Sonnet: Spectral Operator Neural Network for Multivariable Time Series Forecasting

Short-term multivariate time series load data forecasting at low-voltage level using optimised deep-ensemble learning-based models

Prediction of pink esthetic score using deep learning: A proof of concept.

To develop a deep learning (DL) model to measure the sagittal Cobb angle of the cervical spine on computed tomography (CT). Two VB-Net-based DL models for cervical vertebra segmentation and key-point detection were developed. Four-points and line-fitting methods were used to calculate the sagittal Cobb angle automatically. The average value of the sagittal Cobb angle was manually measured by two doctors as the reference standard. The percentage of correct key points (PCK), matched samples t test, intraclass correlation coefficient (ICC), Pearson correlation coefficient, mean absolute error (MAE), and Bland‒Altman plots were used to evaluate the performance of the DL model and the robustness and generalization of the model on the external test set. A total of 991 patients were included in the internal data set, and 112 patients were included in the external data set. The PCK of the DL model ranged from 78 to 100% in the test set. The four-points method, line-fitting method, and reference standard measured sagittal Cobb angles were - 1.10 ± 18.29°, 0.30 ± 13.36°, and 0.50 ± 12.83° in the internal test set and 4.55 ± 20.01°, 3.66 ± 18.55°, and 1.83 ± 12.02° in the external test set, respectively. The sagittal Cobb angle calculated by the four-points method and the line-fitting method maintained high consistency with the reference standard (internal test set: ICC = 0.75 and 0.97; r = 0.64 and 0.94; MAE = 5.42° and 3.23°, respectively; external test set: ICC = 0.74 and 0.80, r = 0.66 and 0.974, MAE = 5.25° and 4.68°, respectively). The DL model can accurately measure the sagittal Cobb angle of the cervical spine on CT. The line-fitting method shows a higher consistency with the doctors and a minor average absolute error.

This paper proposed an operational trapezoid interpolation disaggregation model (OTI_DisTrad) that disaggregates coarse resolution thermal data to fine spatial resolutions using the Vegetation fractional cover (VFC) versus Land surface temperature (LST) space. Developed model attempted to weaken the influences of soil moisture on LST variability at the finer-resolution level. The main assumption underlying our model is that the quantitative relationship between VFC, Bowen ratio and LST that is retrieved at a coarse resolution also works well at finer resolutions for the pixels inside the VFC-LST space. Moreover, the Support Vector Machine (SVM) regression was utilized to address the pixels outside the VFC-LST space by fitting the relationship between Index-based Built-up Index (IBI), surface albedo and LST. The OTI_DisTrad model was tested on two heterogeneous landscapes of the central Iowa, USA (watershed region) and Beijing, China (urban region). Statistical performance measures were used to compare OTI_DisTrad to other five classical approaches (M1, M2, M3, M4 and M5). Results revealed that OTI_DisTrad was more accurate than the three classical ones for the vegetated-dominated watershed region. The average Root mean square error (RMSE) and Mean absolute error (MAE) between disaggregated and reference LST is decreased from 1.83, 1.86 and 1.75 °C (M1, M2 and M3) to 1.59 °C (OTI_DisTrad) and from 1.32, 1.39 and 1.28 °C (M1, M2 and M3) to 1.10 °C (OTI_DisTrad), respectively. Statistical results also reported that OTI_DisTrad was slightly superior to other two classical methods for the urban regions, with the average RMSE and MAE between disaggregated and reference temperature decreased from 2.43 and 2.26 °C (M4 and M5) to 1.97 °C (OTI_DisTrad) and from 1.69 and 1.47 °C (M4 and M5) to 1.21 °C (OTI_DisTrad), respectively.

The health monitoring of CRF (circulation water) pumps is essential for prognostics and management in nuclear power plants. However, the operational status of CRF pumps can vary due to environmental factors and human intervention, and the interrelationships between monitoring parameters are often complex. Consequently, the existing methods face challenges in effectively assessing the health status of CRF pumps. In this study, we propose a health monitoring model for CRF pumps utilizing a meta graph transformer (MGT) observer. Initially, the meta graph transformer, a temporal-spatial graph learning model, is employed to predict trends across the various monitoring parameters of the CRF pump. Subsequently, a fault observer is constructed to generate early warnings of potential faults. The proposed model was validated using real data from CRF pumps in a nuclear power plant. The results demonstrate that the average Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE) of normal predictions were reduced to 1.2385, 0.5614, and 2.6554, respectively. These findings indicate that our model achieves higher prediction accuracy compared to the existing methods and can provide fault warnings at least one week in advance.

Within precision agriculture, yield mapping is important in the evaluation of crop management and delineation of management zones. It can also be used to assess within-field yield potential, in order to guide different precision agriculture practices. However, some farmers do not have a yield monitoring system, and some who do may obtain incomplete or erroneous yield data. This study examined the accuracy with which winter wheat (Triticum aestivum L.) yield could be mapped in 18 fields in southern Sweden using a simple empirical relationship between Sentinel-2 (ESA, Paris, France) data, vegetation index (VI) maps and combined harvester data collected in nearby fields. The results showed that a decrease in map resolution to 40 m reduced the error in the yield maps obtained. Normalised difference water index (NDWI) was the most efficient VI, while a combination of satellite data from earlier and later plant development (booting and milk development stages) performed slightly better than data for other development stages and combinations. The best-performing model at a within-field scale (40-m resolution) had an average mean absolute error (MAE) of 0.40 tonnes ha−1 in a leave-one-field-out cross validation. When the prediction model at field-means scale was applied on 69 farms in a 1055 km2 area, MAE was 0.75 tonnes ha−1 when comparing predictions with mean yields reported by farmers in a phone survey. Therefore, if adequate combined harvester and/or mean yield data are available, a modelling framework that translates satellite imagery into yield maps on-the-fly could be made available for different stakeholders via decision support systems for precision agriculture.

Battery state of health (SOH) is a significant metric for evaluating battery life and predicting battery safety. Currently, SOH research is largely based on laboratory data, with a dearth of research on electric vehicle (EV) operating data. Due to the difficulty in obtaining complete charge data under EV operating conditions, this study presents a SOH estimation method utilizing deep network adaptation. First, a data-driven approach is employed to extract voltage, current, state of charge (SOC), and incremental capacity (IC) data features. To compensate for the lack of aging information in the EV operation data domain, transfer learning is employed to construct the SOH estimation model. Additionally, to resolve inconsistent data distribution between the source laboratory battery data domain and the target EV operation data domain, an adaptive layer is added to the network, and adaptation of deep network (ADN) is utilized to enhance the model’s performance. Finally, the model is validated using electric bus operational data. Results indicate that this model’s average Mean Absolute Error (MAE) is less than 3.0%, and, compared to support vector machine (SVM) regression and Gaussian Process Regression (GPR) algorithms, the MAE is reduced by 27.7% and 38.4%, respectively.

This paper presents a hybrid time-series and adaptive wavelet neural network (AWNN) model for the day-ahead electricity market clearing price forecast. Instead of using price series, one-period continuously compounded return series is used to achieve more attractive statistical properties. The autoregressive moving average with exogenous variables (ARMAX) model is used to catch the linear relationship between price return series and explanatory variable load series, the generalized autoregressive conditional heteroscedastic (GARCH) model is used to unveil the heteroscedastic character of residuals, and AWNN is used to present the nonlinear, nonstationary impact of load series on electricity prices. The Monte Carlo method is adopted to generate more evenly distributed random numbers used for time series and AWNN models to accelerate the convergence. Several criteria such as average mean absolute percentage error (AMAPE) and the variance of forecast errors are used to assess the model and measure the forecasting accuracy. Illustrative price forecasting examples of the PJM market are presented to show the efficiency of the proposed method.

Study regionAmu Darya River Basin (ADRB), Central Asia. Study focusReference-evapotranspiration (ETo) is the most prominent variable for agriculture water-resource management. The Penman-Monteith (PM) equation has been accepted as universal-standard for estimating ETo but it requires extensive climatic-variables. The PM alternative Hargreaves-Samani (HS) equation inadequately estimates ETo with an average error of RMSE 1.84 mm/day, NSE 0.27, and MAE 0.61 mm/day, respectively, in the ADRB. New hydrological insight for the regionThe present study proposes the calibration of original HS-equation based on actual vapor pressure and linear regression to estimate daily ETo at 114 meteorological stations in the ADRB. The locally-calibrated HS-equation coefficient (C) was found variated-widely across the different climatic-conditions and months, ranging from 0.0025 to 0.0186, due to hyper-aridity conditions in the study area. Compared with proposed methods by other studies, the modified-HS equation and empirical-coefficient using newly-proposed variable leads to a significant reduction of the average root means square error and mean absolute error by about 55 % and 52 %, respectively, demonstrates their feasibility and accuracy in the ADRB, Central Asia. In-addition, the modified-HS-equation based on climatic zones were found to be more accurate and comprehensive than geographical localizations. The modified-HS equation and empirical-coefficient using newly-proposed variable and approach is suitable to estimate ETo in a data scarce region (e.g., ADRB), when only temperatures, solar radiation and vapor pressure data are available.

Tea-garden pest control is crucial to ensure tea quality. In this context, the time-series prediction of insect pests in tea gardens is very important. Deep-learning-based time-series prediction techniques are advancing rapidly but research into their use in tea-garden pest prediction is limited. The current study investigates the time-series prediction of whitefly populations in the Tea Expo Garden, Jurong City, Jiangsu Province, China, employing three deep-learning algorithms, namely Informer, the Long Short-Term Memory (LSTM) network, and LSTM-Attention. The comparative analysis of the three deep-learning algorithms revealed optimal results for LSTM-Attention, with an average root mean square error (RMSE) of 2.84 and average mean absolute error (MAE) of 2.52 for 7 days' prediction length, respectively. For a prediction length of 3 days, LSTM achieved the best performance, with an average RMSE of 2.60 and an average MAE of 2.24. These findings suggest that different prediction lengths influence model performance in tea garden pest time series prediction. Deep learning could be applied satisfactorily to predict time series of insect pests in tea gardens based on LSTM-Attention. Thus, this study provides a theoretical basis for the research on the time series of pest and disease infestations in tea plants. © 2024 Society of Chemical Industry.

Geometry-aware PointNet for rapid prediction of cerebral aneurysm hemodynamics.

Road infrastructures are fundamental parts of peoples’ lives, allowing them to access various destinations and activities. Accordingly, infrastructure should be in an appropriate condition. A pavement maintenance plan should be optimized, and pavement condition should be predicted accurately to obtain optimal pavement maintenance solutions. Therefore, the prediction of pavement conditions with high accuracy has been an immense concern. This study aims to introduce a new approach to accurately predict pavement international roughness index (IRI) over the long term. To this end, all the vital parameters, including initial IRI, pavement age, lane width, traffic loadings, structural characteristics, climatic features, and pavement distresses, are considered. With all the vital parameters, the prediction problem includes 58 variables. Thus, the application of a proper feature-selection technique is vital. To this end, a novel hybrid feature-selection method is introduced by a combination of arithmetic optimization algorithm and stochastic gradient descent regression (AOA-SGDR). Moreover, the performance of the proposed feature-selection method is compared with Lasso and all features. Five machine-learning algorithms, including random forest regression (RFR), support vector machine, multi-layer perceptron, decision-tree regression, and multiple linear regression, are employed for the prediction process. By employing AOA-SGDR, the average testing-data mean absolute error (MAE) reduces by at least 7.92%. Meanwhile, RFR provides the highest accuracy, with average testing-data MAE of 0.095 m/km. Moreover, analyzing the parameters indicates that initial IRI, pavement age, equivalent single axle load (ESAL), and structural number (SN) have the most significant relative influence on IRI.

The reliability of hourly PM2.5 data obtained from air quality monitoring stations is compromised as a result of the missing values, thereby impeding the thorough examination of crucial information. In this paper, we present a spatiotemporal (ST) stacking machine learning (ML) method with daily‐cycle restrictions for reconstructing missing hourly PM2.5 records. First, the ST neighbors for the target station with missing values are selected at a daily scale. Subsequently, the non‐null data within the ST neighbors undergo an iterative P‐BSHADE interpolation process for re‐interpolation. Next, a stacking ML model is constructed using the re‐interpolation values and several environmental factors associated with PM2.5 as the predictors, while the observed PM2.5 is taken as the independent variable. Finally, the missing values are reconstructed by inputting the predictors into the trained stacking model. The study utilized hourly PM2.5 data in the Beijing‐Tianjin‐Hebei region as a case study to assess the effectiveness of the proposed method, using daily missing ratios of 10%, 30%, and 50%, respectively. The accuracy of the proposed method was then compared to four contemporary ST interpolation methods. The results indicate that the proposed method exhibits superior performance compared to the classical methods. Specifically, it achieves a reduction in the average root mean square error and mean absolute error by at least 40.6% and 40.1%, respectively. Additionally, the proposed method demonstrates the successful recovery of extreme values in the hourly PM2.5 records, in contrast to the classical methods which often exhibit a tendency to overestimate low values and underestimate high values. Overall, the proposed method presents a viable and efficient approach to recover missing values in the hourly PM2.5 records that demonstrate evident daily periodic patterns.

In this study, the assimilative capacity changes of the Gharehsou River has been studied. Gharehsou River, in the East of Ardabil province of Iran in its flow path, is the entrance and discharge domestic sewage, industrial wastewater and agricultural drainage and for determining of assimilative capacity of this river, first, the calibrated model of Qual2kw simulated parameters change of NO3, BOD, DO, pH and temperature for two months of January and July, and then it was compared with observational data in Aladizgeh, Samian, Anzob and ArbabKandi stations. Given the mean correlation coefficient R 2 and the mean average absolute error (MAE) for two months, the best simulation for pH (R 2 =0.43 and MAE=0.95) and then respectively, for temperature (R 2 =0.56 and MAE=5.75), DO (R 2 =0.69 and MAE=1.75), BOD (R 2 =0.75 and MAE=5.78), NO 3 (R 2 =0.39 and MAE=2.17), have been done and with the exception of nitrate, for other parameters, the Simulation in July was better than January. Results showed that measured and modeled values have a good agreement with each other and the simulation in these parameters has been done well.

Accurate and consistent wind speed forecasting is vital for efficient energy management and the market economy. Wind speed is non-linear, non-stationary, and irregular, so it is very difficult to forecast. There are many forecasting methods currently in use; however, selecting and developing the most appropriate method for a particular region in wind speed forecasting is still a hot topic. This study presents a new and unique neural network-based ensemble system for forecasting wind speed, which is very difficult to predict but is directly related to the power generated by wind farms for individual and different sites. With the developed ensemble model, average mean absolute error, mean absolute percentage error and root mean square error values are obtained as 0.1269, 3.074%, 0.1596 respectively. Test results demonstrate significant contributions of the proposed system compared to existing statistical, heuristic and ensemble models, indicating that the developed model is a promising alternative for wind speed forecasting models. The obtained results show that this system is an effective and useful intelligent tool that can be used by various companies and government facilities that invest and operate in intelligent wind energy technologies.