• Energy Research

In this paper, a low-cost monitoring system for an off-grid photovoltaic (PV) system, installed at an isolated location (Sahara region, south of Algeria), is designed. The PV system is used to supply a small-scale greenhouse farm. A simple and accurate fault diagnosis algorithm was developed and integrated into a low-cost microcontroller for real time validation. The monitoring system, including the fault diagnosis procedure, was evaluated under specific climate conditions. The Internet of Things (IoT) technique is used to remotely monitor the data, such as PV currents, PV voltages, solar irradiance, and cell temperature. A friendly web page was also developed to visualize the data and check the state of the PV system remotely. The users could be notified about the state of the PV system via phone SMS. Results showed that the system performs better under this climate conditions and that it can supply the considered greenhouse farm. It was also shown that the integrated algorithm is able to detect and identify some examined defects with a good accuracy. The total cost of the designed IoT-based monitoring system is around 73 euros and its average energy consumed per day is around 13.5 Wh.

Solar photovoltaic (PV) energy is becoming used gradually as an alternative source to classical fossil fuel because of being renewable, clean and abundant. In agricultural regions such as Hail, Saudi Arabia, farmers can invest in using PV sources for pumping water from relatively deep wells. In such case, forecasting the potential of PV energy is among the primary factors that may affect the design of a PV water pumping system efficiency and reliability. In this paper, a novel combined method based on Hammerstein—autoregressive with exogenous input—Wiener model optimized by particle swarm optimization was developed. Four models including, saturation, dead-zone and polynomial nonlinear blocks have been evaluated using global horizontal irradiation (GHI) as model output and separately, the temperature, the clearness index, the relative humidity and the wind speed as model inputs on daily basis collected from Hail region, Saudi Arabia over 20 years (2000–2019). The collected 7305 observations were divided into a training subset (6205) and a testing subset (1095). Six other models including persistence forecast, smart persistence forecast, time series autoregressive and artificial neural networks have been also implemented on the same dataset for comparison purpose. Based on robust performance indicators (mean absolute percentage error, coefficient of determination, the root mean square error and skill score), the Hammerstein–Saturation with temperature as input model outperformed all the other developed models. To validate the suitability of the proposed approach for forecasting GHI in arid climates, our best model was implemented on two locations picked randomly in the Hail region (Saudi Arabia). The results of validation were in the same scale of accuracy as the primary best model.

L'irradiation horizontale globale prévue (GHI) peut aider à la conception, au dimensionnement et à l'analyse des performances des systèmes photovoltaïques (PV), y compris les systèmes de pompage PV à eau utilisés pour les applications d'irrigation. Dans cet article, divers modèles de réseaux neuronaux profonds (DNN) pour la prédiction à un jour du GHI à Hail city (Arabie saoudite) sont développés et étudiés. Les modèles DNN considérés comprennent la mémoire à long terme (LSTM), le LSTM bidirectionnel (BiLSTM), l'unité récurrente fermée (GRU), le GRU bidirectionnel (Bi-GRU), le réseau neuronal convolutionnel unidimensionnel (CNN 1D ) et d'autres configurations hybrides telles que CNN-LSTM et CNN-BiLSTM.A. L'ensemble de données des enregistrements quotidiens GHI collectés entre le 1er janvier 2000 et le 30 juin 2020 auprès de la National Aeronautics and Space Administration (NASA) à un endroit aride (Hail, Arabie Saoudite) est utilisé pour développer et comparer les modèles basés sur DNN ci-dessus. Les paramètres affectant la précision des modèles ont également été analysés en profondeur. Seules les valeurs historiques du GHI quotidien ont été utilisées pour construire les modèles basés sur DNN, tandis que des paramètres météorologiques supplémentaires tels que la température de l'air, la vitesse du vent, la direction du vent, la pression atmosphérique et l'humidité relative ne sont pas pris en compte dans ce travail. La bibliothèque Keras et le langage Python ont été utilisés utilisé pour développer et comparer les modèles de prévision GHI. Les métriques d'évaluation telles que le coefficient de corrélation (r), l'erreur absolue moyenne en pourcentage (MAPE), l'erreur absolue moyenne (MAE), la fonction de distribution cumulative (CDF) et l'écart type (σ ) sont choisies pour évaluer la performance des modèles de prévision. Les résultats obtenus ont montré que les modèles DNN ont fourni de bonnes performances à l'échelle mondiale avec une valeur maximale atteinte de r = 96 %, pour la prévision quotidienne GHI. La irradiación horizontal global pronosticada (GHI) puede ayudar a diseñar, dimensionar y analizar el rendimiento de los sistemas fotovoltaicos (PV), incluidos los sistemas de bombeo de agua PV utilizados para aplicaciones de riego. En este documento, se desarrollan e investigan varios modelos de redes neuronales profundas (DNN) para la predicción de un día de anticipación de GHI en la ciudad de Hail (Arabia Saudita). Los modelos DNN considerados incluyen memoria a largo plazo (LSTM), LSTM bidireccional (BiLSTM), unidad recurrente cerrada (GRU), GRU bidireccional (Bi-GRU), red neuronal convolucional unidimensional (CNN 1D ) y otras configuraciones híbridas como CNN-LSTM y CNN-BiLSTM. Un conjunto de datos de grabaciones diarias de GHI recopiladas durante el 1 de enero de 2000 al 30 de junio de 2020 de la Administración Nacional de Aeronáutica y del Espacio (NASA) en una ubicación árida (Hail, Arabia Saudita) se utiliza para desarrollar y comparar los modelos basados en DNN anteriores. Los parámetros que afectan la precisión de los modelos también se han analizado profundamente. Solo se han utilizado valores históricos de GHI diarios para construir los modelos basados en DNN, mientras que los parámetros climáticos adicionales como la temperatura del aire, la velocidad del viento, la dirección del viento, la presión atmosférica y la humedad relativa no se consideran en este trabajo. Biblioteca Keras y lenguaje Python han sido utilizado para desarrollar y comparar los modelos de pronóstico de GHI. Las métricas de evaluación como el coeficiente de correlación (r), el error porcentual absoluto medio (MAPE), el error absoluto medio (MAE), la función de distribución acumulativa (CDF) y la desviación estándar (σ ) se optan para evaluar el rendimiento de los modelos de predicción. Los resultados obtenidos mostraron que los modelos DNN han proporcionado un buen rendimiento a nivel mundial con un valor máximo alcanzado de r = 96%, para el pronóstico diario de GHI. Forecasted global horizontal irradiation (GHI) can help for designing, sizing and performances analysis of photovoltaic (PV) systems including water PV pumping systems used for irrigation applications.In this paper, various deep neural networks (DNN) models for one day-ahead prediction of GHI at Hail city (Saudi Arabia) are developed and investigated.The considered DNN models include long-shortterm memory (LSTM), bidirectional-LSTM (BiLSTM), gated recurrent unit (GRU), bidirectional-GRU (Bi-GRU), one-dimensional convolutional neural network (CNN 1D ) and other hybrid configurations such as CNN-LSTM and CNN-BiLSTM.A dataset of daily GHI recordings collected during January 1, 2000 to June 30, 2020 from National Aeronautics and Space Administration (NASA) at an arid location (Hail, Saudi Arabia) is used to develop and compare the above DNN-based models.The parameters affecting the accuracy of the models have been also deeply analyzed.Only historical values of daily GHI have been used to build the DNN-based models whereas additional weather parameters such as air temperature, wind speed, wind direction, atmospheric pressure and relative humidity are not considered in this work.Keras library and Python language have been used to develop and compare the GHI forecasting models.The evaluation metrics such as correlation coefficient (r), Mean Absolute Percent Error (MAPE), Mean Absolute Error (MAE), cumulative distribution function (CDF) and standard deviation (σ ) are opted to evaluate the performance of the prediction models.The obtained results showed that the DNN models have provided globally good performances with a maximum reached value of r = 96%, for daily GHI forecasting. يمكن أن يساعد الإشعاع الأفقي العالمي المتوقع (GHI) في تصميم وتحجيم وتحليل أداء الأنظمة الكهروضوئية (PV) بما في ذلك أنظمة ضخ المياه الكهروضوئية المستخدمة في تطبيقات الري. في هذه الورقة، يتم تطوير نماذج مختلفة للشبكات العصبية العميقة (DNN) للتنبؤ قبل يوم واحد من GHI في مدينة حائل (المملكة العربية السعودية) والتحقيق فيها. تشمل نماذج DNN التي يتم النظر فيها الذاكرة قصيرة المدى (LSTM)، ثنائية الاتجاه - LSTM (BiLSTM)، وحدة متكررة مسورة (GRU)، ثنائية الاتجاه - GRU (Bi - GRU)، شبكة عصبية التفافية أحادية البعد (CNN 1D ) والتكوينات الهجينة الأخرى مثل CNN - LSTM و CNN - BiLSTM. يتم استخدام مجموعة بيانات من تسجيلات GHI اليومية التي تم جمعها خلال الفترة من 1 يناير 2000 إلى 30 يونيو 2020 من الإدارة الوطنية للملاحة الجوية والفضاء (ناسا) في موقع قاحل (حائل، المملكة العربية السعودية) لتطوير ومقارنة النماذج القائمة على DNN المذكورة أعلاه. كما تم تحليل المعلمات التي تؤثر على دقة النماذج بعمق. تم استخدام القيم التاريخية فقط لـ GHI اليومية لبناء النماذج القائمة على DNN بينما لا يتم النظر في معلمات الطقس الإضافية مثل درجة حرارة الهواء وسرعة الرياح واتجاه الرياح والضغط الجوي والرطوبة النسبية في هذا العمل. تستخدم لتطوير ومقارنة نماذج التنبؤ بمؤشر GHI. يتم اختيار مقاييس التقييم مثل معامل الارتباط (r)، متوسط النسبة المئوية المطلقة (MAPE)، متوسط الخطأ المطلق (MAE)، وظيفة التوزيع التراكمي (CDF) والانحراف المعياري (σ ) لتقييم أداء نماذج التنبؤ. أظهرت النتائج التي تم الحصول عليها أن نماذج DNN قد قدمت أداءً جيدًا عالميًا بحد أقصى للقيمة التي تم الوصول إليها r = 96 ٪، للتنبؤ اليومي بمؤشر GHI.

In this study, a novel technique for identifying and categorizing flaws in small-scale photovoltaic systems is presented. First, a supervised machine learning (neural network) was developed for the fault detection process based on the estimated output power. Second, an extra tree supervised algorithm was used for extracting important features from a current-voltage (I–V) curve. Third, a multi-stacking-based ensemble learning algorithm was developed to effectively classify faults in solar panels. In this work, single faults and multiple faults are investigated. The benefit of the stacking strategy is that it can combine the strengths of several machine learning-based algorithms that are known to deliver good results on classification tasks, producing results that are more precise and efficient than those produced by a single algorithm. The approach was tested using an experimental dataset and the findings show that it could accurately diagnose faults (a detection rate of around 98.56% and a classification rate of around 96.21%). A comparison study with different ensemble learning algorithms (AdaBoost, CatBoost, and XGBoost) was conducted to evaluate the effectiveness of the suggested method.

In order to satisfy increasing energy demand and mitigate global warming worldwide, the implementation of photovoltaic (PV) clean energy installations needs to become common practice. However, solar energy is known to be dependent on several random factors, including climatic and geographic conditions. Prior to promoting PV systems, an assessment study of the potential of the considered location in terms of power yield should be conducted carefully. Manual assessment tools are unable to handle high amounts of data. In order to overcome this difficulty, this study aims to investigate various artificial intelligence (AI) models—with respect to various intuitive prediction benchmark models from the literature—for predicting solar energy yield in the Ha’il region of Saudi Arabia. Based on the daily data, seven seasonal models, namely, naïve (N), simple average (SA), simple moving average (SMA), nonlinear auto-regressive (NAR), support vector machine (SVM), Gaussian process regression (GPR) and neural network (NN), were investigated and compared based on the root mean square error (RMSE) and mean absolute percentage error (MAPE) performance metrics. The obtained results showed that all the models provided good forecasts over three years (2019, 2020, and 2021), with the naïve and simple moving average models showing small superiority. The results of this study can be used by decision-makers and solar energy specialists to analyze the power yield of solar systems and estimate the payback and efficiency of PV projects.

This study investigates the asymmetric impacts of Bitcoin prices on Bitcoin energy consumption. Two series are shown to be chaotic and non-linear using the BDS Independence test. To take into consideration this nonlinearity, we employed the QNARDL model as a traditional technique and Support Vector Machine (SVM) and eXtreme Gradient Boosting (XGBoost) as non-conventional approaches to study the link between Bitcoin energy usage and Bitcoin prices. Referring to QNARDL estimates, results show that the relationship between Bitcoin energy use and prices is asymmetric. Additionally, results demonstrate that changes in Bitcoin prices have a considerable effect, both short- and long-run, on energy consumption. As a result, any upsurge in the price of Bitcoin leads to an immediate boost in energy use. Furthermore, the short-term drop in Bitcoin values causes an increase in energy use. However, higher Bitcoin prices reduce energy use in the long run. Otherwise, every decline in Bitcoin prices leads to a long-term reduction in energy use. In addition, the performance metrics and convergence of the cost function provide evidence that the XGBoost model dominates the SVM model in terms of Bitcoin energy consumption forecasting. In addition, we analyze the effectiveness of several modeling approaches and discover that the XGBoost model (MSE: 0.52%; RMSE: 0.72 and R2: 96%) outperforms SVM (MSE: 4.89; RMSE: 2.21 and R2: 75%) in predicting. Results indicate that the forecast of Bitcoin energy consumption is more influenced by positive shocks to Bitcoin prices than negative shocks. This study gives insights into the policies that should be implemented, such as increasing the sustainable capacity, efficiency, and flexibility of mining operations, which would allow for the reduction of the negative impacts of Bitcoin price shocks on energy consumption.