Normalized Root Mean Square Error Research Articles

A Modified Method for Reducing the Scale Effect in Land Surface Temperature Downscaling at 10 m Resolution

Satellite-derived Land Surface Temperature (LST) plays an important role in research on natural energy balance and water cycle. Considering the tradeoff between spatial and temporal resolutions, accurate fine-resolution LST must be obtained through the use of LST downscaling (DLST) technology. Various methods have been proposed for DLST at fine resolutions (e.g., 10 m) and small scales. However, the scale effect of these methods, which is inherent to DLST processes at different extents, has rarely been addressed, thus limiting their application. In this study, a modified daily 10 m resolution DLST method based on Google Earth Engine, called mDTSG, is proposed in order to reduce the scale effect at fine spatial resolutions. The proposed method introduces a convolution-based moving window into the DLST process for the fusion of different remote sensing data. The performance of the modified method is compared with the original method in six regions characterized by various extents and landscape heterogeneity. The results show that the scale effect is significant in the DLST process at fine resolutions across extents ranging from 100 km2 to 22,500 km2. Compared with the original method, mDTSG can effectively reduce the LST value differences between tile edges, especially when considering large extents (>22,500 km2) with an average R2 improvement of 33.75%. The average MAE is 1.63 °C, and the average RMSE is 2.3 °C in the mDTSG results, when compared with independent remote sensing products across the six regions. A comparison with in situ observations also shows promising results, with an MAE of 2.03 °C and an RMSE of 2.63 °C. These findings highlight the robustness and scalability of the mDTSG method, making it a valuable tool for fine-resolution LST applications in diverse and extensive landscapes.

A Copernicus-based evapotranspiration dataset at 100 m spatial resolution over four Mediterranean basins

Abstract. Evapotranspiration (ET) is responsible for regulating the hydrological cycle, with a relevant impact on air humidity and precipitation that is particularly important in the context of acute drought events in recent years. With the intensification of rainfall deficits and extreme heat events, the Mediterranean region requires regular monitoring to enhance water resource management. Even though remote sensing provides spatially continuous information for estimating ET on large scales, existing global products with spatial resolutions ≥ 0.5 km are insufficient for capturing spatial detail at a local level. In the framework of ESA's 4DMED-Hydrology project, we generate an ET dataset at both high spatial and high temporal resolutions using the Priestley–Taylor Two-Source Energy Balance (TSEB-PT) model driven by Copernicus satellite data. We build an automatic workflow to generate a 100 m ET product by combining data from Sentinel-2 (S2) MultiSpectral Instrument (MSI) and Sentinel-3 (S3) land surface temperature (LST) with ERA5 climate reanalysis derived within the period 2017–2021 over four Mediterranean basins in Italy, Spain, France, and Tunisia (Po, Ebro, Hérault, and Medjerda). First, original S2 data are pre-processed before deriving 100 m inputs for the ET estimation. Next, biophysical variables, like leaf area index and fractional vegetation cover, are generated, and then they are temporally composited within a 10 d window according to S3 acquisitions. Consequently, decadal S2 mosaics are used to derive the remaining TSEB-PT inputs. In parallel, we sharpen 1 km S3 by exploiting the dependency between coarse-resolution LST and 100 m S2 reflectances using a decision tree algorithm. Afterwards, climate forcings are utilized to model energy fluxes and then for daily ET retrieval. The daily ET composites demonstrate reasonable TSEB-PT estimates. Based on the validation results against eight eddy covariance (EC) towers between 2017 and 2021, the model predicts 100 m ET with an average RMSE of 1.38 mm d−1 and a Pearson coefficient equal to 0.60. Regardless of some constraints mostly related to the high complexity of EC sites, TSEB-PT can effectively estimate 100 m ET, which opens up new opportunities for monitoring the hydrological cycle on a regional scale. The full dataset is freely available at https://doi.org/10.48784/b90a02d6-5d13-4acd-b11c-99a0d381ca9a, https://doi.org/10.48784/fb631817-189f-4b57-af6a-38cef217bad3, https://doi.org/10.48784/70cd192c-0d46-4811-ad1d-51a09734a2e9, and https://doi.org/10.48784/7abdbd94-ddfe-48df-ab09-341ad2f52e47 for the Ebro, Hérault, Medjerda, and Po catchments, respectively (Bartkowiak et al., 2023a–d).

Exploring charge sharing compensation using inter-pixel coincidence counters for photon counting detectors by deep-learning from local information

Photon counting detectors (PCDs) have well-acknowledged advantages in computed tomography (CT) imaging. However, charge sharing and other problems prevent PCDs from fully realizing the anticipated potential in diagnostic CT. PCDs with multi-energy inter-pixel coincidence counters (MEICC) have been proposed to provide particular information about charge sharing, thereby achieving lower Cramér-Rao Lower Bound (CRLB) than conventional PCDs when assessing its performance by estimating material thickness or virtual monochromatic attenuation integrals (VMAIs). This work explores charge sharing compensation using local spatial coincidence counter information for MEICC detectors through a deep-learning method.
Approach: By analyzing the impact of charge sharing on photon count detection, we designed our network with a focus on individual pixels. Employing MEICC data of patches centered on POIs as input, we utilized local information for effective charge sharing compensation. The output was VMAI at different energies to address real detector issues without knowledge of primary counts. To achieve data diversity, a fast and online data generation method was proposed to provide adequate training data. A new loss function was introduced to reduce bias for training with high-noise data. The proposed method was validated by Monte Carlo (MC) simulation data for MEICC detectors that were compared with conventional PCDs. 
Main-Results: For conventional data as a reference, networks trained on low-noise data yielded results with a minimal bias (about 0.7%) compared with > 3% for the polynomial fitting method. The results of networks trained on high-noise data exhibited a slightly increased bias (about 1.3%) but a significantly reduced standard deviation (STD) and normalized root mean square error (NRMSE). The simulation study of the MEICC detector demonstrated superior compared to the conventional detector across all the metrics. Specifically, for both networks trained on high-noise and low-noise data, their biases were reduced to about 1% and 0.6%, respectively. Meanwhile, the results from a MEICC detector were of about 10% lower noise than a conventional detector. Moreover, an ablation study showed that the additional loss function on bias was beneficial for training on high-noise data.
Significance: We demonstrated that a network-based method could utilize local information in PCDs effectively by patch-based learning to reduce the impact of charge sharing. MEICC detectors provide very valuable local spatial information by additional coincidence counters. Compared with MEICC detectors, conventional PCDs only have limited local spatial information for charge sharing compensation, resulting in higher bias and standard deviation in VMAI estimation with the same patch strategy.&#xD.

Imaging error reduction in radial cine-MRI with deep learning-based intra-frame motion compensation.

Radial cine-MRI allows for sliding window reconstruction at nearly arbitrary frame rate, promising high-speed imaging for intra-fractional motion monitoring in MRgRT. However, motion within the reconstruction window may determine the location of the reconstructed target to deviate from the true real-time position (target positioning errors), particularly in cases of fast breathing or for anatomical structures affected by the heartbeat. In this work, we present a proof-of-concept study aiming to enhance radial cine-MR imaging by implementing deep-learning-based intra-frame motion compensation techniques.
Approach: A novel network (TransSin-UNet) was proposed to continuously estimate the final-position image of the target, corresponding to end of the frame acquisition. Within the radial k-space reconstruction window, the spatial-temporal dependencies among the sinogram representation of the spokes were modeled by a transformer encoder subnetwork, followed by a UNet subnetwork operating in the spatial domain for pixel-level fine-tuning. By simulating motion-dependent radial sampling with (tiny) golden angles, we generated datasets from 25 4D digital anthropomorphic lung cancer phantoms. The network was then trained and extensively evaluated across datasets characterized by varying azimuthal radial profile increments.
Main Results: The method required additional 4.8ms per frame over the conventional approach involving direct image reconstruction with motion-corrupted spokes. TransSin-UNet outperformed architectures relying solely on transformer encoders or UNets across all the comparative evaluations, leading to a noticeable enhancement in image quality and target positioning accuracy. The normalized root mean-squared error (NRMSE) decreased by 50% from the initial value of 0.188 on average, whereas the mean Dice similarity coefficient (DSC) of the gross tumor volume (GTV) increased from 85.1% to 96.2% in the investigated cases. Furthermore, the final-positions of anatomical structures undergoing substantial intra-frame deformations were precisely derived.
Significance: The proposed approach enables an effective intra-frame motion compensation, offering an opportunity to reduce errors in radial cine-MR imaging for real-time motion management.&#xD.

Assimilation of Satellite-Derived Reservoir Storage Data to Improve Global Hydrodynamic Modeling

In this study, we explore the potential of assimilating satellite-derived reservoir storage data into the global-scale hydrodynamic model CaMa-Flood, focusing on the Yangtze River basin. We evaluated three data assimilation (DA) methods: direct assimilation (DIR), anomaly based assimilation (ANO), and normalized assimilation (NOM). Our results show that the DIR method achieved the most significant improvements in reservoir storage and downstream discharge simulations. DIR reduced the average relative root mean square error (rRMSE) of reservoir storage estimates by 80.5%, and increased discharge correlation (ΔCC) by 78.6% in the 14 validated discharge stations. ANO, while effective in certain cases, led to mixed results, with 56.4% of the 39 assimilated dams showing improved storage estimates and a modest 7.8% reduction in average RMSE. NOM had minimal impact, with negligible changes in RMSE or discharge correlation (ΔCC). The direct assimilation method (DIR) consistently outperformed the others, improving both reservoir storage and downstream discharge estimates. However, the magnitude of improvement varied across locations, highlighting the need for the further refinement of DA techniques and input data, especially for regions with complex reservoir operations. Our findings enhance reservoir representation in global hydrodynamic models and improve the predictability of river dynamics and water resource management.

Spatio-Temporal Model to Forecast COVID-19 Confirmed Cases in High-Density Areas of Malaysia

The coronavirus 2019 disease has spread across the world. The number of coronaviruses 2019 (COVID-19) cases throughout Malaysia is high in the densely populated state of Selangor. In assisting the early preventive measures, this study utilises time series methods to model and forecast the number of daily positive cases in three Selangor districts: Petaling, Hulu Langat, and Klang. Specifically, the study compares the effectiveness of the Autoregressive Integrated Moving Average (ARIMA), a univariate model and the Generalized Space-Time autoregressive integrated (GSTARI), a multivariate model. For the GSTARI model, uniform and inverse distance weights represent the relationship between locations. The analysed data are from January to August 2021, and the lowest root mean square error (RMSE) is chosen as the best model. The results show GSTARI (1,1) with both spatial weights outperformed ARIMA (0,1,1) in Petaling and Klang but not in Hulu Langat. However, the average RMSE values show that the most accurate and effective for forecasting the number of daily confirmed positive cases in Selangor is using GSTARI. In conclusion, by utilising advanced time series methods such as spatial analysis, this study provides important insights into forecasting trends of infectious diseases like COVID-19 and can help in early preventive measures.

Design a novel algorithm for enhancing UWB positioning accuracy in GPS denied environments

Accurate indoor positioning is the key to the development of the Internet of Things and intelligent devices. In view of GPS-denied indoor environments, we propose to build the indoor local positioning system by using ultra-wide band (UWB) system. In order to enhance the localization accuracy of UWB system, we propose a novel algorithm which integrates the Maximum Correntropy Criterion (MCC) and unscented Kalman filter (UKF) method to reconstruct the measurement distance by using the maximum entropy principle to reduce the influence of outliers and unknown process noise on the smooth effect. Subsequently, the least square (LS) method is implemented to attain the target node (TN) initial position coordinates, and the Taylor algorithm is then performed to further optimize the localization results of the LS method. Lastly, the experimental investigation is conducted to assess the effectiveness and applicability of the developed method via the UWB system in indoor scenarios. The experimental outcomes demonstrate that the developed MCCUKF-LS method can achieve the lowest root mean square error (RMSE), and enhance the positioning accuracy of the TN compared with the LS, KF-LS, and UKF-LS methods. The overall average RMSE of MCCUKF-LS method is reduced by 45.7% contracted with the LS algorithm. The average error of x-, y- and z-axis orientation for the LS method is reduced from 0.074 m, 0.067 m, 0.098 m to 0.036 m, 0.034 m, 0.044 m, and the achieved accuracy in the orientation of the three axes is increased by 51.4%, 49.3% and 55.1% respectively, which reveals that the designed fusion technique is capable of enhancing the positioning accuracy of the TN effectively, providing a new positioning methodology and reference for indoor positioning in GPS-denied environments.

Linked Data-Driven, Physics-Based Modeling of Pumping-Induced Subsidence with Application to Bangkok, Thailand.

Research into land subsidence caused by groundwater withdrawal is hindered by the availability of measured heads, subsidence, and forcings. In this paper, a parsimonious, linked data-driven and physics-based approach is introduced to simulate pumping-induced subsidence; the approach is intended to be applied at observation well nests. Time series analysis using response functions is applied to simulate heads in aquifers. The heads in the clay layers are simulated with a one-dimensional diffusion model, using the heads in the aquifers as boundary conditions. Finally, simulated heads in the layers are used to model land subsidence. The developed approach is applied to the city of Bangkok, Thailand, where relatively short time series of head and subsidence measurements are available at or near 23 well nests; an estimate of basin-wide pumping is available for a longer period. Despite the data scarcity, data-driven time series models at observation wells successfully simulate groundwater dynamics in aquifers with an average root mean square error (RMSE) of 2.8 m, relative to an average total range of 21 m. Simulated subsidence matches sparse (and sometimes very noisy) land subsidence measurements reasonably well with an average RMSE of 1.6 cm/year, relative to an average total range of 5.4 cm/year. Performance is not good at eight out of 23 locations, most likely because basin-wide pumping is not representative of localized pumping. Overall, this study demonstrates the potential of a parsimonious, linked data-driven, and physics-based approach to model pumping-induced subsidence in areas with limited data.

Evaluation of wavelet decomposition level to enhance numerical analysis of seismic responses

In this study, the discrete wavelet transform (DWT) is used to enhance the efficiency of numerical analysis by decomposing earthquake excitation signals into low- and high-frequency components. The low-frequency components, considered as the main damaging factor in earthquake waves, stored in approximation coefficients have been used to perform time response analysis. In the selection of the ground motions, the special attention is given to the ratio of peak ground acceleration (PGA) to peak ground velocity (PGV). An ensemble of nine earthquake records have been selected and used. For each earthquake, displacement response spectra have been generated for both the original signal and its approximations at different DWT levels. To determine the most accurate decomposition level, the normalized root-mean-squared error (NRMSE) of wavelet energy has been computed in spectral displacement responses. The outcomes reveal a dependency of the decomposition level on the specific characteristics of each earthquake. Subsequently, structural analyses are conducted on base-isolated structures with varying story numbers, to examine the accuracy of decomposed signals. Employing both the original signal and its approximations to the structures, base mat displacements and roof accelerations are computed. The findings of the seismic analyses demonstrate the effectiveness of the proposed decomposition levels.

Remaining Useful Life of the Rolling Bearings Prediction Method Based on Transfer Learning Integrated with CNN-GRU-MHA

To accurately predict the remaining useful life (RUL) of rolling bearings under limited data and fluctuating load conditions, we propose a new method for constructing health indicators (HI) and a transfer learning prediction framework, which integrates Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and Multi-head attention (MHA). Firstly, we combined Convolutional Neural Networks (CNN) with Gated Recurrent Units (GRU) to fully extract temporal and spatial features from vibration signals. Then, the Multi-head attention mechanism (MHA) was added for weighted processing to improve the expression ability of the model. Finally, a new method for constructing Health indicators (HIs) was proposed in which the noise reduction and normalized vibration signals were taken as a HI, the L1 regularization method was added to avoid overfitting, and the model-based transfer learning method was used to realize the RUL prediction of bearings under small samples and variable load conditions. Experiments were conducted using the PHM2012 dataset from the FEMTO-ST research institute and XJTU-SY dataset. Three sets of 12 migration experiments were conducted under three different operating conditions on the PHM2012 dataset. The results show that the average RMSE of the proposed method was 0.0443, indicating high prediction accuracy under variable loads and small sample conditions. Three different operating conditions and two sets of four migration experiments were conducted on the XJTU-SY dataset, and the results show that the average RMSE of the proposed method was 0.0693, verifying the good generalization of the model under variable load conditions. In summary, the proposed HI construction method and prediction framework can effectively reduce the differences between features, with high stability and good generalizability.

Deep Learning for Traffic Prediction and Trend Deviation Identification: A Case Study in Hong Kong

This paper introduces a robust methodology for predicting traffic volume and speed on major strategic routes in Hong Kong by leveraging data from data.gov.hk and utilizing deep learning models. The approach offers predictions from 6 min to 1 h, considering detector reliability. By extracting hidden deep features from historical detector data to establish detector profiles and grouping detectors into clusters based on profile similarities, the method employs a CNN-LSTM prediction model for each cluster. The study demonstrates the model’s resilience to detector failures, with tests conducted across failure rates from 1% to 20%, highlighting its ability to maintain accurate predictions despite random failures. In scenarios without failed detectors, the method achieves favorable performance metrics: MAE, RMSE, and MAPE for traffic volume prediction over the next 6 min stand at 5.17 vehicles/6 min, 7.64 vehicles/6 min, and 14.07%, respectively, while for traffic speed prediction, the values are 3.70 km/h, 6.32 km/h, and 6.33%. Considering a failure rate of approximately 6% in the Hong Kong dataset, in simulated scenarios with 6% failures, the model maintains its predictive accuracy, with average MAE, RMSE, and MAPE for traffic volume prediction at 5.24 vehicles/6 min, 7.81 vehicles/6 min, and 14.21%, and for traffic speed prediction at 3.87 km/h, 6.55 km/h, and 6.68%. However, the limitation of the proposed method is its potential to underperform when predicting rare or unseen scenarios, indicating the need for future research to incorporate additional data sources and methods to enhance predictive performance.

Novel optimized models to enhance performance forecasting of grid-connected PERC PV string operating under semi-arid climate conditions

This study focuses on modeling and forecasting the performance of a grid-connected photovoltaic (PV) string, aiming to enhance the accuracy of output power prediction, which is crucial for the effective management and stability of the electrical grid. While previous research has extensively examined PV modules, there is a notable lack of studies specifically targeting PV string behaviors, especially in harsh climates prevailing in semi-arid regions. This study addresses this gap by investigating the modeling and performance prediction of a Passivated Emitter Rear Cell (PERC) PV string under semi-arid climate conditions using analytical and predictive approaches based on both implicit and explicit models as Single Diode Model (SDM), Das Model (DM), Boutana et al. Model (BM) and Power Law Model (PLM). New mathematical models of DM and BM shape parameters versus temperature and irradiance along with novel analytical formulas giving DM shape parameters as a function of PV metrics have been introduced. The proposed approaches are validated using real measurements of a PERC PV string incorporating 12 JKM405M-72H PV modules operating at Green Energy Park (GEP) research facility in Morocco. The reliability of these approaches is assessed by comparing I-V curves, P-V curves and peak power generated and predicted by SDM, DM, BM and PLM to actual measurements. The comparison reveals that generated and predicted outcomes align well with measured data, with an average value of Normalized Root Mean Square Error (NRMSE) not exceeding 4.16 % for all models throughout the day. The findings indicate that the proposed approaches effectively predict the performance of the PERC PV string, accounting for climatic influences and providing insights into optimizing PV system performance, thereby contributing to improved grid stability in semi-arid regions.

High precision aerodynamic heat prediction method based on data augmentation and transfer learning

Data-driven modeling methods have become one of the main technologies for predicting aerodynamic heat in hypersonic conditions. However, due to the limitations of wind tunnel experimental conditions, the spatial distribution of aerothermal wind tunnel experimental data is often sparse, and the sample size is relatively small. Furthermore, there is a lack of direct correlation in the aerodynamic heat distribution data among different shapes of vehicles, which poses challenges for constructing high-performance data-driven aerodynamic heat prediction models. To address these issues, this paper proposes a high-precision aerodynamic heat modeling and prediction method based on data augmentation and transfer learning. First, integrating the concept of data fusion, we propose to enhance the sparse aerothermal wind tunnel experimental data by using deep neural networks and introducing low-precision numerical computation results. Next, based on the close physical correlation between boundary layer outer edge information and wall surface aerodynamic heat, we construct the aerodynamic heat prediction model ED-ResNet using a double-series residual neural network. Finally, by fine-tuning the ED-ResNet model for transfer learning, high-precision predictions of aerothermal wind tunnel experimental results for different shaped vehicles are achieved under small sample conditions. Verification using hypersonic double-ellipsoid, blunt cone, and blunt bicone shows that after data augmentation, the prediction error of the aerodynamic heat prediction model is significantly reduced to 1/3 of that when data augmentation is not used. Moreover, through transfer learning, the model effectively leverages existing hypersonic double-ellipsoid aerothermal wind tunnel experimental data to achieve high-precision predictions of aerodynamic heat distribution for blunt cone and blunt double cone under different incoming flow conditions, with normalized root mean square error(NRMSE) maintained below 10%.

PAOFCDN: A novel method for predictive analysis of solar irradiance

Solar photovoltaic power is the most feasible Renewable Energy Source (RES) for Pakistan, due to ample sunlight availability throughout the year. Since solar photovoltaic power is primarily dependent on solar irradiance, forecasting of solar irradiance is essential for reliable, secure and effective incorporation of solar photovoltaic power in power systems. Considering the importance of solar irradiance forecasting, in this study, predictive analysis of Islamabad’s solar irradiance is performed by using a novel proposed model named as Pelican Algorithm-based Optimized Fully-Connected Deep Network (PAOFCDN). The initial weights of Fully-Connected Deep Network (FCDN) are optimized through an effective optimization technique known as Pelican optimization. The accuracy of the optimized network PAOFCDN is enhanced many fold as compared to the FCDN network trained with randomly initialized weights. The inherent issue of poor generalization in FCDN is also resolved by optimization. The superior performance of PAOFCDN is evident from its comparative evaluation with existing benchmark methods, i.e., Long Short-Term Memory (LSTM), Support Vector Regression (SVR), Least Square Boosting (LSBoost) and standard FCDN. PAOFCDN achieves the least Normalized Root Mean Square Error (NRMSE) of 0.0503 as compared to 0.1179 of LSTM, 0.1256 of FCDN, and 0.2992 of SVR and LSBoost. The proposed model is applied to three real-world solar irradiance datasets having different resolutions of 10-minutes, hourly and daily. This study took the initiative of performing predictions on three datsets having multiple resolutions in perspective of south asia.

A hybrid deep learning-based short-term forecast model for ionospheric foF2 in East Asia region

The short-term forecast of the critical frequency of the ionospheric F2 layer (foF2) is particularly challenging due to its nonlinear and non-stationary characteristics. To improve the short-term forecast accuracy of foF2, we propose a short-term hybrid deep learning model that integrates the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and long short-term memory (LSTM) algorithms. The proposed model leverages the ICEEMDAN decomposition algorithm to expand foF2 time series data into multi-dimensional space and utilizes the LSTM algorithm to preserve long-term data information, capturing detailed changes in foF2 parameters and obtaining internal information about foF2. The 2014 and 2017 foF2 data from four observation stations located in Mohe, Beijing, Jeju, and Kokubunji in the mid-latitudes of East Asia were utilized for modeling and 1-hour forecast analysis. The forecast values of the proposed model align closely with the measured data, exhibiting minimal fluctuation in forecast error and effectively tracking the changing trend of foF2. The average root mean square error (RMSE) across the four stations in 2014 is 0.40 MHz, with a relative RMSE (RRMSE) of 7.11 % and a coefficient of determination (R2) of 0.98. The average RMSE in 2017 is 0.43 MHz, RRMSE is 7.68 %, and R2 is 0.92. The proposed model reveals significant improvements in the forecast accuracy of the comparison model compared with the CCIR, URSI, NN, and LSTM models. Especially compared with the CCIR and URSI models in the International Reference Ionosphere (IRI), demonstrating superior real-time tracking capability for foF2 and effectively addressing the challenge of poor short-term forecast accuracy observed in the IRI model.

A Short-term Vehicle Speed Prediction Approach considering Dynamic Traffic Scene

Vehicle speed serves as a crucial indicator in traffic flow efficiency and safety evaluation. Previous researches typically utilize the time-series data of target vehicle speed or the relative speed and distance between the target and leading vehicles as speed prediction model input, which are suitable for stable and single-vehicle scenes. However, the surrounding traffic scenes dynamically changes during the vehicle’s driving process, the driver needs to allocate attention to the key scene vehicles in the surrounding traffic scenes and adjust the current speed based on these vehicles’ behavior. Therefore, this paper proposes a short-time vehicle speed prediction model considering the dynamic traffic scenes. A dynamic grid method is adopted to divide the target vehicle’s surrounding areas into left front, front, and right front regions. And key scene vehicles are identified based on each region’s state. Then the relative distance, speed, and acceleration data between the target vehicle and key scene vehicles are selected as model inputs. Vehicle trajectory data from the South-North No. 1 Expressway in Hefei and Ranger optimized Temporal Convolutional Network (TCN) are utilized to train the speed prediction model. Experiment results indicated that under 1s, 2s, and 3s prediction lengths, the model’s average Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) were 0.430 and 0.271 respectively. Compared to the speed prediction model considering the static traffic scenes, the average RMSE and MAE for the three prediction lengths have been reduced by 49.15% and 56.02% respectively.

Dynamic mode decomposition and short-time prediction of PM2.5 using the graph Neural Koopman network

Analyzing and accurately predicting the spatiotemporal dynamics of PM2.5 remain challenging. The existing spatiotemporal prediction approaches are associated with high model complexity and limited interpretability. Conventional methods combining Koopman theory and deep learning often neglect spatial correlations in spatiotemporal data. This study used the hourly PM2.5 dataset of the Beijing-Tianjin-Hebei region to reveal its spatiotemporal hierarchy using Koopman mode decomposition to identify the key dynamic modes. Furthermore, a Spatial Physics Constrained Learning (SPCL) model utilizing a graph representation learning method was proposed to combine the graph topological information of the PM2.5 spatial features with the Koopman feature function. The results showed that PM2.5 has growth, decay, and oscillation modes as well as daily, weekly, monthly, and yearly periods. SPCL achieved mean absolute error, root mean square error (RMSE), correlation r, and index of agreement values of 9.678, 13.922, 0.864, and 0.921, respectively. The average RMSE at 12 h improved by 16.1%, 12.7%, 0.9%, and 3.5% compared with using Long short-term Memory, Graph Convolutional Networks and Long Short-Term Memory Networks, Spatio-Temporal Graph Convolutional Networks, and Dynamic Spatiotemporal Graph Convolution Network, respectively. By discretizing the neural network hidden layers, the explanatory key of PM2.5 modes was elucidated, which demonstrated enhanced stability.

CGAformer: Multi-scale feature Transformer with MLP architecture for short-term photovoltaic power forecasting

Accurately predicting the output power of photovoltaic (PV) systems is an effective means to ensure the reliable and economical operation of grid-connected PV systems. Aiming at the characteristics of PV power generation such as strong volatility, high intermittency and obvious periodicity, a hybrid model named CGAformer based on One-Dimensional Convolutional Neural Networks (CNN1D), Global Additive Attention (GADAttention), and Auto-Correlation is proposed for short-term PV power generation prediction. The model uses CNN1D to extract local features and obtains global weights by improving the GADAttetion obtained by additive attention. Auto-Correlation integrates local features and global weights and identifies repeated patterns in the sequence to obtain highly coupled multi-scale features, and finally generates the final prediction results through Multilayer Perceptron (MLP). In order to verify the effectiveness of the model, this paper uses a historical dataset from a PV system located in Uluru, Australia for sufficient experiments. In the comparative experiments, The overall average RMSE and MAE of CGAformer are improved by 6.82% and 20.46% respectively compared with long short-term memory (LSTM). In addition, ablation experiments and seasonal analysis are used to verify the effectiveness of the model and its excellent generalization ability for different seasons.

Development of a three-dimensional variational data assimilation system for 137Cs based on WRF-Chem model and applied to the Fukushima nuclear accident

Abstract Nuclear explosions and accidents release large amounts of radionuclides that harm human health and the environment. Accurate forecasting of nuclide pollutants and assessment of the ramifications of nuclear incidents are necessary for the emergency response and disaster assessment of nuclide pollution. In this study, we developed a three-dimensional variational (3Dvar) system to assimilate 137Cs based on the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) model. The distribution of 137Cs after the Fukushima nuclear accident in Japan on 15 March 2011 was analysed. The 137Cs background field at 06:00 UTC was assimilated using a 3Dvar system and surface observational data to optimise the 137Cs analysis field. Compared with the background field, the root mean square error (RMSE) and mean bias in the 137Cs analysis field decreased by 98% and 94%, respectively. The average fraction of predictions within factors of 2 (FAC2), 5 (FAC5), and 10 (FAC10) increased from 0.67, 0.72, and 0.72 to 0.90, 1.00, and 1.00, respectively. This substantial enhancement indicated the effectiveness of the 3DVar system in mitigating the uncertainty associated with the background field. Two 12 h forecast experiments were conducted to gauge the advancement in 137Cs forecasting facilitated by data assimilation (DA). The control experiment was conducted without DA, whereas the assimilation experiment was conducted with DA. Compared with the control experiment, the average FAC2, FAC5, and FAC10 in the assimilation experiment increased by 28%, 30%, and 29%, respectively. The average RMSE decreased by 33%. The mean bias and correlation coefficient increased by 41% and 36%, respectively. These results indicated that the 3Dvar method improves the forecast accuracy of 137Cs concentration.

Effect of Ultrasound Conditions on Germination and Drying Characteristics of Bengal Gram (Cicer arietinum L.) and Determination of Physicochemical and Functional Properties of Flours

ABSTRACTThis study aims to investigate the effect of ultrasound conditions on the germination kinetics and drying characteristics of a germinated Bengal gram at different drying temperatures. Ultrasound treatment was given to the Bengal gram seeds at two different conditions, that is, before (US) and after soaking (SU), which was then followed by germination. This study also determines mass transfer parameters at drying temperatures of 45°C, 55°C, and 65°C and assesses the influence on the physicofunctional characteristics of a germinated Bengal gram. The germination rate behavior was effectively predicted using a zero‐order kinetic model with the highest R2 value of 0.7974–0.8857 in each nontreated (S) and treated (SU, US) Bengal gram seed, respectively. This study showed that the ultrasound treatment effectively enhanced the germination rate in both conditions, and the highest germination rate was found in pretreated ultrasound Bengal gram samples. The logarithmic thin layer drying model, with the highest average R2 of 0.9954 and the lowest average RMSE value of 0.0160, is the best‐fitted model to predict the changes in moisture ratio in both treated and nontreated conditions. The moisture diffusivity values at drying temperatures ranging from 45°C to 65°C were found in treated (US, SU) and nontreated (S) germinated samples ranging from 3.34 × 10−8 to 4.03 × 10−8 m2/s, 1.30 × 10−8 to 2.01 × 10−8 m2/s, and 6.6 × 10−9 to 8.06 × 10−9 m2/s, respectively. The protein content increased in the ultrasound‐treated sample between 12.37% and 13.50%. The solubility ranged from 8% to 10.36% throughout the treated and nontreated germinated Bengal gram flour. This study provides a unified approach to utilizing ultrasound‐treated germinated Bengal gram seeds as an alternative option to develop a functional product with better nutritional and functional properties.