Components Of Time Series Research Articles

Prediction model of transformer top oil temperature based on data quality enhancement.

Top oil temperature (TOT) is an important parameter to evaluate the running state of a transformer. According to the variation trend of TOT, the internal thermal state of transformers can be predicted so as to arrange operation and maintenance reasonably and prevent the occurrence of accidents. However, due to the complex working environment in the field, there are often a large number of missing values in online monitoring data, which seriously affects the prediction of TOT. At the same time, it is affected by various factors such as load, ambient temperature, wind speed, and solar radiation, which cause the information of different time scales to be mixed in its monitoring data. Therefore, it is difficult to achieve the desired accuracy with a single model. In this article, a model for predicting TOT based on data quality enhancement is proposed. First, the Markov model is used to complete the online monitoring data containing missing values to obtain a complete and continuous time series. Then, using the ensemble empirical modal decomposition method, the time series of TOT is decomposed into multiple time series components to eliminate the interaction between different time scales of information, thus reducing the prediction difficulty. Finally, the sub-prediction model of the extreme learning machine is constructed, and the prediction results of all the sub-models are reconstructed to obtain the final prediction results of TOT. In order to verify the effectiveness of the model, the TOT of an operating transformer for the next two days is predicted in the article, and its mean absolute percentage error (MAPE) is 5.27% and its root mean square error (RMSE) is 2.46. Compared with the BP neural network model and the support vector machines (SVM) model, the MAPE is reduced by 69.56% and 61.92%, respectively, and the RMSE is reduced by 67.02% and 59.87%. The results of this study will provide important support for the fault diagnosis of the top oil temperature online monitoring system.

Visualising the Optimistic, Realistic, and Pessimistic Financial Distress Outlooks for Airport Operations in Malaysia

This paper aims to visualise three financial distress outlooks using computer simulations. The financial distress exposure for airport operations in Malaysia between 1991 and 2021 is given by Altman Z”-score and modelled by the multivariate generalized linear model (MGLM). Seven determinants contributing to the financial distress from literature are examined. The determinant series are fitted individually by using linear model with time series components and autoregressive integrated moving average models to forecast values for the next 10 financial years. Future short- to long-term memory effects following COVID-19 are apparent in time series plots. In the simulations, the MGLM procedure utilised Gaussian, gamma, and Cauchy probability distributions associated with expectations and challenges of doing business as well as uncertainties in the economy. The underlying trends of realistic, optimistic, and pessimistic financial distress outlooks insinuate that the increasing risk of financial distress of airport operations in Malaysia is expected to continue for the next decade.

Exploring Dynamic Structures in Matrix-Valued Time Series via Principal Component Analysis

Time-series data are widespread and have inspired numerous research works in machine learning and data analysis fields for the classification and clustering of temporal data. While there are several clustering methods for univariate time series and a few for multivariate series, most methods are based on distance and/or dissimilarity measures that do not fully utilize the time-dependency information inherent to time-series data. To highlight the main dynamic structure of a set of multivariate time series, this study extends the use of standard variance–covariance matrices in principal component analysis to cross-autocorrelation matrices at time lags k=1,2,…. This results in “principal component time series”. Simulations and a sign language dataset are used to demonstrate the effectiveness of the proposed method and its benefits in exploring the main structural features of multiple time series.

Network connectivity and structural correlates of survival in progressive supranuclear palsy and corticobasal syndrome.

There is a pressing need to understand the factors that predict prognosis in progressive supranuclear palsy (PSP) and corticobasal syndrome (CBS), with high heterogeneity over the poor average survival. We test the hypothesis that the magnitude and distribution of connectivity changes in PSP and CBS predict the rate of progression and survival time, using datasets from the Cambridge Centre for Parkinson-plus and the UK National PSP Research Network (PROSPECT-MR). Resting-state functional MRI images were available from 146 participants with PSP, 82 participants with CBS, and 90 healthy controls. Large-scale networks were identified through independent component analyses, with correlations taken between component time series. Independent component analysis was also used to select between-network connectivity components to compare with baseline clinical severity, longitudinal rate of change in severity, and survival. Transdiagnostic survival predictors were identified using partial least squares regression for Cox models, with connectivity compared to patients' demographics, structural imaging, and clinical scores using five-fold cross-validation. In PSP and CBS, between-network connectivity components were identified that differed from controls, were associated with disease severity, and were related to survival and rate of change in clinical severity. A transdiagnostic component predicted survival beyond demographic and motion metrics but with lower accuracy than an optimal model that included the clinical and structural imaging measures. Cortical atrophy enhanced the connectivity changes that were most predictive of survival. Between-network connectivity is associated with variability in prognosis in PSP and CBS but does not improve predictive accuracy beyond clinical and structural imaging metrics.

Spatial-Temporal Changes of Abarkuh Playa Landform from Sentinel-1 Time Series Data

Playas, as the flattest landforms in semiarid and arid regions, are extremely sensitive to climate changes, such as changes in the hydrologic regime of the landscape. The changes in these landforms due to irrigation, anthropogenic activities, and climate change could be a source of disasters. In this study, we explored the spatial-temporal changes of the Abarkuh Playa in Central Iran using the time series of the Sentinel-1 backscatter dataset in the three scales. Our results showed that the western area of the Abarkuh Playa has been changed to other landforms with different characteristics, which is clear from all backscatter maps. The spatial-temporal analysis of the time series of backscatter data using the independent component analysis and time series of precipitation revealed that the backscatter variations were associated with direct rainfall across the playa and the surface was reacting to changes in the soil moisture content. The results of the power scale showed that the boundary of the playa could successfully be recognized as the oscillating pattern from other landforms in the study area. Moreover, the spatial-temporal analysis of backscatter in the power scale showed that different polarizations could reveal different patterns of surface changes for the playa.

Solar Hysteresis Pattern and Spectral Components in TEC Time Series (GPS and TIE‐GCM) of the Quadrilaterally Coupled Geomagnetic Conjugate Low‐Latitude Stations

AbstractThe current study is a first‐of‐its‐kind in that it compares the Total Electron Content (TEC) of the solar cycle from the Global Positioning System (GPS) and Thermosphere‐Ionosphere‐Electrodynamics General Circulation Model (TIE‐GCM) over the geomagnetically conjugate low‐latitude stations with the particular scenario of lower atmospheric conditions over land‐ and sea‐locked locations. For this, TEC data for the period 2009–2017 is used from northern hemispheric stations Varanasi (25.31°N; 82.97°E) and LHAZ (29.65°N; 91.10°E) and southern hemispheric stations DGAR (7.27°S; 72.37°E) and COCO Island (12.18°S; 96.83°E). The solar cycle variation in TEC is identified by two distinct maxima and hysteresis between the ascending and descending phases. The solar cycle trends are modulated by the equatorial ionization anomaly as well as longitudinal biases. The Lomb‐Scargle periodogram shows that the improved TIE‐GCM version 2.0, which incorporates variable eddy diffusion to provide an accurate simulation of seasonal variability, is largely successful in simulating semi‐annual and annual oscillations but still needs to resolve the seasonal anomaly feature, particularly in the case of southern low latitude stations. Terannual (120‐day) and 1.4‐year (500‐day) periodicities in the TEC time series are observed only at EIA region stations, not at off‐crest location LHAZ, and are most likely caused by E x B drift. The wavelet coherence analysis reveals that the Quasi Biannual Oscillations (QBO) in the TEC time series (597‐, 773‐, and 930‐day) have a strong physical affinity with the QBO oscillation of F10.7 flux. Results indicate that both solar activity and equatorial electrodynamics significantly influence the QBO oscillation in the TEC.

Resilient Embedded Systems Designs via on the Fly Generation of Adaptive Degenerate Components Using Machine Learning

Software defined radio (SDR) provides significant advantages over traditional analog radio systems and are becoming increasingly relied on for “mission critical” applications. This along with risk of trojans, single-event upsets and human error creates the necessity for fault tolerant systems. Redundancy has been traditionally used to implement fault tolerance but incurs a substantial area overhead which is undesirable in most applications. Advancements in field-programmable gate array and system on a chip technologies have made implementing machine learning (ML) algorithms within embedded systems feasible. In this paper we explore the use of ML to implement fault tolerance in an SDR. Our approach, which we call adaptive component-level degeneracy (ACD), uses a ML model to learn the functionality of an SDR component. Once trained, the model can detect when the component is compromised and mitigate the issue with its own output. We demonstrate the ability of our model to learn multiple simulated SDR components. We compare the one-dimensional convolutional neural network and bidirectional recurrent neural network architectures at modeling time series components. We also implement ACD within a real-time SDR system using GNU Radio Companion. The results show great potential for the utilization of ML techniques for improving embedded system reliability.

Artificial intelligence and sensor-based autism spectrum disorder diagnosis using brain connectivity analysis

Autism spectrum disorder (ASD) is a complicated neurodevelopmental condition whose cause is unclear, and due to its unusual pattern, it is difficult to diagnose the disease at the right time. Because of its ability to automatically uncover complicated patterns in high-dimensional data, artificial intelligence (AI) can be a beneficial tool. Recent improvements in neuroimaging technologies using biosensors have enabled the quantification of functional abnormalities associated with ASD. This work proposes an approach to constructing a functional connectivity network from resting state functional Magnetic Resonance Image (rs-fMRI) data. For obtaining a functional connectivity network, the time series component of fMRI data is used, and from it, a correlation matrix is calculated showing the degree of interaction among the brain regions. Several brain atlases have been considered in the experiment. With the majority voting concept based on the results from the atlas, the proposed technique reveals an ASD detection accuracy of 84.79%.

Predicting Top Five Cryptocurrency Prices via Linear Structural Time Series (STS) Approach

Predicting cryptocurrency prices are difficult due to dynamic data. At the same time, the hidden market behavior of trend and seasonal components in the history data is also critical as it provides an idea of what the price pattern will be in the future. Hence, this research proposes to identify and model the hidden pattern behavior in terms of component time series instead of removing it via the linear structural time series (STS) model approach. This study focuses on the top five cryptocurrencies relying on the highest market capitalization. From the results obtained, the top five cryptocurrencies have a different trend model, either deterministic or stochastic, which relies on the behavior of data. The five cryptocurrencies also show the crypto winter event, where the trend is downward after six months every year. The linear STS is the best model for predicting three cryptocurrencies’ prices for nonstationary and volatility data behavior. It can also handle the hidden component behavior and is easy to interpret. Since the linear STS model can indirectly retain the information of data, it will assist investors and traders in accurately predicting cryptocurrency prices.

On the anomalous structure of the Southeast Vietnam Offshore Current during 1994 to 2015

The summer circulation in the South China Sea (SCS) often has a dipole structure associated with the eastward Southeast Vietnam Offshore Current (SVOC). The dipole has an anti-cyclone to the south and a cyclone to the north. A 22-year (1994–2015) HYbrid Coordinate Ocean Model (HYCOM) global reanalysis is used to study why the SVOC in the SCS is anomalous in summer of 1995, 1998, 2007, 2010, and 2015. The normal SVOC Sea Surface Height (SSH) pattern is defined based on previous observations of the dipole associated with it. An Empirical Orthogonal Function (EOF) analysis is performed on the monthly mean SSH and wind stress curl to seek the answer. SSH EOF1 and EOF2 explain 43.3% and 18.9% of the total SSH variance, respectively. Wind stress curl EOF1 and EOF2 explain 63.4% and 14.9% of the total variance, respectively. SSH EOF1 is a basin-scale gyre mode that represents the barotropic response to the basin-scale monsoon wind stress curl EOF1. SSH EOF2 is a dipole mode that represents a baroclinic response to both wind stress curl EOF1 and EOF2 (western North Pacific anticyclone mode) via forced baroclinic Rossby waves. SSH EOF2 explains 39.6% of the SSH variance in the SVOC dipole and is more important for the SVOC formation than SSH EOF1, which explains 28.2% of the SVOC dipole SSH variance. SSH EOF2 principal component time series indicate that the summer dipole mode in 1995, 1998, 2010, and 2015 is associated with an anti-cyclone to the north and a cyclone to the south, opposite to the normal SVOC dipole. In 2007, the SSH EOF2 is very weak with its principal component close to zero.

Clustering multivariate time series using energy distance

A novel methodology is proposed for clustering multivariate time series data using energy distance defined in Székely and Rizzo (2013). Specifically, a dissimilarity matrix is formed using the energy distance statistic to measure the separation between the finite‐dimensional distributions for the component time series. Once the pairwise dissimilarity matrix is calculated, a hierarchical clustering method is then applied to obtain the dendrogram. This procedure is completely nonparametric as the dissimilarities between stationary distributions are directly calculated without making any model assumptions. In order to justify this procedure, asymptotic properties of the energy distance estimates are derived for general stationary and ergodic time series. The method is illustrated in a simulation study for various component time series that are either linear or nonlinear. Finally, the methodology is applied to two examples; one involves the GDP of selected countries and the other is the population size of various states in the U.S.A. in the years 1900–1999.

A Framework for Estimating Global River Discharge From the Surface Water and Ocean Topography Satellite Mission

AbstractThe Surface Water and Ocean Topography (SWOT) mission will vastly expand measurements of global rivers, providing critical new data sets for both gaged and ungaged basins. SWOT discharge products (available approximately 1 year after launch) will provide discharge for all river that reaches wider than 100 m. In this paper, we describe how SWOT discharge produced and archived by the US and French space agencies will be computed from measurements of river water surface elevation, width, and slope and ancillary data, along with expected discharge accuracy. We present for the first time a complete estimate of the SWOT discharge uncertainty budget, with separate terms for random (standard error) and systematic (bias) uncertainty components in river discharge time series. We expect that discharge uncertainty will be less than 30% for two‐thirds of global reaches and will be dominated by bias. Separate river discharge estimates will combine both SWOT and in situ data; these “gage‐constrained” discharge estimates can be expected to have lower systematic uncertainty. Temporal variations in river discharge time series will be dominated by random error and are expected to be estimated within 15% for nearly all reaches, allowing accurate inference of event flow dynamics globally, including in ungaged basins. We believe this level of accuracy lays the groundwork for SWOT to enable breakthroughs in global hydrologic science.

Bayesian herd detection for dynamic data

This article analyzes multiple agents who forecast an underlying dynamic state based on streams of (partially overlapping) information. Each agent minimizes a convex combination of their forecasting error and deviation from the other agents’ forecasts. As a result, the agents exhibit herding behavior, a bias well-recognized in the economics and psychology literature. Our first contribution is a general framework for analyzing agents’ forecasts under different levels of herding. The underlying state dynamics can be non-linear with seasonality, trends, shocks, or other time series components. Our second contribution describes how models within our framework can be estimated from data. We apply our estimation procedure to surveys of equity price forecasts and find that the agents concentrate 37% of their efforts on making similar forecasts on average. However, there is substantial variation in the level of herding over time; even though herding fell substantially during the 2007–2008 financial crisis, it rose after the crisis.

Wavelet Analysis for Studying Rainfall Variability and Regionalizing Data: An Applied Study of the Moulouya Watershed in Morocco

This study analyzes the spatiotemporal variability of precipitation at the scale of the Moulouya watershed in eastern Morocco, which is very vulnerable to the increasing water shortage. For this purpose, we opted for wavelet transformation, a method based on the spectral analysis of data which allows for periodic components of a rainfall time series to change with time. The results obtained from this work show spectral power across five frequency ranges of variability: 1 to 2 years, 2 to 4 years, 4 to 8 years, 8 to 16 years, and 16 to 32 years. The duration of significant power at these frequencies is generally not homogeneous and varies from station to station. The most widespread frequency over the entire study area was found in the 4- to 8-year range. This mode of variability can last up to 27 consecutive years. In most of the basin, this mode of variability was observed around the period between 1990 and 2010. Oscillations at 8 to 16 years in frequency appear in only five series and over different time periods. The 16- to 32-year mode of variability appears in 15 stations and extends over the period from 1983 to 2008. At this level, signal strength is very weak compared to other higher-frequency modes of variability. On the other hand, the mode of variability at the 1- to 2-year frequency range appeared to be continuous in some stations and intermittent in others. This allowed us to regionalize our study basin into two homogeneous clusters that only differ in variability and rainfall regime.

On the capabilities of the SWOT satellite to monitor the lake level change over the Third Pole

The lake level dynamics of the Qinghai–Tibetan Plateau (QTP, also called the ‘Third Pole’) are a crucial indicator of climate change and human activities; however, they remain poorly measured due to extremely high elevation and cold climate. The existing satellite altimeters also suffer from relatively coarse temporal resolution or low spatial coverage, preventing effective monitoring of lake level change at such a large spatial scale. The recently launched surface water and ocean topography (SWOT) mission is expected to greatly enhance the current lake level monitoring capabilities. However, a systematic evaluation is still lacking in the region. To elucidate this potential, here, we generated SWOT-like lake products for 38 major lakes (>150 km2) over the QTP during 2000–2018 using a large-scale SWOT hydrology simulator with the input of satellite altimetry and water mask databases. The comparative assessments between the satellite altimetry data and SWOT simulations using various statistical metrics and decomposed time series components demonstrate that SWOT can successfully monitor both short-term dynamics and long-term trends. Extended experiments to derive SWOT-like data of 783 lakes (>1 km2) based on the synthetic lake level series present the spatial pattern of SWOT performance that tends to improve with the increasing lake area. Our findings provide comprehensive inferences and confidence for lake level monitoring in the Third Pole in the early period of the SWOT satellite.

Multicomponent Activity Cycles Using Hilbert–Huang Analysis

The temporal analysis of stellar activity evolution is usually dominated by a complex trade-off between model complexity and interpretability, often by neglecting the nonstationary nature of the process. Recent studies appear to indicate that the presence of multiple coexisting cycles in a single star is more common than previously thought. The correct identification of physically meaningful cyclic components in spectroscopic time series is therefore a crucial task, which cannot overlook local behaviors. Here we propose a decomposition technique that adaptively recovers amplitude- and frequency-varying components. We present our results for the solar activity as measured both by the sunspot number and the K-line emission index, and we consistently recover the Schwabe and Gleissberg cycles as well as the Gnevyshev–Ohl pattern probably related to the Hale cycle. We also recover the known 8 yr cycle for 61 Cygni A, in addition to evidence of a three-cycles-long pattern reminiscent of the Gnevyshev–Ohl rule. This is particularly interesting as we cannot discard the possibility of a relationship between the measured field polarity reversals and this Hale-like periodicity.

From point forecasts to multivariate probabilistic forecasts: The Schaake shuffle for day-ahead electricity price forecasting

Modeling price risks is crucial for economic decision making in energy markets. Besides the risk associated with a single price, the dependence structure of multiple prices is often relevant. We therefore propose a generic and easy-to-implement method for generating multivariate probabilistic forecasts based on univariate point forecasts of day-ahead electricity prices. While each univariate point forecast refers to one of the day’s 24 hours, the multivariate forecast distribution models dependencies across hours. The proposed method is based on simple copula techniques and an optional time series component. We illustrate the method for five benchmark data sets. Furthermore, we demonstrate an example for constructing realistic prediction intervals for the weighted sum of consecutive electricity prices as needed for pricing individual load profiles.

Statistical modelling and forecasting of wheat and meslin export from Ukraine using the singular spectral analysis

The article addresses the problems related to the functioning of the worldwide market of wheat and meslin. The authors identify the countries that over the past 17 years have been among the top 10 world leaders in terms of the value of export and import of wheat and meslin. The structure of wheat export by Ukrainian regions is analysed in comparison with the total export. The localisation coefficient is applied to measure the regional unevenness of the distribution of wheat export volumes and the total export by regions of the country. The modelling and forecasting of the volumes and prices of export of wheat and meslin from Ukraine are based on Singular Spectrum Analysis. The study particularly focuses on the individual components of time series, such as trend, annual, semi-annual, four-month, three-month seasonal components. The reliability of the forecast is confirmed by the calculation of the MAPE forecast error and Henry Theil’s inequality coefficient. The article proposes an algorithm for calculating the relative indicators of the structure for the individual components of the reconstructed time series, identified through the singular spectral analysis.

A data-driven adversarial machine learning for 3D surrogates of unstructured computational fluid dynamic simulations

This paper presents a general workflow to generate and improve the forecast of model surrogates of computational fluid dynamics simulations using deep learning, and most specifically adversarial training. This adversarial approach aims to reduce the divergence of the forecasts from the underlying physical model. Our two-step method integrates a Principal Components Analysis (PCA) based adversarial autoencoder (PC-AAE) with adversarial Long short-term memory (LSTM) networks. Once the reduced-order model (ROM) of the CFD solution is obtained via PCA, an adversarial autoencoder is used on the principal components time series. Subsequentially, a LSTM is adversarially trained on the latent space produced by the PC-AAE to make forecasts. Here we show, that the application of adversarial training improves the rollout of the latent space predictions. Our workflow is applied to three different case studies including two models of urban air pollution in London.

Bayesian inference and wind field statistical modeling applied to multiple source estimation

We present a methodology to identify multiple pollutant sources in the atmosphere that combines a data-driven dispersion model with Bayesian inference and uncertainty quantification. The dispersion model accounts for a realistic wind field based on the output of a multivariate dynamic linear model (DLM), estimated from measured wind components time series. The forward problem solution, described by an adjoint transient advection-diffusion partial differential equation, is then obtained using an appropriately stabilized finite element formulation. The Bayesian inference tool accounts for uncertainty in the concentration data and automatically states the balance between the prior and the likelihood. The source parameters are estimated by a Metropolis in Gibbs Monte Carlo Markov chain (MCMC) algorithm with adaptive steps. The MCMC algorithm is initialized with a maximum a posteriori estimator obtained with particle swarm optimization to accelerate convergence. Finally, the proposed methodology seems to outperform inversion techniques from previous works.