in efficiently planning the healthcare workforce,ensuring appropriate staffing levels,and providing high-quality care to patients.The intricacy and variety of contemporary healthcare systems and a growing patient populace call for advanced forecasting models.Factors like technological advancements,novel treatment protocols,and the increasing prevalence of chronic illnesses have diminished the efficacy of traditional estimation approaches.Novel forecasting methodologies,including time-series analysis,machine learning,and simulation-based techniques,have been developed to tackle these challenges.Time-series analysis recognizes patterns from past data,whereas machine learning uses extensive datasets to uncover concealed trends.Simulation models are employed to assess diverse scenarios,assisting in proactive adjustments to staffing.These techniques offer distinct advantages,such as the identification of seasonal patterns,the management of large datasets,and the ability to test various assumptions.By integrating these sophisticated models into workforce planning,organizations can optimize staffing,reduce financial waste,and elevate the standard of patient care.As the healthcare field progresses,the utilization of these predictive models will be pivotal for fostering adaptable and resilient workforce management.展开更多
Model-free,data-driven prediction of chaotic motions is a long-standing challenge in nonlinear science.Stimulated by the recent progress in machine learning,considerable attention has been given to the inference of ch...Model-free,data-driven prediction of chaotic motions is a long-standing challenge in nonlinear science.Stimulated by the recent progress in machine learning,considerable attention has been given to the inference of chaos by the technique of reservoir computing(RC).In particular,by incorporating a parameter-control channel into the standard RC,it is demonstrated that the machine is able to not only replicate the dynamics of the training states,but also infer new dynamics not included in the training set.The new machine-learning scheme,termed parameter-aware RC,opens up new avenues for data-based analysis of chaotic systems,and holds promise for predicting and controlling many real-world complex systems.Here,using typical chaotic systems as examples,we give a comprehensive introduction to this powerful machine-learning technique,including the algorithm,the implementation,the performance,and the open questions calling for further studies.展开更多
The reverse design of solid rocket motor(SRM)propellant grain involves determining the grain geometry to closely match a predefined internal ballistic curve.While existing reverse design methods are feasible,they ofte...The reverse design of solid rocket motor(SRM)propellant grain involves determining the grain geometry to closely match a predefined internal ballistic curve.While existing reverse design methods are feasible,they often face challenges such as lengthy computation times and limited accuracy.To achieve rapid and accurate matching between the targeted ballistic curve and complex grain shape,this paper proposes a novel reverse design method for SRM propellant grain based on time-series data imaging and convolutional neural network(CNN).First,a finocyl grain shape-internal ballistic curve dataset is created using parametric modeling techniques to comprehensively cover the design space.Next,the internal ballistic time-series data is encoded into three-channel images,establishing a potential relationship between the ballistic curves and their image representations.A CNN is then constructed and trained using these encoded images.Once trained,the model enables efficient inference of propellant grain dimensions from a target internal ballistic curve.This paper conducts comparative experiments across various neural network models,validating the effectiveness of the feature extraction method that transforms internal ballistic time-series data into images,as well as its generalization capability across different CNN architectures.Ignition tests were performed based on the predicted propellant grain.The results demonstrate that the relative error between the experimental internal ballistic curves and the target curves is less than 5%,confirming the validity and feasibility of the proposed reverse design methodology.展开更多
文摘The rapid integration of Internet of Things(IoT)technologies is reshaping the global energy landscape by deploying smart meters that enable high-resolution consumption monitoring,two-way communication,and advanced metering infrastructure services.However,this digital transformation also exposes power system to evolving threats,ranging from cyber intrusions and electricity theft to device malfunctions,and the unpredictable nature of these anomalies,coupled with the scarcity of labeled fault data,makes realtime detection exceptionally challenging.To address these difficulties,a real-time decision support framework is presented for smart meter anomality detection that leverages rolling time windows and two self-supervised contrastive learning modules.The first module synthesizes diverse negative samples to overcome the lack of labeled anomalies,while the second captures intrinsic temporal patterns for enhanced contextual discrimination.The end-to-end framework continuously updates its model with rolling updated meter data to deliver timely identification of emerging abnormal behaviors in evolving grids.Extensive evaluations on eight publicly available smart meter datasets over seven diverse abnormal patterns testing demonstrate the effectiveness of the proposed full framework,achieving average recall and F1 score of more than 0.85.
基金The National Natural Science Foundation of China under contract No.42406188the Natural Science Foundation of Liaoning Province under contract No.2024-BS-022+1 种基金the Dalian High-Level Talent Innovation Program under contract No.2022RG02the Fundamental Research Funds for the Central Universities under contract No.3132025107.
文摘Red tide is an ecological disaster caused by the excessive proliferation of photosynthetic algae in the ocean.The frequent occurrences of red tide have brought serious harms to the marine aquaculture and caused significant economic losses to the marine industry.Red tide prediction can alleviate and even stop the long-term damages to marine ecosystems,which helps maintain the ecological balance of the ocean environment and contributes to the Sustainable Development Goal of“life below water”formulated by the United Nations.Aiming at red tide prediction using remote sensing technology,this study proposed a novel approach of red tide prediction using time-series hyperspectral observations,and examined the proposed method in the Xinghai Bay,China.Three spectral indices,namely the twoband ratio(TBR),the three-band spectral index(TBSI),and the fluorescence baseline height(FLH),were used to reduce the dimensionality of hyperspectral data and extract spectral features.Two machine learning models including the random forest(RF)and the support vector machine(SVM)were employed to predict whether red tide would occur on a target day based on the time-series spectral indices obtained in the previous days.By comparing and analyzing the prediction results of multiple machine learning models trained with different spectral indices and temporal lengths,it is found that both the RF and the SVM models can predict the red tide outbreaks at the accuracies over 0.9 using adequate temporal lengths of input data.When the temporal length of input data is limited,however,it is suggested to use the RF model,which accurately predicts red tide outbreaks using the temporal input of the 2-d TBSI.The proposed method is expected to provide oceanic and maritime agencies with early warnings on red tide outbreaks and ensure the safety of the coastal environment in large spatial scales using optical remote sensing technology.
基金National Key Research and Development Program of China,Grant/Award Number:2018YFB2101003National Natural Science Foundation of China,Grant/Award Numbers:51991395,U1806226,51778033,51822802,71901011,U1811463,51991391Science and Technology Major Project of Beijing,Grant/Award Number:Z191100002519012。
文摘It is crucial to predict future mechanical behaviors for the prevention of structural disasters.Especially for underground construction,the structural mechanical behaviors are affected by multiple internal and external factors due to the complex conditions.Given that the existing models fail to take into account all the factors and accurate prediction of the multiple time series simultaneously is difficult using these models,this study proposed an improved prediction model through the autoencoder fused long-and short-term time-series network driven by the mass number of monitoring data.Then,the proposed model was formalized on multiple time series of strain monitoring data.Also,the discussion analysis with a classical baseline and an ablation experiment was conducted to verify the effectiveness of the prediction model.As the results indicate,the proposed model shows obvious superiority in predicting the future mechanical behaviors of structures.As a case study,the presented model was applied to the Nanjing Dinghuaimen tunnel to predict the stain variation on a different time scale in the future.
基金Under the auspices of National Natural Science Foundation of China(No.42171407,42077242)Key Program of National Natural Science Foundation of China(No.42330607)。
文摘Recently,the outbreak and spread of larch caterpillar(Dendrolimus superans)pests have emerged as significant contributors to forest degradation in the Changbai Mountains,China.Understanding the spatiotemporal distribution patterns of these pests is crucial for effective management and protection of forest ecosystems.This study proposes a pest monitoring approach based on Sentinel imagery.Through time-series analysis,we extracted pest-sensitive features and developed a random forest classifier that integrated Sentinel-1,Sentinel-2,and field sampling data from 2019–2023 to monitor larch caterpillar pests in the Changbai Mountains National Nature Reserve(CMNNR),Northeast China.Our findings indicated that bands green(B3),near-infrared(B8),short wave infrared(B11 and B12)from Sentinel-2 remote sensing images exhibited notable discriminative capabilities for identifying larch caterpillar pests.Specifically,the Normalized Difference Vegetation Index(NDVI)at the end of the growing season emerged as the most valuable feature for pest extraction.Incorporating Synthetic Aperture Radar(SAR)features along with optical data marginally enhances model performance.Furthermore,our approach unveiled the outbreak of larch caterpillar pests,achieving classification map with overall accuracy exceeding 85%and Kappa coefficient surpassing 0.8 for five study years.The pest outbreak began in 2019 and progressively intensified over time.In September 2019,the affected area spanned 114.23 km^(2).The infested area exhibited a declining trend from 2020 to 2023.This study introduces a novel method for the high-precision identification of larch caterpillar pests,offering technical advancements and theoretical underpinnings to support forest management strategies.
基金supported by the National Natural Science Foundation of China(62073140,62073141)the Shanghai Rising-Star Program(21QA1401800).
文摘Fault diagnosis is important for maintaining the safety and effectiveness of chemical process.Considering the multivariate,nonlinear,and dynamic characteristic of chemical process,many time-series-based data-driven fault diagnosis methods have been developed in recent years.However,the existing methods have the problem of long-term dependency and are difficult to train due to the sequential way of training.To overcome these problems,a novel fault diagnosis method based on time-series and the hierarchical multihead self-attention(HMSAN)is proposed for chemical process.First,a sliding window strategy is adopted to construct the normalized time-series dataset.Second,the HMSAN is developed to extract the time-relevant features from the time-series process data.It improves the basic self-attention model in both width and depth.With the multihead structure,the HMSAN can pay attention to different aspects of the complicated chemical process and obtain the global dynamic features.However,the multiple heads in parallel lead to redundant information,which cannot improve the diagnosis performance.With the hierarchical structure,the redundant information is reduced and the deep local time-related features are further extracted.Besides,a novel many-to-one training strategy is introduced for HMSAN to simplify the training procedure and capture the long-term dependency.Finally,the effectiveness of the proposed method is demonstrated by two chemical cases.The experimental results show that the proposed method achieves a great performance on time-series industrial data and outperforms the state-of-the-art approaches.
基金supported by Graduate Funded Project(No.JY2022A017).
文摘The frequent missing values in radar-derived time-series tracks of aerial targets(RTT-AT)lead to significant challenges in subsequent data-driven tasks.However,the majority of imputation research focuses on random missing(RM)that differs significantly from common missing patterns of RTT-AT.The method for solving the RM may experience performance degradation or failure when applied to RTT-AT imputation.Conventional autoregressive deep learning methods are prone to error accumulation and long-term dependency loss.In this paper,a non-autoregressive imputation model that addresses the issue of missing value imputation for two common missing patterns in RTT-AT is proposed.Our model consists of two probabilistic sparse diagonal masking self-attention(PSDMSA)units and a weight fusion unit.It learns missing values by combining the representations outputted by the two units,aiming to minimize the difference between the missing values and their actual values.The PSDMSA units effectively capture temporal dependencies and attribute correlations between time steps,improving imputation quality.The weight fusion unit automatically updates the weights of the output representations from the two units to obtain a more accurate final representation.The experimental results indicate that,despite varying missing rates in the two missing patterns,our model consistently outperforms other methods in imputation performance and exhibits a low frequency of deviations in estimates for specific missing entries.Compared to the state-of-the-art autoregressive deep learning imputation model Bidirectional Recurrent Imputation for Time Series(BRITS),our proposed model reduces mean absolute error(MAE)by 31%~50%.Additionally,the model attains a training speed that is 4 to 8 times faster when compared to both BRITS and a standard Transformer model when trained on the same dataset.Finally,the findings from the ablation experiments demonstrate that the PSDMSA,the weight fusion unit,cascade network design,and imputation loss enhance imputation performance and confirm the efficacy of our design.
文摘Accurate mapping and timely monitoring of urban redevelopment are pivotal for urban studies and decisionmakers to foster sustainable urban development.Traditional mapping methods heavily depend on field surveys and subjective questionnaires,yielding less objective,reliable,and timely data.Recent advancements in Geographic Information Systems(GIS)and remote-sensing technologies have improved the identification and mapping of urban redevelopment through quantitative analysis using satellite-based observations.Nonetheless,challenges persist,particularly concerning accuracy and significant temporal delays.This study introduces a novel approach to modeling urban redevelopment,leveraging machine learning algorithms and remote-sensing data.This methodology can facilitate the accurate and timely identification of urban redevelopment activities.The study’s machine learning model can analyze time-series remote-sensing data to identify spatio-temporal and spectral patterns related to urban redevelopment.The model is thoroughly evaluated,and the results indicate that it can accurately capture the time-series patterns of urban redevelopment.This research’s findings are useful for evaluating urban demographic and economic changes,informing policymaking and urban planning,and contributing to sustainable urban development.The model can also serve as a foundation for future research on early-stage urban redevelopment detection and evaluation of the causes and impacts of urban redevelopment.
基金funded by the project supported by the Natural Science Foundation of Heilongjiang Provincial(Grant Number LH2023F033)the Science and Technology Innovation Talent Project of Harbin(Grant Number 2022CXRCCG006).
文摘Stock price prediction is a typical complex time series prediction problem characterized by dynamics,nonlinearity,and complexity.This paper introduces a generative adversarial network model that incorporates an attention mechanism(GAN-LSTM-Attention)to improve the accuracy of stock price prediction.Firstly,the generator of this model combines the Long and Short-Term Memory Network(LSTM),the Attention Mechanism and,the Fully-Connected Layer,focusing on generating the predicted stock price.The discriminator combines the Convolutional Neural Network(CNN)and the Fully-Connected Layer to discriminate between real stock prices and generated stock prices.Secondly,to evaluate the practical application ability and generalization ability of the GAN-LSTM-Attention model,four representative stocks in the United States of America(USA)stock market,namely,Standard&Poor’s 500 Index stock,Apple Incorporatedstock,AdvancedMicroDevices Incorporatedstock,and Google Incorporated stock were selected for prediction experiments,and the prediction performance was comprehensively evaluated by using the three evaluation metrics,namely,mean absolute error(MAE),root mean square error(RMSE),and coefficient of determination(R2).Finally,the specific effects of the attention mechanism,convolutional layer,and fully-connected layer on the prediction performance of the model are systematically analyzed through ablation study.The results of experiment show that the GAN-LSTM-Attention model exhibits excellent performance and robustness in stock price prediction.
基金supported by the French Agence Nationale de la Recherche(19-CE31-0002 AstroMeso)the European Research Council under the European Union’s Horizon 2020 Research and Innovation Program(Advanced Grant AstroGeo-885250).
文摘When interpreting results,it is imperative to have some understanding of the degree to which the results are replicable.If the results cannot be replicated with independent data,then interpretations from the results become questionable.To minimize the potential for misinterpretations,the current study analyzes six time-series derived from globally sampled U-Pb zircon databases–of which,two are independent igneous databases,one being a quasi-independent igneous database,and three being independent detrital databases.These time-series are then analyzed with standard statistical methods to evaluate replicability.The methods include bandpass filtering to transform the raw time-series into stationary sequences,Student’s t-test,Monte Carlo simulations,periodograms from spectral analysis,correlation studies,and correlograms.Each test is designed to determine the replicability of a specific time-series,as well as the replicability of periodicities found from the time-series.The results show at least three key components to assessing replicability:(a)U-Pb igneous and detrital zircon age distributions are highly replicable,(b)time-series replicability gradually deteriorates with age,and(c)replicability is scale dependent,with low frequency cycles being more replicable than high frequency cycles.From the tests,we conclude that four harmonic cycles are highly replicable and statistically significant,these being periodicities of 810,270,90,and 67.5-myr.
基金funded by the National Plan for Science,Technology and Innovation(MAARIFAH)-King Abdulaziz City for Science and Technology-The Kingdom of Saudi Arabia-award number(13-MAT377-08).
文摘Preterm birth remains a leading cause of neonatal complications and highlights the need for early and accurate prediction techniques to improve both fetal and maternal health outcomes.This study introduces a hybrid approach integrating Long Short-Term Memory(LSTM)networks with the Hybrid Greylag Goose and Particle Swarm Optimization(GGPSO)algorithm to optimize preterm birth classification using Electrohysterogram signals.The dataset consists of 58 samples of 1000-second-long Electrohysterogram recordings,capturing key physiological features such as contraction patterns,entropy,and statistical variations.Statistical analysis and feature selection methods are applied to identify the most relevant predictors and enhance model interpretability.LSTM networks effectively capture temporal patterns in uterine activity,while the GGPSO algorithm finetunes hyperparameters,mitigating overfitting and improving classification accuracy.The proposed GGPSO-optimized LSTM model achieved superior performance with 97.34%accuracy,96.91%sensitivity,97.74%specificity,and 97.23%F-score,significantly outperforming traditional machine learning approaches and demonstrating the effectiveness of hybrid metaheuristic optimization in enhancing deep learning models for clinical applications.By combining deep learning withmetaheuristic optimization,this study contributes to advancing intelligent auto-diagnosis systems,facilitating early detection of pretermbirth risks and timely medical interventions.
基金funded by the Russian Federation Ministry of Science and Higher Education according to the research project FUUW-2025-0033 of the FECIAR UrB RASthe Arkhangelsk branch of the Territorial Fund of Geological Information for the North-Western Federal District for providing geological and geophysical reports for the objects under consideration。
文摘Studying the inner structure of intraplate earthquakes originating in aseismic areas,which are poorly covered by seismic networks or as historical earthquakes is usually the only way to get knowledge about their source mechanisms,which is partially essential for a deeper understanding of intraplate geodynamics.The epicentral zones of earthquakes are situated in hard-toreach areas,so,using active seismic methods for such purposes is unreasonable or even impossible because of high cost and logistical difficulties.We propose a novel approach that combines diverse passive seismic methods,which allows us to get sufficient information about geological environment structure for such task solutions.As an example,we investigated the inner structure of platform earthquake epicentral zone originated up north of Russia.We used four passive seismic methods:microseismic sounding method,passive seismic interferometry,HVSR method,and microseismic activity method.We show that passive seismic data,recoded in the same installation and processed by these different methods,can provide sufficient information about structure of studied environment,needed to explain source mechanism.In sum,the hypocenter zone is presented by intersection of vertical faults and a lateral fractured zone in the middle crust.Results were confirmed by comparison with results by active seismic methods.
基金supported by the National Key R&D Program of China(No.2022YFC2803900)the National Natural Science Foundation of China(No.42276191)Zhejiang University Students Science and Technology Innovation Activity Plan(New Talent Plan)(No.2024R401185).
文摘The spatiotemporal characterization of plume sedimentation and microorganisms is critical for developing plume ecological monitoring model.To address the limitations of traditional methods in obtaining high-quality sediment,a novel sampling system with 6000 m operational capability and three-month endurance was developed.It is equipped with three sediment samplers,a set of formaldehyde preservation solution injection devices.The system is controlled by a low-power,timing-triggered controllers.To investigate low-disturbance rheological mechanisms,gap controlled rheological tests were conducted to optimize the structural design of the sampling and sealing assembly.Stress-controlled shear rheological tests were employed to investigate the mechanisms governing yield stress in sediments under varying temperature conditions and boundary roughness.Additionally,the coupled Eulerian-Lagrangian(CEL)method and sediment rheological constitutive models were employed to simulate tube-soil interaction dynamics and sediment disturbance.The radial heterogeneity of sediment disturbance and friction variation of the sampling tube were revealed.The tube was completely“plugged”at a penetration depth of 261 mm,providing critical data support to the penetration depth parameters.The deep-sea pressure test and South China Sea field trials demonstrated the system’s capability to collect and preserve quantitative time-series sediment samples with high fidelity.
基金financially supported by The National Natural Science Foundation of China (42374004)the Open Fund of Hubei Luojia Laboratory (220100045)the Natural Science Foundation of Sichuan Province (2022NSFSC1047)。
文摘Since April 2002,the Gravity Recovery and Climate Experiment Satellite(GRACE)has provided monthly total water storage anomalies(TWSAs)on a global scale.However,these TWSAs are discontinuous because some GRACE observation data are missing.This study presents a combined machine learning-based modeling algorithm without hydrological model data.The TWSA time-series data for 11 large regions worldwide were divided into training and test sets.Autoregressive integrated moving average(ARIMA),long short-term memory(LSTM),and an ARIMA-LSTM combined model were used.The model predictions were compared with GRACE observations,and the model accuracy was evaluated using fi ve metrics:the Nash-Sutcliff e effi ciency coeffi cient(NSE),Pearson correlation coeffi cient(CC),root mean square error(RMSE),normalized RMSE(NRMSE),and mean absolute percentage error.The results show that at the basin scale,the mean CC,NSE,and NRMSE for the ARIMA-LSTM model were 0.93,0.83,and 0.12,respectively.At the grid scale,this study compared the spatial distribution and cumulative distribution function curves of the metrics in the Amazon and Volga River basins.The ARIMA-LSTM model had mean CC and NSE values of 0.89 and 0.61 and 0.92 and 0.61 in the Amazon and Volga River basins,respectively,which are superior to those of the ARIMA model(0.86 and 0.48 and 0.88 and 0.46,respectively)and the LSTM model(0.80 and 0.41 and 0.89 and 0.31,respectively).In the ARIMA-LSTM model,the proportions of grid cells with NSE>0.50 for the two basins were 63.3%and 80.8%,while they were 54.3%and 51.3%in the ARIMA model and 53.7%and 43.2%in the LSTM model.The ARIMA-LSTM model significantly improved the NSE values of the predictions while guaranteeing high CC values in the GRACE data reconstruction at both scales,which can aid in fi lling in discontinuous data in temporal gravity fi eld models..
基金funded by NARI Group’s Independent Project of China(Grant No.524609230125)the Foundation of NARI-TECH Nanjing Control System Ltd.of China(Grant No.0914202403120020).
文摘Studies to enhance the management of electrical energy have gained considerable momentum in recent years. The question of how much energy will be needed in households is a pressing issue as it allows the management plan of the available resources at the power grids and consumer levels. A non-intrusive inference process can be adopted to predict the amount of energy required by appliances. In this study, an inference process of appliance consumption based on temporal and environmental factors used as a soft sensor is proposed. First, a study of the correlation between the electrical and environmental variables is presented. Then, a resampling process is applied to the initial data set to generate three other subsets of data. All the subsets were evaluated to deduce the adequate granularity for the prediction of the energy demand. Then, a cloud-assisted deep neural network model is designed to forecast short-term energy consumption in a residential area while preserving user privacy. The solution is applied to the consumption data of four appliances elected from a set of real household power data. The experiment results show that the proposed framework is effective for estimating consumption with convincing accuracy.
文摘Accurate energy demand forecasting is crucial in today’s rapidly electrifying world with decentralized systems and integrated renewables.Traditional models struggle with the dynamic complexities,but AI(artificial intelligence),particularly ML(machine learning)and DL(deep learning),offers transformative solutions.This article explores how AI enhances forecasting accuracy,enables real-time adaptability,and supports strategic energy management.It examines the synergy between AI,IoT(Internet of Things)devices,and smart grids in generating predictive and prescriptive insights.Through case studies,we analyze the benefits and challenges of deploying AI in this domain,including data quality,model explainability,and infrastructure needs.Ultimately,AI emerges as a key enabler for the resilient,data-driven energy systems required to meet modern society’s evolving demands and achieve a sustainable future.
文摘Accurate prediction of nurse demand plays a crucial role in efficiently planning the healthcare workforce,ensuring appropriate staffing levels,and providing high-quality care to patients.The intricacy and variety of contemporary healthcare systems and a growing patient populace call for advanced forecasting models.Factors like technological advancements,novel treatment protocols,and the increasing prevalence of chronic illnesses have diminished the efficacy of traditional estimation approaches.Novel forecasting methodologies,including time-series analysis,machine learning,and simulation-based techniques,have been developed to tackle these challenges.Time-series analysis recognizes patterns from past data,whereas machine learning uses extensive datasets to uncover concealed trends.Simulation models are employed to assess diverse scenarios,assisting in proactive adjustments to staffing.These techniques offer distinct advantages,such as the identification of seasonal patterns,the management of large datasets,and the ability to test various assumptions.By integrating these sophisticated models into workforce planning,organizations can optimize staffing,reduce financial waste,and elevate the standard of patient care.As the healthcare field progresses,the utilization of these predictive models will be pivotal for fostering adaptable and resilient workforce management.
基金Project supported by the National Natural Science Foundation of China(Grant No.12275165)XGW was also supported by the Fundamental Research Funds for the Central Universities(Grant No.GK202202003).
文摘Model-free,data-driven prediction of chaotic motions is a long-standing challenge in nonlinear science.Stimulated by the recent progress in machine learning,considerable attention has been given to the inference of chaos by the technique of reservoir computing(RC).In particular,by incorporating a parameter-control channel into the standard RC,it is demonstrated that the machine is able to not only replicate the dynamics of the training states,but also infer new dynamics not included in the training set.The new machine-learning scheme,termed parameter-aware RC,opens up new avenues for data-based analysis of chaotic systems,and holds promise for predicting and controlling many real-world complex systems.Here,using typical chaotic systems as examples,we give a comprehensive introduction to this powerful machine-learning technique,including the algorithm,the implementation,the performance,and the open questions calling for further studies.
文摘The reverse design of solid rocket motor(SRM)propellant grain involves determining the grain geometry to closely match a predefined internal ballistic curve.While existing reverse design methods are feasible,they often face challenges such as lengthy computation times and limited accuracy.To achieve rapid and accurate matching between the targeted ballistic curve and complex grain shape,this paper proposes a novel reverse design method for SRM propellant grain based on time-series data imaging and convolutional neural network(CNN).First,a finocyl grain shape-internal ballistic curve dataset is created using parametric modeling techniques to comprehensively cover the design space.Next,the internal ballistic time-series data is encoded into three-channel images,establishing a potential relationship between the ballistic curves and their image representations.A CNN is then constructed and trained using these encoded images.Once trained,the model enables efficient inference of propellant grain dimensions from a target internal ballistic curve.This paper conducts comparative experiments across various neural network models,validating the effectiveness of the feature extraction method that transforms internal ballistic time-series data into images,as well as its generalization capability across different CNN architectures.Ignition tests were performed based on the predicted propellant grain.The results demonstrate that the relative error between the experimental internal ballistic curves and the target curves is less than 5%,confirming the validity and feasibility of the proposed reverse design methodology.
