Deep learning models for solar irradiance forecasting: A comprehensive review

Capacitive resistivity subsurface imaging of roads operating at very low frequency is susceptible to antenna characteristic capacitance dynamics that may cause unwanted signal reflection, coupling, and unfavorable effect on reception sensitivity. Antennas are conventionally modeled using a complex and repetitive default mathematical method that is prone to human error and discrete results. To address this emerging challenge, this study has developed a new technique for plate-wire antenna capacitance optimization through equatorial dipole-dipole antenna geometry modeling using genetic programming (GP) integrated with metaheuristic methods, namely Archimedes optimization algorithm (AOA), Lichtenberg algorithm (LA), and Henry gas solubility optimization (HGSO). GP was used to construct the antenna capacitance fitness function based on 241 combinations of wire antenna radius and elevation, and dipole plate elevation, length, width, and thickness measurements. Minimization of antenna capacitance (approaching 1 nF) to achieve quasi-static condition was performed using GP-AOA, GP-LA, and GP-HGSO. The 3 metaheuristic-based antennas were 3D-modeled using Altair Feko and compared from the default antenna’s electrical features. It was found that even with the smallest dipole geometry, hybrid GP-LA antenna model exhibited the most practical outputs at 5 kHz with correct directional propagation based on its radiation pattern, a realistic receiver voltage of -8.86 dBV which is close to the default model, and a high-power efficiency of 99.925%. While hybrid GP-AOA and GP-HGSO resulted in indirect coupled transceiver systems with unsuitable antenna characteristic capacitance inducing anomalous receiver voltages. The experimental results prove the validity of the developed technique for more accurate determination of optimal antenna geometry.

Optimization of Subsurface Imaging Antenna Capacitance through Geometry Modeling using Archimedes, Lichtenberg and Henry Gas Solubility Metaheuristics

A multiple combined method for rebalancing medical data with class imbalances.

Solar energy is one of the most important renewable energies, with many advantages over other sources. Many parameters affect the electricity generation from solar plants. This paper aims to study the influence of these parameters on predicting solar radiation and electric energy produced in the Salt-Jordan region (Middle East) using long short-term memory (LSTM) and Adaptive Network-based Fuzzy Inference System (ANFIS) models. The data relating to 24 meteorological parameters for nearly the past five years were downloaded from the MeteoBleu database. The results show that the influence of parameters on solar radiation varies according to the season. The forecasting using ANFIS provides better results when the parameter correlation with solar radiation is high (i.e., Pearson Correlation Coefficient PCC between 0.95 and 1). In comparison, the LSTM neural network shows better results when correlation is low (PCC in the range 0.5–0.8). The obtained RMSE varies from 0.04 to 0.8 depending on the season and used parameters; new meteorological parameters influencing solar radiation are also investigated.

/pdf/solar-radiation-forecasting-by-pearson-correlation-using-3exg26hf.pdf

Solar Radiation Forecasting by Pearson Correlation Using LSTM Neural Network and ANFIS Method: Application in the West-Central Jordan

As non-renewable energy sources are in the verge of exhaustion, the entire world turns towards renewable sources to fill its energy demand. In the near future, solar energy will be a major contributor of renewable energy, but the integration of unreliable solar energy sources directly into the grid makes the existing system complex. To reduce the complexity, a microgrid system is a better solution. Solar energy forecasting models improve the reliability of the solar plant in microgrid operations. Uncertainty in solar energy prediction is the challenge in generating reliable energy. Employing, understanding, training, and evaluating several forecasting models with available meteorological data will ensure the selection of an appropriate forecast model for any particular location. New strategies and approaches emerge day by day to increase the model accuracy, with an ultimate objective of minimizing uncertainty in forecasting. Conventional methods include a lot of differential mathematical calculations. Large data availability at solar stations make use of various Artificial Intelligence (AI) techniques for computing, forecasting, and predicting solar radiation energy. The recent evolution of ensemble and hybrid models predicts solar radiation accurately compared to all the models. This paper reviews various models in solar irradiance and power estimation which are tabulated by classification types mentioned.

/pdf/systematic-review-on-impact-of-different-irradiance-282w46s3.pdf

Systematic Review on Impact of Different Irradiance Forecasting Techniques for Solar Energy Prediction

Crops that are cultivated in aquaponics setup highly relies on the nutrients supplied by the aqueous system through fish effluents. Continuous monitoring of essential elemental nutrients requires expensive sensors and arrays of it for full scale deployment. However, sustainable agriculture demands energy consumption reduction and cost-effectiveness. This study employed device minimization by utilizing a combination of physical water sensors, namely temperature and electrical conductivity sensors, to predict total organic carbon (TOC) and hydrogen ion (H) concentrations in pond water. Aquaphotomics through ultraviolet (UV) and visible light (Vis) wavelength sweeping from 250 to 500 nm was explored to determine the nutrient biomarkers of pond water samples that undergoes temperature perturbation from 16 to $36^{\circ}C$ with $2^{\circ}C$ increment per testing. Principal component analysis (PCA) selected the most relevant activated water bands which are 275 nm for TOC and 415 nm for H. Direct spectrophotometric TOC concentration data was passed through a Savitzky-Golay filter to smoothen the nutrient signal. Recurrent neural network (RNN) exhibited the fastest inference time of 3.5 seconds on the average with R2 of 0. S583 and 0.9686 for predicting TOC and H concentrations. Multigene symbolic regression genetic programming (MSRGP) exhibited the best R2 performances of 0.9280 and 0.9693 in predicting TOC and H concentrations by using only the temperature and electrical conductivity sensor-acquired data. This developed model is an innovative approach on measuring chemical concentrations of water using physical limnological sensors which resulted to energy consumption reduction of 50% for complete 42-day crop life cycle of lettuce.

Aquaphotomics Determination of Total Organic Carbon and Hydrogen Biomarkers on Aquaponic Pond Water and Concentration Prediction Using Genetic Programming

Energy production of photovoltaic (PV) system is heavily influenced by solar irradiance. Accurate prediction of solar irradiance leads to optimal dispatching of available energy resources and anticipating end-user demand. However, it is difficult to do due to fluctuating nature of weather patterns. In the study, neural network models were defined to predict solar irradiance values based on weather patterns. Models included in the study are artificial neural network, convolutional neural network, bidirectional long-short term memory (LSTM) and stacked LSTM. Preprocessing methods such as data normalization and principal component analysis were applied before model training. Regression metrics such as mean squared error (MSE), maximum residual error (max error), mean absolute error (MAE), explained variance score (EVS), and regression score function (R2 score), were used to evaluate the performance of model prediction. Plots such as prediction curves, learning curves, and histogram of error distribution were also considered as well for further analysis of model performance. All models showed that it is capable of learning unforeseen values, however, stacked LSTM has the best results with the max error, R2, MAE, MSE, and EVS values of 651.536, 0.953, 41.738, 5124.686, and 0.946, respectively.

Using Stacked Long Short Term Memory with Principal Component Analysis for Short Term Prediction of Solar Irradiance based on Weather Patterns

Optimizing crop production is one of the significant advances in urban agriculture and biosystems engineering. This deals with providing possibilities of year-long plant-based food supply in an efficiently limited space or environment; to cater the increasing demand of population necessities. The study to be discussed contributes on a novel methodology on optimizing lettuce production in a smart aquaponics setup by predicting its phytomorphological features _ deemed to be a noteworthy factor for plant phenomics research. Pre-trained deep learning-based networks were customized to train the machine vision-based feature extracted data. The data comprised of the phytomorphological attributes of leaf lettuce, to be predicted by the model through regression, thus setting them as the system's response data. The phytomorphological features considered in the study are the area, centroid, major and minor axis length, and equivdiameter. The sole predictor data will be the raw and resized images gathered. Evaluation shows that DarkN et-53 was the best network used for modelling area prediction with an R2 value of 0.99. For x and y centroids, Xception was the best estimator, yielding R2 values of 0.64 and 0.72 respectively. InceptionResN etV2 derived the highest-performing axis length prediction models with R2 values of 0.82 for the major axis length while 0.75 for the minor. Equivdiameter with the highest R2 value of 0.98 was obtained with DarkN et-53.

Machine Vision-Based Prediction of Lettuce Phytomorphological Descriptors using Deep Learning Networks

chronic kidney disease (CKD) is one of the diseases with high mortality rate. It is a disease resulted from kidney function loss over a long period of time. The disease shows no symptoms during initial stage. When left not medicated, a person may suffer from other complications such as high blood pressure, anemia, malnutrition, increased risk of cardiovascular disease, cognitive impairment and impaired physical function. Automated diagnosis by using classification algorithms has been an interest of researchers. In this study, six machine learning algorithms were used for classification and its prediction performance was compared based on training time and F1 score, with and without hypertuning the parameters. Of all the six algorithms, KNN has the best F1 score of 0.992248 and minimal training time of 46.999ms. The performance of decision trees was improved with hypertuning, having a F1 score from 0.96 to 0.99. Overall, machine learning algorithms are significant tool to assess chronic kidney disease.

Performance Comparison of Classification Algorithms for Diagnosing Chronic Kidney Disease

Fault detection and monitoring system in photovoltaic (PV) energy management system is important in achieving its optimal performance. An effective diagnostic system involves correct analysis of electrical parameters of a PV array on a given weather condition. In the study, mean-shift clustering was applied for pre-classification and anomaly detection of time-series data of electrical parameters from grid-tied inverter, and solar-irradiance. Classification and anomaly detection applied is based in ensemble learning, where its base learners are based from multilayer perceptron. A stacking ensemble is used in classification of energy production profile while bagging ensemble is used detecting anomalous trend in time-series data. A stacking ensemble got a highest accuracy value of 94% compared to single classifiers which have accuracy value of 85.25%, 84.14%, and 63.4%, respectively. The bagging ensemble autoencoders have the lowest mean squared error during model reconstruction compared to single autoencoder. It has a fair performance in classifying anomaly points from normal datapoints, having an AUC value of 0.795 and F1-score of 0.71, given that the hyperparameter is 0.5. Overall, ensemble learners improve the performance in classification and detection tasks.

Application of Ensemble Learning with Mean Shift Clustering for Output Profile Classification and Anomaly Detection in Energy Production of Grid-Tied Photovoltaic System

Justin D. de Guia

Papers

Aquaphotomics Determination of Total Organic Carbon and Hydrogen Biomarkers on Aquaponic Pond Water and Concentration Prediction Using Genetic Programming

Using Stacked Long Short Term Memory with Principal Component Analysis for Short Term Prediction of Solar Irradiance based on Weather Patterns

Machine Vision-Based Prediction of Lettuce Phytomorphological Descriptors using Deep Learning Networks

Performance Comparison of Classification Algorithms for Diagnosing Chronic Kidney Disease

Application of Ensemble Learning with Mean Shift Clustering for Output Profile Classification and Anomaly Detection in Energy Production of Grid-Tied Photovoltaic System