Figure - available from: Energies
This content is subject to copyright.
Comparing different DNI forecasting models in absence of cloud covers (a) Long-Short Term Memory (b) Gated Recurrent Unit.
Source publication
![Gated Recurrent Unit [19]. The activation functions shown have...](publication/343349866/figure/fig1/AS:11431281397508892@1745516745353/Gated-Recurrent-Unit-19-The-activation-functions-shown-have-associated-sets-of_Q320.jpg)
Power grid operators rely on solar irradiance forecasts to manage uncertainty and variability associated with solar power. Meteorological factors such as cloud cover, wind direction, and wind speed affect irradiance and are associated with a high degree of variability and uncertainty. Statistical models fail to accurately capture the dependence bet...
Similar publications
Chalky grains of rice are increased due to high temperature (HT) during a ripening period. However, the underlying mechanisms of the chalkiness are not well known, seemingly due to the obtuse response of wild type and lack of effective mutants. In this study, we isolated and characterized the flo11-2 mutant, which showed higher number of chalky gra...
Citations
... Majid Hosseini et al. introduced a novel method for predicting DNI using a multivariate gated recurrent unit (GRU) model, which they compared to a long short-term memory model based on historical irradiance data. The findings show that the suggested method considerably enhances DNI prediction accuracy, especially when wind direction and speed are incorporated as observed variables 4 . Larbi Mouhamed et al. computed and evaluated DNI to align with SAM in order to model and simulate concentrated thermal power plants in the Mechria Algeria region 5 . ...
... The selection of the design value is based on a thorough statistical analysis of direct radiation data. A comprehensive review of historical irradiance values, meteorological data, and cloud cover statistics is essential to accurately forecast solar irradiance 4,5,17 . SAM uses meteorological files to fulfill its weather data needs, including altitude information. ...
This paper presents a comprehensive techno-economic analysis of three molten salt Concentrated Solar Power (CSP) tower plants located in the regions of Mechria, Adrar, and Tindouf in Algeria. The study evaluates the thermal efficiency, economic feasibility, and performance of these CSP using the System Advisor Model (SAM) software, which accurately models Direct Normal Irradiance (DNI), a critical factor influencing plant performance. Key parameters analyzed include Solar Multiple (SM), Thermal energy storage (TES) hours, capacity factor (CF), and the Levelized Cost of Energy (LCOE). The results demonstrate that an optimal heliostat field configuration with a SM of 1.8 and 10 h of TES achieves a capacity factor of 51.49%, with a minimum LCOE of 0.097 /kWh at a capacity factor of 24.03%. Similarly, in Tindouf, a SM of 1.6 and TES of 8 h result in a capacity factor of 18.95% and an LCOE of 0.17 $/kWh. The analysis reveals that the design of CSP systems, particularly the combination of solar Multiple and TES, plays a pivotal role in optimizing the economic performance of the plants, This approach enables researchers to save time and costs by using satellite-derived DNI estimations, enhancing data accuracy and optimizing CSP deployment.
... Notably, LSTM's enhanced data handling and feature extraction capabilities make it a promising tool for accurate solar irradiance prediction, overcoming defciencies in traditional statistical methods like ARMA, and signaling its suitability for practical applications. Hosseini et al. [18] introduce a GRU-based approach for Direct Normal Irradiance forecasting, ofering computational efciency while maintaining accuracy comparable to LSTMs. Tey assess univariate and multivariate GRU, optimized using historical irradiance, weather, and cloud cover data from the LRSS solar facility in Colorado. ...
Solar energy with hydropower power plants marks a significant leap forward in renewable energy innovation. The combination ensures a consistent power supply by merging the fluctuations of solar energy with the predictable storage provided by hydropower. This research aims to predict high solar irradiance on hydropower plants to maximize active power generation. A novel hybrid decomposed residual ensembling model for deep learning (SBLTSRARW) using models such as autoregressive integrated moving average (ARIMA) and seasonal-trend decomposition using loess (STL) along with prediction and optimization models such as Bidirectional LSTM (Bi-LSTM), and Whale Optimization Algorithm (WOA) methods are used to predict the irradiances. Various forecasting methods, including STL-Bi-LSTM, SBLTSAR, SBLTARS, and SBLTSRAR models, are assessed to determine their effectiveness in predicting solar radiation. The results show the accuracy of the proposed model, with RMSE and MAE values of 1.85 W/m² and 1.31 W/m², respectively. The proposed SBLTSRARW model results are more accurate than the Bi-LSTM, STL-Bi-LSTM, SBLTSAR, SBLTARS, and SBLTSRAR models, with RMSE value reductions of 517%, 217%, 151%, 98%, and 1%, respectively.
... Deep-learning approaches can effectively leverage large volumes of data and automatically learn complex patterns and relationships between various input variables and output irradiance. Hosseini et al. [18] proposed multivariate Gated Recurrent Units (GRU) and Long Short-Term Memory (LSTM) for multi-step DNI forecasting (15 min to 3 h), incorporating data like Solar Zenith Angle, humidity, temperature, wind direction, and wind speed along with cloud cover from the past 24,48, and 72 h. The data, sourced from Denver's Data Center (2009)(2010)(2011)(2012)(2013)(2014)(2015), showed that for one-hour forecasts, multivariate models outperformed univariate ones in RMSE and MAPE. ...
This study addresses the critical need for precise Direct Normal Irradiation forecasting in concentrating solar power systems to enhance performance and manage power generation intermittency. We propose a novel hybrid model that combines Variation Mode Decomposition, Swarm Decomposition Algorithm, Random Forest for feature selection, and Deep Convolutional Neural Networks, aiming to improve the forecasting accuracy. This model covers the entire process from Direct Normal Irradiation forecasting to heliostat field optimization and electricity generation. We validated the model across four globally diverse regions, taking into account their distinct climates and meteorological conditions. The results show that our model aligns closely with actual measurements and outperforms existing forecasting methods in terms of precision and stability. The forecasting performance was assessed using normalized Root Mean Square Error, with results ranging from 0.75% to 3.4% across different regions. This demonstrates the model's potential for real-world application in concentrating solar power systems, optimizing heliostat field effectiveness, and reliably forecasting electricity production for grid management.
... Employing domain-reduction techniques can be beneficial to address these challenges. Hybrid solutions that use sky features as input to machine learning (ML) models are promising [9,[24][25][26]. These solutions do not use raw images but rather features characterizing sky conditions. ...
... These solutions do not use raw images but rather features characterizing sky conditions. The simplest approach is to feed in the cloud cover besides other exogenous meteorological data to the model [25]. Cloud cover can be calculated using sky camera images [24] or obtained from another external source, i.e., online weather databases [25]. ...
... The simplest approach is to feed in the cloud cover besides other exogenous meteorological data to the model [25]. Cloud cover can be calculated using sky camera images [24] or obtained from another external source, i.e., online weather databases [25]. A more sophisticated approach utilizes various image features generated by a traditional image processing algorithm. ...
In recent years, with the growing proliferation of photovoltaics (PV), accurate nowcasting of PV power has emerged as a challenge. Global horizontal irradiance (GHI), which is a key factor influencing PV power, is known to be highly variable as it is determined by short-term meteorological phenomena, particularly cloud movement. Deep learning and computer vision techniques applied to all-sky imagery are demonstrated to be highly accurate nowcasting methods, as they encode crucial information about the sky’s state. While these methods utilize deep neural network models, such as Convolutional Neural Networks (CNN), and attain high levels of accuracy, the training of image-based deep learning models demands significant computational resources. In this work, we present a computationally economical estimation technique, based on a deep learning model. We utilize both all-sky imagery and meteorological data, however, information on the sky’s state is encoded as a feature vector extracted using traditional image processing methods. We introduce six all-sky image features utilizing detailed knowledge of meteorological and physical phenomena, significantly decreasing the amount of input data and model complexity. We investigate the accuracy of the determined global and diffuse radiation for different combinations of meteorological parameters. The model is evaluated using two years of measurements from an on-site all-sky camera and an adjacent meteorological station. Our findings demonstrate that the model provides comparable accuracy to CNN-based methods, yet at a significantly lower computational cost.
... A tree can be seen as a piecewise constant approximation. decision trees are consisted of root node, branches, internal nodes and leaf nodes [47]. Figure 1 shows a decision tree structure. ...
This research provides a comprehensive prediction using machine learning to predict vapor-liquid-equilibrium for CO2 - contained binary mixtures for carbon capture and sequestration projects. One of the best practices to lower the CO2 emissions in the atmosphere is Carbon Capture and Sequestration including capturing carbon dioxide from atmosphere and injecting it into the underground geological formations. One of the key elements in a successful project is to accurately model the phase equilibria which provides us on how the fluid or mixtures of the injected fluids will behave in certain pressures and temperatures underground. In this regard, different machine learning models have been implemented for the prediction. The data set consists experimental results of five different binary mixtures with CO2 presents in all of them. Then the results were compared to each other and the one with the highest accuracy was selected for each mixture. Peng Robinson equation of state was also used and compared with machine learning results. Finally, both machine learning and thermodynamic models were compared to experimental results to determine the accuracy. It was found out that thermodynamic model was unable to predict results for many data points while machine learning could predict results for most of the data points. Also, the accuracy of machine learning models was greatly better than thermodynamic model. In this research, a large data set including 748 data points is used on which machine learning models can be trained more accurate. Also, as a single machine learning model cannot predict accurate results for all mixtures, several models have been run on each mixture, and the one with the highest accuracy was selected for each CO2 -contained binary mixture which to our knowledge, has been never implemented.
... However, statistical methods unperformed in problems with complex non-linear relationship between features [35]. AI methods, principally machine learning (ML) algorithms such as support vector machines [36,37], fuzzy logic [38,39], evolutionary-based algorithms [40,41], artificial neural networks (ANN) [42,43,44], convolutional neural networks (CNN) [45,46], and recurrent neural networks (RNN) [9,47], proficiently describe non-linear relationships between features in time series data [48]. ...
... Hosseini et al. [47] performed multivariate forecasting of direct normal solar irradiance with LSTM and GRU algorithms. They performed the predictions in several time horizons and demonstrated that multivariate GRU performs better than multivariate LSTM. ...
The intermittent nature of renewable resources like solar and wind presents challenges for small-scale energy markets and off-grid regions. Localized forecasting of these resources is crucial to incentives the attention of citizens and urban designers in adopting small-scale generation systems, i.e., hybrid or renewable energy-based microgrids, in their communities. This study employs Encoder–Decoder Sequence-to-Sequence (Seq2Seq) models with two different recurrent neural network architectures (RNN), long short-term memory (LSTM) and gated recurrent unit (GRU), to forecast wind speed and solar radiation on a location appointed to install microgrids. The models were trained using meteorological data obtained from weather stations while taking into account data on extreme weather events, such as wind gusts, to enhance the predictive accuracy of the models. Particularly relevant in the context of the increasing menace of climate change and the resulting rise in the frequency and severity of extreme weather events. An ensemble method is also employed to quantify uncertainty in the model’s prediction to increase the confidence that an end-user of a microgrid can deposit in the forecasting for decision-making. We compared the Seq2Seq-LSTM and Seq2Seq-GRU models using deterministic and probabilistic metrics. Both models showed satisfactory accuracy and confidence in wind speed and solar radiation forecasting. However, the Seq2Seq-GRU model stood out for its significantly faster computation times. On average, it achieved a 20% reduction in processing time for the same dataset. The models performed best for shorter prediction horizons (1-hour) and higher lookback periods (
= 720-hrs). We also analyzed the models’ performances according to the season. The results show that the seasonal variation significantly affects the accuracy and uncertainty of wind speed forecasting, with fall and summer demonstrating the highest accuracy and winter and spring the lowest. For solar radiation, the models show greater reliability, narrower prediction intervals, and higher accuracy during winter, with opposite trends observed during summer. The deep learning framework proposed in this study can help anticipate daily surpluses or shortfalls in energy generation across different seasons, contributing to the efficient operation of microgrids.
... Regarding the input variables, on [36] the results show an increase in accuracy when more weather variables were included, from 17% MAPE to 10%. On [7] the results show an important drop in accuracy due to the presence of clouds. ...
Photovoltaic power forecasting is an important problem for renewable energy integration in the grid. The purpose of this review is to analyze current methods to predict photovoltaic power or solar irradiance, with the aim of summarizing them, identifying gaps and trends, and providing an overview of what has been achieved in recent years. A search on Web of Science was performed, obtaining 60 articles published from 2020 onwards. These articles were analyzed, gathering information about the forecasting methods used, the horizon, time step, and parameters. The most used forecasting methods are machine learning and deep learning based, especially artificial neural networks. Most of the articles make predictions for one hour or less ahead and predict power instead of irradiance, although both parameters are strongly correlated, and output power depends on received irradiance. Finally, they use weather variables as inputs, consisting mainly of irradiance, temperature, wind speed and humidity. Overall, there is a lack of hardware implementations for real-time predictions, being an important line of development in future decades with the use of embedded prediction systems at the photovoltaic installations.
... This study is focused on the statistical analysis of direct radiation data, preceding irradiance values, meteorological information, Here are some relevant studies from this topic .Majid Hosseini et al suggest a new technique for predicting DNI by utilizing multivariate GRU and comparing it to LSTM using past irradiance data. The findings indicate that the proposed approach enhances DNI prediction accuracy, especially when incorporating wind direction and speed as observation states [1].In this paper, a novel aerosol parameterization is presented by Guoping Shi et al. The aim of this parameterization is to enhance the precision of direct normal irradiance (DNI) for the purpose of solar energy forecast. ...
... Concentrated solar power plants use direct solar radiation in order to convert solar energy into electricity or heat or cold, The dimensioning of the opening surface based essentially on the value of the reference radiation, this value is selected as a function of the solar deposit in the site of implantation, and the total investment cost of such a CSP plant is proportional to the reference value of the direct radiation chosen. The choice of the design value is based on a statistical study of direct radiation data, The analysis of previous irradiance values, meteorological information, and cloud cover statistics is imperative for anticipating solar irradiance [1]. We studied the radiation by filtering the intensity of direct radiation in ranges of 200 w /m2, we started with the values less than 200w /m2 and then the larger intervals up to greater than 1000w /m2, for each interval we calculated the number of hours or the direct radiation included in the interval We scan the whole year hour by hour, we find the interval where the direct solar radiation is the number of hours is maximum so the design value is selected from this interval. ...
SUN FLUX calculates solar irradiance on Earth's surface quickly and accurately. Direct solar radiation is critical for concentrated solar power plants, which determine their opening surface size based on reference radiation values. To reduce investment costs, a statistical analysis was performed using direct radiation, meteorological data from NSDRB, and cloud cover statistics to identify optimum ranges for direct solar radiation. CSP plant performance is evaluated and power tower systems are simulated using the SAM software tool.
... To address the fluctuation and intermittency of solar and wind energy, Alharbi and Csala [25] proposed a BI-LSTM (bidirectional long-short term memory) prediction model with high accuracy based on historical fluctuation DNI and wind speed data. Using historical irradiance data and weather information, Hosseini et al. [26] introduced a DNI prediction method that is as accurate as the LSTM model at forecasting solar irradiance results. With DNI as the aim of prediction, Zhu et al. ...
In spite of the fact that the solar power tower system is considered as one of the most valuable power generation facilities, it still faces challenges such as insufficient utilization of the solar salt temperature range and the dependence on high solar radiation intensity. To address this issue, an integrated recompression Brayton cycle and trans-critical regenerative organic Rankine cycle in parallel layout is proposed. The comparison with other literature shows that the specific work of the integrated system (147.8 kW/kg) outperforms the partial cooling Brayton cycle (130.1 kW/kg) under the same solar salt temperature range. In order to forecast future electricity generation and serve as a reference for plant operation, a power prediction strategy using neural networks is developed. Findings indicate that the integrated system can increase power production and maximize solar salt temperature utilization by adjusting the physical constraint strategy based on the changing temperature interval. Its equivalent work and thermal efficiency reach 12.7 MW and 38.36%, respectively, and it can recover the cost in 8 years. These results suggest that the proposed integrated system is a promising solution to enhance the performance of solar power tower systems with significant economic benefits.
... Yet, the weather characteristics are typically random anddepending on the type of marketelectricity prices may exhibit time-varying and volatile statistical properties (Carolina et al., 2020). The findings of previous studies have shown that forecast models developed for DNI and spot prices predictions have a limited accuracy of about 10-15 % mean absolute percentage error (MAPEs) (Hosseini et al., 2020;Wang et al., 2020). Accordingly, the existence of uncertainty in the forecasting can cause the "optimal" dispatch plan to be too aggressive (e.g. ...
The aim at dispatch scheduling for concentrating solar power plants is to utilize thermal storage to maximize profit from electricity generation. The dispatching plan, however, must contend with significant uncertainty in both weather forecasts (particularly solar irradiation) and, in some locations, real-time electricity prices governed by potentially volatile markets. This paper proposes a Model-Predictive Control (MPC) which employs a novel similarity-based forecast for weather variables and exploits the predictions of electricity prices provided by the market operator. A Monte-Carlo simulation is developed to evaluate true performance of the proposed MPC against two benchmarks, with perfect knowledge (PK) forecasts and day-ahead optimized using prototypical weather (DAPW). A case study is performed on a 115 MW Solar Tower plant with 8-hour storage, hypothetically located in South Australia, and operates under either a fixed-price or a real-time spot price scenario. A Monte-Carlo simulation is conducted for 150 tests for January (summer) and August (winter). The results show in fixed-price scenario MPC achieves 82.4 % of optimal profit (i.e. obtained via PK) in January and 72.4 % in August whereas that in DAPW falls to 71.5 % and 56.9 % of optimal, respectively. In the spot market scenario, MPC reaches 71.3 % and 63.6 % of optimal profit in January and August, whereas that for DAPW reaches 61.4 % and 55.1 %, respectively. In conclusion, PK forecast assumption in dispatch planning over-estimates the achievable profit by ∼ 28–40 % particularly in spot market scenario. Moreover, MPC can mitigate the influence of uncertainties on the plant economic performance by 8.5 %–15.5 % compared to DAPW benchmarks.
