Normalized Root Mean Square Error Values Research Articles

Dynamic Harvest Index Estimation of Winter Wheat Based on UAV Hyperspectral Remote Sensing Considering Crop Aboveground Biomass Change and the Grain Filling Process

The crop harvest index (HI) is of great significance for research on the application of crop variety breeding, crop growth simulation, crop management in precision agriculture and crop yield estimation, among other topics. To obtain spatial information on the crop dynamic HI (D-HI), taking winter wheat as the research object and fully considering the changes in crop biomass and the grain filling process from the flowering period to the maturity period, the dynamic fG (D-fG) parameter was estimated as the ratio between the aboveground biomass accumulated in different growth periods, from the flowering stage to the maturity stage, and the aboveground biomass in the corresponding periods. Based on the D-fG parameter estimation using unmanned aerial vehicle (UAV) hyperspectral remote sensing data, a technical method for obtaining spatial information on the winter wheat D-HI was proposed and the accuracy of the proposed method was verified. A correlation analysis was performed between the normalized difference spectral index (NDSI), which was calculated using pairs of any two bands of the UAV hyperspectral spectrum, and the measured D-fG. Based on this correlation analysis, the center of gravity of the local maximum region of R2 was used to determine the sensitive band center to accurately estimate D-fG. On this basis, remote sensing estimation of the D-fG was realized by using the NDSI constructed by the sensitive hyperspectral band centers. Finally, based on the D-fG remote sensing parameters and the D-HI estimation model, spatial information on the D-HI of winter wheat was accurately obtained. The results revealed five pairs of sensitive hyperspectral band centers (i.e., λ(476 nm, 508 nm), λ(444 nm, 644 nm), λ(608 nm, 788 nm), λ(724 nm, 784 nm) and λ(816 nm, 908 nm)) for D-fG estimation, and the results of the D-fG remote sensing estimation showed high precision. The root mean square error (RMSE) was between 0.0436 and 0.0604, the normalized RMSE (NRMSE) was between 10.31% and 14.27% and the mean relative error (MRE) was between 8.28% and 12.55%. In addition, the D-fG parameter estimation, using the NDSI constructed by the above five sensitive remote sensing band centers, yielded highly accurate spatial D-HI information with an RMSE between 0.0429 and 0.0546, an NRMSE between 9.87% and 12.57% and an MRE between 8.33% and 10.90%. The D-HI estimation results based on the hyperspectral sensitive band centers λ(724 nm, 784 nm) had the highest accuracy, with RMSE, NRMSE and MRE values of 0.0429, 9.87% and 8.33%, respectively. The proposed method of acquiring spatial information on the winter wheat D-HI in this study was shown to be feasible, and it might provide a technical reference toward developing satellite-based indices to monitor large-scale crop HI information.

Evaluation of remote sensing-based evapotranspiration products at low-latitude eddy covariance sites

Remote sensing-based evapotranspiration (ET) products have been evaluated primarily using data from northern middle latitudes; therefore, little is known about their performance at low latitudes. To address this bias, an evaluation dataset was compiled using eddy covariance data from 40 sites between latitudes 30° S and 30° N. The flux data were obtained from the emerging network in Mexico (MexFlux) and from openly available databases of FLUXNET, AsiaFlux, and OzFlux. This unique reference dataset was then used to evaluate remote sensing-based ET products in environments that have been underrepresented in earlier studies. The evaluated products were: MODIS ET (MOD16, both the discontinued collection 5 (C5) and the latest collection (C6)), Global Land Evaporation Amsterdam Model (GLEAM) ET, and Atmosphere-Land Exchange Inverse (ALEXI) ET. Products were compared with unadjusted fluxes (ETorig) and with fluxes corrected for the lack of energy balance closure (ETebc). Three common statistical metrics were used: coefficient of determination (R2), root mean square error (RMSE), and percent bias (PBIAS). The effect of a vegetation mismatch between pixel and site on product evaluation results was investigated by examining the relationship between the statistical metrics and product-specific vegetation match indexes. Evaluation results of this study and those published in the literature were used to examine the performance of the products across latitudes. Differences between the MOD16 collection 5 and 6 datasets were generally smaller than differences with the other products. Performance and ranking of the evaluated products depended on whether ETorig or ETebc was used. When using ETorig, GLEAM generally had the highest R2, smallest PBIAS, and best RMSE values across the studied land cover types and climate zones. Neither MOD16 nor ALEXI performed consistently better than the other. When using ETebc, none of the products stood out in terms of both low bias and strong correlations. The use of ETebc instead of ETorig affected the biases more than the correlations. The product evaluation results showed no significant relationship with the degree of match between the vegetation at the pixel and site scale. The latitudinal comparison showed tendencies of lower R2 (all products) but better PBIAS and normalized RMSE values (MOD16 and GLEAM) for forests at low latitudes than for forests at northern middle latitudes. For non-forest vegetation, the products showed no clear latitudinal differences in performance.

Developing pedotransfer functions using Sentinel-2 satellite spectral indices and Machine learning for estimating the surface soil moisture

To estimate the surface soil moisture (SM) using a combination of new spectral indices and methods of Random Forrest (RF) and Multiple Linear Regression (MLR), 11 pedotransfer functions (PTF1-11) were developed by combining basic soil properties (clay, silt/sand, and bulk density) and spectral indices of Sentinel-2 satellite. In this study, 124 surface soil samples were randomly taken from three regions including Telo in Tehran province, Ivaneki in Semnan province, and Borujerd in Lorestan province, Iran. The results showed that the accuracy of the RF method was considerably higher compared to the MLR method. The SM was better estimated using water spectral indices (such as Normalized Difference Water Index (NDWI) and Surface Water Capacity Index (SWCI)), along with the basic properties of soil as inputs of PTF7. In the training and testing steps, the Root Mean Square Error (RMSE) decreased from 0.041 and 0.05 (cm3 cm−3) in PTF1 to 0.028 and 0.039 in PTF7, respectively. The average values of RMSE, Akaike Information Criterion (AIC), coefficient of determination (R2) and Relative Improvement (RI) of the RF method were 0.028 (cm3 cm−3), −559, 0.79 and 0.001, and 0.038 (cm3 cm−3), −279, 0.73 and −0.006 for the training and testing steps of all PTFs, respectively. While, these values for the MLR method were 0.032 (cm3 cm−3), −542, 0.69 and 0.0003, and 0.043 (cm3 cm−3), −269, 0.63 and 0.002 for the training and testing steps, respectively. Due to the low values of MBE, it was possible to disregard the overestimation of the results. The evaluation of results for predictor importance indicated that among the basic properties, clay percent has a significant effect on estimation of SM. These results show that spectral indices alone are not suitable estimators for SM estimation. It suggests using basic soil properties and spectral indices to estimate the SM.

Iterative bicluster-based Bayesian principal component analysis and least squares for missing-value imputation in microarray and RNA-sequencing data.

Microarray and RNA-sequencing (RNA-seq) techniques each produce gene expression data that can be expressed as a matrix that often contains missing values. Thus, a process of missing-value imputation that uses coherence information of the dataset is necessary. Existing imputation methods, such as iterative bicluster-based least squares (bi-iLS), use biclustering to estimate the missing values because genes are only similar under correlative experimental conditions. Also, they use the row average to obtain a temporary complete matrix, but the use of the row average is considered to be a flaw. The row average cannot reflect the real structure of the dataset because the row average only uses the information of an individual row. Therefore, we propose the use of Bayesian principal component analysis (BPCA) to obtain the temporary complete matrix instead of using the row average in bi-iLS. This alteration produces new missing values imputation method called iterative bicluster-based Bayesian principal component analysis and least squares (bi-BPCA-iLS). Several experiments have been conducted on two-dimension independent gene expression datasets, which are microarray (e.g., cell-cycle expression dataset of yeast saccharomyces cerevisiae) and RNA-seq (gene expression data from schizosaccharomyces pombe) datasets. In the case of the microarray dataset, our proposed bi-BPCA-iLS method showed a significant overall improvement in the normalized root mean square error (NRMSE) values of 10.6% from the local least squares (LLS) and 0.6% from the bi-iLS. In the case of the RNA-seq dataset, our proposed bi-BPCA-iLS method showed an overall improvement in the NRMSE values of 8.2% from the LLS and 3.1% from the bi-iLS. The additional computational time of bi-BPCA-iLS is not significant compared to bi-iLS.

Comparing linear and non-linear data-driven approaches in monthly river flow prediction, based on the models SARIMA, LSSVM, ANFIS, and GMDH.

River flow variations directly affect the hydro-climatological, environmental, and ecological characteristics of a region. Therefore, an accurate prediction of river flow can critically be important for water managers and planners. The present study aims to compare different data-driven models in predicting monthly flow. Two river catchments located in the Guilan province in Iran, where rivers play an essential role in agricultural productions (mainly rice), are studied. The monthly river flow dataset was provided by Guilan Regional Water Authority during 1986-2015. The models are derived from two different numerical types of stochastic and machine learning (ML) models. The stochastic model is seasonal autoregressive integrated moving average (SARIMA), and the MLs are least square support vector machine (LSSVM), adaptive neuro-fuzzy inference system (ANFIS), and group method of data handling (GMDH). The inputs were selected by autocorrelation and partial autocorrelation functions (ACF and PACF) from the flow rates of the previous months. The data was divided into 75% of training and 25% of testing phases, and then the mentioned models were implemented. Predictions were evaluated by the criteria of root mean square error (RMSE), normalized RMSE (NRMSE), and Nash Sutcliff (NS) coefficient. According to the calculated values of different criteria during the test phase, RMSE = 1.138 cms, NRMSE = 0.109, and NS = 0.826, it can be concluded that the SARIMA model was superior to its ML competitors. Among the ML models, GMDH had the best performance (by RMSE = 1.290 cms, NRMSE = 0.124, and NS = 0.777) because it has more optimization parameters and sample space for network make-up. The models were also evaluated in hydrological drought conditions of both rivers. It was resulted that the rivers' flow can be well predicted in drought conditions by using these models, especially the SARIMA stochastic model. According to the NRMSE values (ranged between 0.1 and 0.2), the accuracy of predictions is evaluated in the appropriate range, and the present study shows promising results of the current approaches. Consequently, a comparison between the performance of linear stochastic models and complex black-box MLs, reveals that linear stochastic models are more suitable for the current region's monthlyriver flow prediction.

Freundlich and Langmuir Isotherm Studies of Phosphorus Sorption unto Soils Derived from Basement Complex Rock, Alluvium, Coastal Plain Sand and Shale Parent Materials

To provide information on best model to predict Phosphorus (P) Sorption unto Soils derived from Basement Complex Rock, Alluvium, Coastal Plain Sand and Imo Shale Parent Materials in 3 states of Nigeria. Completely randomized design was used to collect surface soil samples in 3 replications from 4 locations in Nigeria.
 Samples were collected from Idanre, Koko, NIFOR and Uhonmora in Ondo, Delta and Edo states Nigeria, laboratory analysis was carried out in the Central analytical laboratory of Nigerian Institute for Oil-Palm Research (NIFOR) Benin City, Nigeria between march 2016 and September 2017.
 Soil samples were equilibrated in 25 ml of 0.01 M CaCl2 containing various concentration of P as KH2PO4 to give 0, 50, 100, 150, 200 and 250 mg/L P for 24 hours (h) at room temperature 25 ± 2oC. 3 drops of CHCl3 was added to inhibit P mineralization. The suspension was shaken for 24 h on a reciprocating mechanical shaker, centrifuged at 7000 rpm After equilibration, decanted and P determined using spectrophotometer. The sorption data were fitted to linear Freundlich and Langmuir sorption isotherm.
 Considering the Freundlich model, P adsorption capacity (a) and P sorption energy (n) was highest in soils B (1400 mg kg-1) and (2.806 L kg-1) respectively. The Freundlich model fitted better to the data obtained with average root mean square error (RMSE) and R2 value of 0.69 and 0.951 respectively, as against average RMSE and R2 value of 1.60 and 0.883 respectively obtained from Langmuir model.
 The sorption data fitted well to Freundlich and Langmuir isotherms of which Freundlich Adsorption model was found to be better based on lowest RMSE (0.69) and highest regression (R2 = 0.951) value. Freundlich model should be adopted to determine P sorption characteristics of the soils studied. These predictors, however, need further works to validate reliability.

Improvement of megavoltage computed tomography image quality for adaptive helical tomotherapy using cycleGAN-based image synthesis with small datasets.

Megavoltage computed tomography (MVCT) offers an opportunity for adaptive helical tomotherapy. However, high noise and reduced contrast in the MVCT images due to a decrease in the imaging dose to patients limits its usability. Therefore, we propose an algorithm to improve the image quality of MVCT. The proposed algorithm generates kilovoltage CT (kVCT)-like images from MVCT images using a cycle-consistency generative adversarial network (cycleGAN)-based image synthesis model. Data augmentation using an affine transformation was applied to the training data to overcome the lack of data diversity in the network training. The mean absolute error (MAE), root-mean-square error (RMSE), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM) were used to quantify the correction accuracy of the images generated by the proposed algorithm. The proposed method was validated by comparing the images generated with those obtained from conventional and deep learning-based image processing method through non-augmented datasets. The average MAE, RMSE, PSNR, and SSIM values were 18.91 HU, 69.35 HU, 32.73dB, and 95.48 using the proposed method, respectively, whereas cycleGAN with non-augmented data showed inferior results (19.88 HU, 70.55 HU, 32.62dB, 95.19, respectively). The voxel values of the image obtained by the proposed method also indicated similar distributions to those of the kVCT image. The dose-volume histogram of the proposed method was also similar to that of electron density corrected MVCT. The proposed algorithm generates synthetic kVCT images from MVCT images using cycleGAN with small patient datasets. The image quality achieved by the proposed method was correspondingly improved to the level of a kVCT image while maintaining the anatomical structure of an MVCT image. The evaluation of dosimetric effectiveness of the proposed method indicates the applicability of accurate treatment planning in adaptive radiation therapy.

Object-based modelling and accuracy assessment of Gordion tumuli using very high-resolution stereo satellite images

ABSTRACT Very high-resolution (VHR) stereo satellite images provide detailed spatial and spectral information for detecting and monitoring artificial or natural objects. This information can also be used for object-based change detection. This study aims to reveal the potential of three-dimensional (3D) information generated from VHR stereo satellite images to determine the physical changes that may occur on Gordion tumuli. For this purpose, point cloud data were generated from WorldView-3 and KOMPSAT-3 stereo satellite images. The accuracy of these datasets was tested using 161 check points (CPs). On the other hand, digital surface models (DSMs) generated from VHR stereo satellite images were used for object-based modelling of tumuli. The vertical accuracy of these DSMs was tested by comparing them with reference DSMs extracted from Unmanned Aerial System (UAS) images. As a result of the vertical accuracy assessment of the point cloud data, 0.16 m and 0.53 m root mean square error (RMSE) values were obtained for WorldView-3 and KOMPSAT-3 data, respectively. The average RMSE values obtained as a result of the model-based comparison of DSMs are 0.28 m for WorldView-3 and 0.57 m for KOMPSAT-3. These results demonstrate that the sub-metre level physical changes occurring on tumuli can be determined by VHR stereo satellite images.

Comparative Study of Univariate and Multivariate Long Short-Term Memory for Very Short-Term Forecasting of Global Horizontal Irradiance

Accurate global horizontal irradiance (GHI) forecasting is crucial for efficient management and forecasting of the output power of photovoltaic power plants. However, developing a reliable GHI forecasting model is challenging because GHI varies over time, and its variation is affected by changes in weather patterns. Recently, the long short-term memory (LSTM) deep learning network has become a powerful tool for modeling complex time series problems. This work aims to develop and compare univariate and several multivariate LSTM models that can predict GHI in Guntur, India on a very short-term basis. To build the multivariate time series models, we considered all possible combinations of temperature, humidity, and wind direction variables along with GHI as inputs and developed seven multivariate models, while in the univariate model, we considered only GHI variability. We collected the meteorological data for Guntur from 1 January 2016 to 31 December 2016 and built 12 datasets, each containing variability of GHI, temperature, humidity, and wind direction of a month. We then constructed the models, each of which measures up to 2 h ahead of forecasting of GHI. Finally, to measure the symmetry among the models, we evaluated the performances of the prediction models using root mean square error (RMSE) and mean absolute error (MAE). The results indicate that, compared to the univariate method, each multivariate LSTM performs better in the very short-term GHI prediction task. Moreover, among the multivariate LSTM models, the model that incorporates the temperature variable with GHI as input has outweighed others, achieving average RMSE values 0.74 W/m2–1.5 W/m2.

Fully connected deep network: An improved method to predict TOC of shale reservoirs from well logs

Shale oil and shale gas are important unconventional resources. As the total organic carbon (TOC) of shales is related to the generation potentiality of hydrocarbons, a model of TOC prediction can improve development efficiency and reduce cost. However, the intelligent prediction of TOC is not widely studied. This study works on hydrocarbon-rich shales from four basins in China and Canada. ΔLogR, support vector regression (SVR), single-layer artificial neural network (ANN), and fully connected deep network (FCDN) were built to predict the TOC of the studied shales using resistivity, sonic, density, and gamma ray logs. The predicted and geochemical-measured TOC were compared. The correlation of determination (R2) and normalized-root-mean-square-error (NRMSE) were used to evaluate the models. The results indicate that ΔLogR, SVR, and ANN have low R2 and high NRMSE values and are unsatisfactory to predict TOC. These models greatly underestimate the TOC, which provide evident deviations of predicted TOC from their true values.Conversely, FCDN provides accurate TOC predictions. The optimum results are obtained from an eight-layer network that consists of one input layer, six hidden layers, and one output layer. Values of R2 and NRMSE are 0.89 and 0.044, suggesting that the TOC predictions using the FCDN are close to their true values. The FCDN greatly outperforms other models by extracting the complicate relationships between well logs and TOC values. The results of this study may suggest the great potentiality of deep learning techniques in the evaluation of unconventional resources.

Experimental Validation of a Thermo-Electric Model of the Photovoltaic Module under Outdoor Conditions

An operating temperature of the photovoltaic (PV) module greatly affects performance and its lifetime. Therefore, it is essential to evaluate operating temperature of the photovoltaic module in different weather conditions and how it affects its performance. The primary objective of this paper is to present a dynamic thermo-electric model for determining the temperature and output power of the photovoltaic module. The presented model is validated with field measurement at the Institute of Energy Technology, Faculty of Energy Technology, University of Maribor, Slovenia. The presented model was compared with other models in different weather conditions, such as clear, cloudy and overcast. The evaluation was performed for the operating temperature and output power of the photovoltaic module using Root-Mean-Square-Error (RMSE) and Mean-Absolute-Error (MAE). The average RMSE and MAE values are 1.75 °C and 1.14 °C for the thermal part and 20.34 W and 10.97 W for the electrical part.

Prediction of water-adulteration within honey by air-line de-embedding waveguide measurements

An efficient de-embedding air-line microwave method has been proposed for accurate relative complex permittivity, εr=εr′−iεr′′, measurement of water-adulteration level within honey. It could be effectively applied to eliminate the errors arising from usage of imperfect calibration standards because it bypasses the requirement of these standards. Its accuracy is improved by utilizing the unitary and similarity properties of a passive two-port network, and then is compared with the accuracy of a calibration-dependent method present in the literature by using normalized root-mean-square-error (N-RMSE) values of εr′ and εr′′ of distilled water, in reference to the Debye value. From this comparison, it is observed that N-RMSE values calculated for εr′ (and εr′′) by using this calibration-dependent method and the (improved) proposed method are, respectively, around 0.1955 (0.1002) and 0.1962 (1.1067), indicating a good agreement between them. After validation the proposed de-embedding method using distilled water measurements, tested pure honey was adulterated with distilled water by different percentage values δ ranging from 1% to 10% in 1% increments. It is observed that the maximum distance between extracted εr′ (or εr′′) values of adulterated honey by the applied calibration-dependent method and the proposed method is less than 2%. Afterward, an empirical formula was devised to fit εr′ and εr′′ values from measured εr of water-adulterated honey with various δ levels. It is noted that extracted εr′ is much more better fitted than extracted εr′′, especially for δ≤4. Next, an optimization process is followed to evaluate the frequency for optimum prediction of adulteration levels using the empirical formula based on εr′ or εr′′. It is noticed that optimized δ values using the empirical formula based on εr′ (with an average prediction error of around 0.071 at 4.5 GHz) are superior to optimized δ values using the empirical formula based on εr′′ (with an average prediction error of around 0.085 at 4.2 GHz) for prediction of previously known δ values. Sensitivity and uncertainty analyses were performed to assess and improve the accuracy of the proposed method.

Flow interval prediction based on deep residual network and lower and upper boundary estimation method

Interval prediction is an efficient approach for quantifying the uncertainties of future flow. In this paper, a novel method, based on the combination of a deep residual neural network (ResNet) and lower and upper bound estimation (LUBE), is proposed to forecast future flow and construct prediction intervals. LUBE is proposed by optimizing a LUBE-based objective function and adjusting the type and quantity of residual blocks to design a combined residual network. The final proposed interval prediction model is stResNet-LUBEproposed. The performance of the stResNet-LUBEproposed model is verified using the spatiotemporal dataset of Tunxi, which is a small and medium watershed. The performance of proposed model is mainly evaluated by the root mean square error (RMSE), coefficient of determination (R2), and coverage width-based criterion (CWC). The experimental results show that the average values of RMSE, R2, and CWC of the proposed model are better than those of the spatiotemporal deep learning model stCNN-LUBEproposed, the deep learning model LSTM-LUBEproposed, and the machine learning model MLP-LUBEproposed by 1.881%, 10.574%, and 14.113%; 2.198%, 15.754%, and 18.546%; and 10.572%, 35.907%, and 46.819%, respectively (predict 6-step).

Model-Based Control of Torque and Nitrogen Oxide Emissions in a Euro VI 3.0 L Diesel Engine through Rapid Prototyping

In the present paper, a model-based controller of engine torque and engine-out Nitrogen oxide (NOx) emissions, which was previously developed and tested by means of offline simulations, has been validated on a FPT F1C 3.0 L diesel engine by means of rapid prototyping. With reference to the previous version, a new NOx model has been implemented to improve robustness in terms of NOx prediction. The experimental tests have confirmed the basic functionality of the controller in transient conditions, over different load ramps at fixed engine speeds, over which the average RMSE (Root Mean Square Error) values for the control of NOx emissions were of the order of 55–90 ppm, while the average RMSE values for the control of brake mean effective pressure (BMEP) were of the order of 0.25–0.39 bar. However, the test results also highlighted the need for further improvements, especially concerning the effect of the engine thermal state on the NOx emissions in transient operation. Moreover, several aspects, such as the check of the computational time, the impact of the controller on other pollutant emissions, or on the long-term engine operations, will have to be evaluated in future studies in view of the controller implementation on the engine control unit.

NASA’s MODIS/VIIRS Global Water Reservoir Product Suite from Moderate Resolution Remote Sensing Data

Global reservoir information can not only benefit local water management but can also improve our understanding of the hydrological cycle. This information includes water area, elevation, and storage; evaporation rate and volume values; and other characteristics. However, operational wall-to-wall reservoir storage and evaporation monitoring information is lacking on a global scale. Here we introduce NASA’s new MODIS/VIIRS Global Water Reservoir product suite based on moderate resolution remote sensing data—the Moderate Resolution Imaging Spectroradiometer (MODIS), and the Visible Infrared Imaging Radiometer Suite (VIIRS). This product consists of 8-day (MxD28C2 and VNP28C2) and monthly (MxD28C3 and VNP28C3) measurements for 164 large reservoirs (MxD stands for the product from both Terra (MOD) or Aqua (MYD) satellites). The 8-day product provides area, elevation, and storage values, which were generated by first extracting water areas from surface reflectance data and then applying the area estimations to the pre-established Area–Elevation (A–E) relationships. These values were then further aggregated to monthly, with the evaporation rate and volume information added. The evaporation rate and volume values were calculated after the Lake Temperature and Evaporation Model (LTEM) using MODIS/VIIRS land surface temperature product and meteorological data from the Global Land Data Assimilation System (GLDAS). Validation results show that the 250 m area classifications from MODIS agree well with the high-resolution classifications from Landsat (R2 = 0.99). Validation of elevation and storage products for twelve Indian reservoirs show good agreement in terms of R2 values (0.71–0.96 for elevation, and 0.79–0.96 for storage) and normalized root-mean-square error (NRMSE) values (5.08–19.34% for elevation, and 6.39–18.77% for storage). The evaporation rate results for two reservoirs (Lake Nasser and Lake Mead) agree well with in situ measurements (R2 values of 0.61 and 0.66, and NRMSE values of 16.25% and 21.76%). Furthermore, preliminary results from the VIIRS reservoir product have shown good consistency with the MODIS based product, confirming the continuity of this 20-year product suite. This new global water reservoir product suite can provide valuable information with regard to water-sources-related studies, applications, management, and hydrological modeling and change analysis such as drought monitoring.

A Novel Short-Term Photovoltaic Power Forecasting Approach based on Deep Convolutional Neural Network

ABSTRACT In this study, a novel photovoltaic power forecasting system that utilizes a deep Convolutional Neural Network (CNN) structure and an input signal decomposition algorithm is proposed. The proposed CNN architecture extracts deep features to forecast short-term power using transfer learning-based AlexNet. The historical power, solar radiation, wind speed, and temperature data are selected as the input. The signal decomposition algorithm called Empirical Mode Decomposition (EMD) is utilized to decompose the historical power signal into sub-components. In order to extract deep features, all input parameters are converted to 2D feature maps and feed to the input of the CNN. The experiments are realized on a grid-tied Photovoltaic Power Plant (PVPP) that has 1000 kW installed capacity located in Turkey. The experiments are performed under four weather conditions as partial cloudy, cloudy-rainy, heavy-rainy, and sunny days to show the effectiveness of the proposed method. The obtained results are compared with the benchmark regression algorithms. When the results are analyzed, the proposed method gives the highest Correlation Coefficient (R) and the lowest Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and SMAPE values under all horizons and weather conditions. For 1-h to 5-h ahead, the average R values of the proposed method are obtained as 97.28%, 95.77%, 94.49%, 93.61%, and 92.62%, respectively. The average RMSE values are observed as 4.90%, 6.30%, 7.50%, 8.00%, and 9.17% for 1-h to 5-h ahead. The experimental results confirm that the proposed method outperforms the conventional regression algorithms and reveals effective results with its competitive performance.

Fast online-customized (FOCUS) parallel transmission pulses: A combination of universal pulses and individual optimization.

To mitigate spatial flip angle (FA) variations under strict specific absorption rate (SAR) constraints for ultra-high field MRI using a combination of universal parallel transmit (pTx) pulses and fast subject-specific optimization. Data sets consisting of B0 , maps, and virtual observation point (VOP) data were acquired from 72 subjects (study groups of 48/12 healthy Europeans/Asians and 12 Europeans with pathological or incidental findings) using an 8Tx/32Rx head coil on a 7T whole-body MR system. Combined optimization values (COV) were defined as combination of spiral-nonselective (SPINS) trajectory parameters and an energy regularization weight. A set of COV was optimized universally by simulating the individual RF pulse optimizations of 12 training data sets (healthy Europeans). Subsequently, corresponding universal pulses (UPs) were calculated. Using COV and UPs, individually optimized pulses (IOPs) were calculated during the sequence preparation phase (maximum 15 s). Two different UPs and IOPs were evaluated by calculating their normalized root-mean-square error (NRMSE) of the FA and SAR in simulations of all data sets. Seven additional subjects were examined using an MPRAGE sequence that uses the designed pTx excitation pulses and a conventional adiabatic inversion. All pTx pulses resulted in decreased mean NRMSE compared to a circularly polarized (CP) pulse (CP = ~28%, UPs = ~17%, and IOPs = ~12%). UPs and IOPs improved homogeneity for all subjects. Differences in NRMSE between study groups were much lower than differences between different pulse types. UPs can be used to generate fast online-customized (FOCUS) pulses gaining lower NRMSE and/or lower SAR values.

A Novel Prediction Error-Based Power Forecasting Scheme for Real PV System Using PVUSA Model: A Grey Box-Based Neural Network Approach

This paper presents a prediction error-based power forecasting (PEBF) method for a Photovoltaic (PV) system, using Photovoltaics for Utility Scale Applications (PVUSA) model based grey box neural network (GBNN). First, the differential equation based PVUSA model is transformed into a neural network. In the proposed PEBF scheme, the neural network is set to train whenever the difference between predicted and output powers increases from a certain threshold defined based on system dynamics and requirements. The unique design of the PVUSA model based grey box neural network takes far less training time than usual black-box neural network based models. This gives the proposed prediction scheme an advantage of updating the prediction model parameters from frequent training of neural networks with the change in metrological variables. The effectiveness of the proposed prediction scheme is demonstrated by a real case study regarding a 20MW grid-connected PV system located in Dongying city of Shandong province China. To evaluate the efficiency of the developed scheme, different assessment metrics, mean absolute error (MAE), root mean square error (RMSE), weighted mean absolute error (WMAE) and coefficient of determination ( R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) are applied. The average values of MAE, RMSE, and WMAE were 0.12 %, 0.20% and 0.23% respectively for all cases. The results demonstrate that the proposed scheme predicts the PV power efficiently within the defined error tolerance level, which shows the effectiveness and feasibility of the proposed prediction scheme. The prediction accuracy of the proposed scheme has been compared with the conventional black box neural network models and reveals outperformed performance with respect to prediction accuracy improvement. The proposed prediction scheme will help to balance power production and demands across integrated networks through economic dispatch decisions between the power sources.

Continuous Estimation of Knee Joint Angle Based on Surface Electromyography Using a Long Short-Term Memory Neural Network and Time-Advanced Feature.

Continuous joint angle estimation based on a surface electromyography (sEMG) signal can be used to improve the man-machine coordination performance of the exoskeleton. In this study, we proposed a time-advanced feature and utilized long short-term memory (LSTM) with a root mean square (RMS) feature and its time-advanced feature (RMSTAF; collectively referred to as RRTAF) of sEMG to estimate the knee joint angle. To evaluate the effect of joint angle estimation, we used root mean square error (RMSE) and cross-correlation coefficient ρ between the estimated angle and actual angle. We also compared three methods (i.e., LSTM using RMS, BPNN (back propagation neural network) using RRTAF, and BPNN using RMS) with LSTM using RRTAF to highlight its good performance. Five healthy subjects participated in the experiment and their eight muscle (i.e., rectus femoris (RF), biceps femoris (BF), semitendinosus (ST), gracilis (GC), semimembranosus (SM), sartorius (SR), medial gastrocnemius (MG), and tibialis anterior (TA)) sEMG signals were taken as algorithm inputs. Moreover, the knee joint angles were used as target values. The experimental results showed that, compared with LSTM using RMS, BPNN using RRTAF, and BPNN using RMS, the average RMSE values of LSTM using RRTAF were respectively reduced by 8.57%, 46.62%, and 68.69%, whereas the average ρ values were respectively increased by 0.31%, 4.15%, and 18.35%. The results demonstrated that LSTM using RRTAF, which contained the time-advanced feature, had better performance for estimating the knee joint motion.

A wavelet artificial neural network method for medium‐term rainfall prediction in Queensland (Australia) and the comparisons with conventional methods

AbstractThis study aims to develop a medium‐term rainfall forecast model to predict monthly rainfall 1, 3, 6, and 12 months in advance using hybrid wavelet‐artificial neural networks (WANN) for Queensland, Australia. To assess the performance of the models, seven input sets comprised of either past rainfall magnitudes, selected climate anomalies, or combinations of rainfall with different climate anomalies as independent variables were defined. The data of the years 1908–1999 and 2000–2016 from 10 weather stations in Queensland were used for training and verification of the models, respectively. The results showed that the WANN enhances the average prediction accuracy in terms of root‐mean‐square‐error (RMSE) with 90, 52, 32, and 15% at 1, 3, 6, and 12 months lead time, respectively compared to of the artificial neural networks (ANN). Also, the current prediction system of the Australian Bureau of Meteorology (BOM), Australian Community Climate Earth‐System Simulator–Seasonal (ACCESS‐S) provided the predictions with the average RMSE values of 91.0, 102.7, and 100.3 mm for 1, 3, and 6 months lead times, respectively. In contrast, the corresponding RMSE values of 8.2, 37.6, and 52.4 mm are obtained through the WANN. Moreover, performance of the WANN method was compared with the autoregressive integrated moving average (ARIMA) and multiple linear regression (MLR) models. ARIMA and MLR produced similar performances at same lead times; corresponding average RMSE values of 78.8, 82.72, 80.9, and 79.5 mm were obtained through ARIMA compared to the 75.9, 79.5, 79.0, and 78.9 mm through the MLR. The results of the current study indicate than the performance of the WANN is more accurate than the ANN, ARIMA, MLR, and ACCESS‐S forecasts.