Irradiance Forecasting Research Articles

New soft computing model for multi-hours forecasting of global solar radiation

New soft computing model for multi-hours forecasting of global solar radiation

How Can Probabilistic Solar Power Forecasts Be Used to Lower Costs and Improve Reliability in Power Spot Markets? A Review and Application to Flexiramp Requirements

Net load uncertainty in electricity spot markets is rapidly growing. There are five general approaches by which system operators and market participants can use probabilistic forecasts of wind, solar, and load to help manage this uncertainty. These include operator situation awareness, resource risk hedging, reserves procurement, definition of contingencies, and explicit stochastic optimization. We review these approaches, and then provide a case study in which a method for using probabilistic solar forecasts to define needs for reserves is developed and evaluated. The case study has three parts. First, we describe building blocks for enhancing the Watt-Sun solar forecasting system to produce probabilistic irradiance and power forecasts. Second, relationships between Watt-Sun forecasts for multiple sites in California and the system’s need for flexible ramp capability (flexiramp) are defined by machine learning and statistical methods. Third, the performance of present methods to defining flexiramp requirements, which are not conditioned on weather and renewables forecasts, is compared with that of probabilistic solar forecast-based requirements, using a multi-timescale production costing model with an 1820-bus representation of the WECC power system. Significant potential savings in fuel and flexiramp procurement costs from using solar-informed reserve requirements are found.

Extreme Gradient Boosting: A Machine Learning Technique for Daily Global Solar Radiation Forecasting on Tilted Surfaces

Enhancing solar irradiance and accurate forecasting is required for improved performance of photovoltaic and solar thermal systems. In this study, Extreme Gradient Boosting (XGBoost) model was developed using three input parameters (time, day number, and horizontal solar radiation) and was utilized to forecast daily global solar radiation on tilted surfaces. The proposed model was built using XGBRegressor with five generations, 100 n estimators, and a learning rate of 0.1. Three statistical metrics, such as the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE), were used to compare the model’s results to observed solar radiation data from the Nation Centre for Energy, Research and Development, University of Nigeria, Nsukka. The results showed improved prediction accuracy and XGBoost capability to estimate daily global solar radiation on tilted surfaces. In the training section, the proposed model had a statistical performance of R2 = 0.9977, RMSE = 1.6988, and MAE = 1.081, and in the testing section, R2 = 0.9934, RMSE = 2.8558, and MAE = 2.033. XGBoost model demonstrated a better performance when compared with other models in the literature. As a result, the proposed model provides an effective approach for estimating solar radiation.

Automated Deep CNN-LSTM Architecture Design for Solar Irradiance Forecasting

Accurate prediction of solar energy is an important issue for photovoltaic power plants to enable early participation in energy auction industries and cost-effective resource planning. This article introduces a new deep learning-based multistep ahead approach to improve the forecasting performance of global horizontal irradiance (GHI). A deep convolutional long short-term memory is used to extract optimal features for accurate prediction of the GHI. The performance of such deep neural networks directly depends on their architectures. To deal with this problem, a swarm evolutionary optimization method, called the sine-cosine algorithm, is applied and advanced to automatically optimize the network architecture. A three-phase modification model is proposed to increase the diversity of population and avoid premature convergence in the optimization mechanism. The performance of the proposed method is investigated using three datasets collected from three solar stations in the east of the United States. The experimental results demonstrate the superiority of the proposed method in comparison to other forecasting models.

Optimized Feature Selection Based on a Least-Redundant and Highest-Relevant Framework for a Solar Irradiance Forecasting Model

Optimized Feature Selection Based on a Least-Redundant and Highest-Relevant Framework for a Solar Irradiance Forecasting Model

Week-ahead solar irradiance forecasting with deep sequence learning

Abstract In order to enable widespread integration of solar energy into the power system, there is an increasing need to reduce the uncertainty associated with solar power output which requires major improvements in solar irradiance forecasting. While most recent works have addressed short-term (minutes or hours ahead) forecasting, through this work, we propose using deep sequence learning models for forecasting at longer lead times such as a week in advance, as this can play a significant role in future power system storage applications. Along with point forecasts, we also produce uncertainty estimates through probabilistic prediction and showcase the potential of our machine learning frameworks for a new and important application of longer lead time forecasting in this domain. Our study on the SURFRAD data over seven US cities compares various deep sequence models and the results are encouraging, demonstrating their superior performance against most benchmarks from the literature and a current machine learning-based probabilistic prediction baseline (previously applied to short-term solar forecasting).

Multibranch Attentive Gated ResNet for Short-Term Spatio-Temporal Solar Irradiance Forecasting

The increasing penetration of solar generation into power grids has promoted the need for accurate and reliable short-term solar irradiance forecasting. This article introduces a novel multibranch attentive gated recurrent residual network (ResAttGRU) consisting of multiple branches of residual networks, gated recurrent units (GRUs), and the attention mechanism. The proposed multibranch ResAttGRU is capable of modeling data at various resolutions, extracting hierarchical features, and capturing short- and long-term dependencies. Moreover, this network also presents a strong multitimescale representative within the proposed architecture, while GRUs can exploit temporal information at less computational cost than the popular long short-term memory (LSTM). The novelty of the proposed architecture is to employ multiple convolutional-based branches with different filter lengths to learn multitimescale features jointly, accelerate the learning process, and reduce overfitting by leveraging shared representations as the auxiliary information. This study also compares the multibranch ResAttGRU networks with state-of-the-art deep learning methods using 18 years of NSRDB data at 12 solar sites. Finally, the proposed multibranch ResAttGRU requires 7.1% fewer parameters than multibranch residual LSTM while achieving similar average RMSE, MAE, and R-squared values.

Review of Different Error Metrics: A Case of Solar Forecasting

Renewable energy systems (RES) are no longer confined to being used as a stand-alone entity in the modern era. These RES, especially solar panels are also used with the grid power systems to supply electricity. However, precise forecasting of solar irradiance is necessary to ensure that the grid operates in a balanced and planned manner. Various solar forecasting models (SFM) are presented in the literature to produce an accurate solar forecast. Nevertheless, each model has gone through the step of evaluation of its accuracy using some error measures. Many error measures are discussed in the literature for deterministic as well as probabilistic solar forecasting. But, each study has its own selected error measure which sometimes landed on a wrong interpretation of results if not selected appropriately. As a result, this paper offers a critical assessment of several common error metrics with the goal of discussing alternative error metrics and establishing a viable set of error metrics for deterministic and probabilistic solar forecasting. Based on highly cited research from the last three years (2019-2021), error measures for both types of forecasting are presented with their basic functionalities, advantages & limitations which equipped the reader to pick the required compatible metrics

Intra-day global horizontal irradiance forecast using FY-4A clear sky index

Accurate forecasting of Global horizontal irradiance (GHI) plays an irreplaceable role in power system expansion, power generation production scheduling, maintenance scheduling and ensuring continuous power supply. A method for GHI forecasting based on FY-4A satellite images is proposed, called FY-4A-Heliosat. This method is derived from the clear sky model (MCclear) and the future clear sky index (CSI), which is obtained by comparing the continuous cloud albedo (CAL) maps derived from AGRI satellite images according to Heliosat-2 method. The forecast horizon is 30 min to 3 h with a 30-min horizon resolution. The combination of climatology and persistence model (CP) is chosen as reference forecasts. Under all sky conditions, the results show that compared with CP model, FY-4A-Heliosat model is better in most cases and seasons. The FY-4A-Heliosat method underestimates the GHI in four seasons for all time horizons. Annual studies indicate that the FY-4A-Heliosat method performs better than the CP method for all forecast horizons; the nRMSE is 17% to 25% for the annual performance. Forecasts with change in cloud conditions show that: the FY-4A-Heliosat method performs better than the CP method in clear, partially cloudy, overcast and clear-to-cloudy conditions. For the cloudy-to-clear set, the FY-4A-Heliosat method needs to be improved future. The results demonstrate that the proposed method is suitable for very short term GHI forecasting.

The Satellite Risk Prediction and Radiation Forecast System (SaRIF)

AbstractWith new satellite constellations being launched into low Earth orbit, the growing use of medium Earth orbit for radio‐navigation and timing signals, slot region orbits for telecommunications and the introduction of electric propulsion to reach geostationary orbit, there is a growing need to develop services to protect satellites from space weather. Here we highlight two recent events in relation to satellite operations. We summarize 10 user needs that arose out of meetings with satellite operators, designers, underwriters and space agency staff. We present the satellite risk prediction and radiation forecast (SaRIF) system which is designed to meet most of these needs. The system uses real‐time data as input to the BAS radiation belt model (BAS‐RBM) to solve the Fokker Planck equation and provides a forecast of the electron flux throughout the outer radiation belt with 1‐hr resolution up to 24 hr ahead. The electron flux is used to calculate charging currents, and is combined with GOES near real time proton fluxes to calculate dose rate and total ionizing dose behind set levels of shielding for satellites in Medium Earth orbit, Geostationary orbit and slot region orbits. The results are compared against design standards and presented as risk indicators to forecast the risk of damage. The system works automatically and is updated every hour. We also present data and a best reconstruction of the radiation environment which are held in a searchable archive for satellite anomaly resolution. The SaRIF system is available via the European Space Agency space weather web portal.

Global Solar Radiation Forecasting using Artificial Neural Network and Support Vector Machine

Global solar radiation (GSoR) forecasting involves predicting future energy from the sun based on past and present data. Literature reveals that not all meteorological stations record solar radiation, some equipments are faulty, and are not available in every location due to high cost. Hence, the need to predict and forecast using predictors such as land surface temperature (LST). Satellite data when were used to complement ground-based stations have been yielding good results. Different artificial intelligence (AI) methods such as Support Vector Machine (SVM) and Artificial Neural Network (ANN) present different forecasting performances. Motivated by existing literature-related contradictions on the performance superiority of ANN and SVM in GSoR forecasting, the two techniques were compared based on several statistical tests. Experimental results show that ANN outperformed SVM by 2.9864% accuracy, making it superior in the forecast of GSoR.

Short-term forecasting of global solar irradiance in tropical environments with incomplete data

Electricity access is a common issue around the world. Many countries that face this problem are located in the tropical region and solar energy might be an alternative to mitigate this limitation. Accurate mechanisms for forecasting solar irradiance and insolation provide important information regarding the potential for generating solar energy. Furthermore, this data is relevant for the planning of renewable energy projects, and energy policy formulation. This research introduces a pipeline for the one-day ahead forecasting of solar irradiance and insolation that only requires solar irradiance historical data for training. Our approach includes a data imputation stage to handle missing data. In the prediction stage, we consider four data-driven approaches: Autoregressive Integrated Moving Average, Single Layer Feed Forward Network, Multiple Layer Feed Forward Network, and Long Short-Term Memory. The experiments are performed in a real-world dataset collected by 12 Automatic Weather Stations located in Nariño - Colombia. Our results show that the neural network-based models outperform Autoregressive Integrated Moving Average in most cases, and that Long Short-Term Memory exhibits better performance in cloudy environments (where more randomness is expected).

Evaluation of SaRIF High‐Energy Electron Reconstructions and Forecasts

AbstractIncreasing numbers of satellites are orbiting through the Earth's radiation belts, and the range of orbits being commonly used is also growing. As a result, there is an increasing need for services to help protect satellites from space weather. The Satellite RIsk prediction and radiation Forecast (SaRIF) system provides reconstructions and forecasts of the high‐energy electron flux throughout the outer radiation belt and translates these predictions into charging currents, dose rates, total ionizing dose and risk indicators. SaRIF both informs satellite operators of current and expected conditions and provides a tool to aid in post‐event analysis. The reconstructions and forecasts are provided by the British Antarctic Survey Radiation Belt Model (BAS‐RBM) running as part of an automatic system using real‐time data to specify the boundary conditions and drive processes within the physics‐based model. If SaRIF is to provide a useful tool, then the accuracy of the reconstructions and forecasts needs to be understood. Here we assess the accuracy of the simulations for geostationary orbit by comparing the model output with measurements made by the GOES 14 spacecraft for the period March–September 2019. No GOES 14 data was used to create the reconstruction or forecasts. We show that, with some improvements to the original system, the reconstructions have a prediction efficiency of 0.82 for >800 keV electrons and 0.87 for >2 MeV electrons, with corresponding prediction efficiencies of 0.59 and 0.78 for the forecasts.

Explicit basis function kernel methods for cloud segmentation in infrared sky images

Photovoltaic systems are sensitive to cloud shadow projection, which needs to be forecasted to reduce the noise impacting the intra-hour forecast of global solar irradiance. We present a comparison between different kernel discriminative models for cloud detection. The models are solved in the primal formulation to make them feasible in real-time applications. The performances are compared using the j-statistic. The infrared cloud images have been preprocessed to remove debris, which increases the performance of the analyzed methods. The use of the pixels’ neighboring features also leads to a performance improvement. Discriminative models solved in the primal yield a dramatically lower computing time along with high performance in the segmentation.

Short term solar irradiance forecasting using sky images based on a hybrid CNN–MLP model

High penetration of photovoltaics (PV) has been observed in the energy market over the last decade. However, its integration into electrical grids is challenging, as solar energy is highly fluctuating given its dependence on different weather variables. Consequently, short-term forecasting of solar irradiance provides a pivotal solution to ensure optimal use of the produced energy and reduce its uncertainty. This study proposes a hybrid convolutional neural network and Multilayer perceptron (CNN–MLP) model to forecast the global irradiance 15 min ahead. The model uses images from a hemispherical sky imager, time series of GHI, and weather variables collected from a ground meteorological station in Morocco. The evaluation of the proposed model under clear, mixed, and overcast days shows that the proposed model performs better than the persistence model. The root mean square error (RMSE) varies between 13.05 W/m2 and 49.16 W/m2 for CNN–MLP and between 45.76 W/m2 and 114.19 W/m2 for persistence. The coefficient of determination (R2) varies between 0.99 and 0.94 for the MLP–CNN and between 0.98 and 0.79 for persistence. The results show that the proposed model could be an appropriate choice for short-term forecasting even under cloudy conditions.

On the Similarity and Repeatability of Fast Radiation Belt Loss: Role of the Last Closed Drift Shell

AbstractProperly characterizing fast relativistic electron losses in the terrestrial Van Allen belts remains a significant challenge for accurately simulating their dynamics. In particular, magnetopause shadowing losses can deplete the radiation belt within hours or even minutes, but can have long‐lasting impacts on the subsequent belt dynamics. By statistically analyzing seven years of data from the entire Van Allen Probes mission in the context of the last closed drift shell, here we show how these losses are much more organized and predictable than previously thought. Once magnetic storm electron dynamics are properly analyzed in terms of the location of the last closed drift shell, not only is the loss shown to be repeatable but its energy‐dependent spatio‐temporal evolution is also revealed to follow a very similar pattern from storm to storm. Employing an energy‐dependent ULF wave radial diffusion model, we further show for the first time how the similar and repeatable fractional loss of the pre‐storm electron population in each storm can be reproduced and explained. Empirical characterization of this loss may open a pathway toward improved radiation belt specification and forecast models. This is especially important since underestimates of loss can also create unrealistic sources in models, creating phantom electron radiation and leading to the prediction of an overly harsh radiation environment.

Evaluation of ECMWF model to predict daily and monthly solar radiation over Indonesia region

Solar radiation forecast is a pivotal information needed in the operational activity of large-scale solar energy production. In this study, the reliability of SSRD (surface solar radiation downward) forecast from the 51 ensemble members in the ECMWF (European Centre for Medium Range Forecast) long-range forecast to predict daily and monthly radiation in 5 climatological stations in Indonesia is evaluated. The global horizontal irradiance (GHI) data from the solar radiation observation network from January 2018 – December 2020 are used in the quantitative evaluation of the SSRD forecast. Post-processing methods are applied to the model output, namely the bilinear interpolation method and the empirical quantile mapping to reduce consistent biases in the model output. The evaluation was carried out for different cloud covers based on the calculation of clearness index (k_t). The cloud condition affects the performance of the model, where the highest correlation value is achieved during sunny days (0.18 – 0.65) and the lowest correlation happens in overcast days (0.05 – 0.35). Models also tend to underestimate radiation when the sky is clear and overestimate it in cloudy days, based on negative MBE values during clear days (-0.47 kWh/m2 – -1.29 kWh/m2). The spatial averaging method did not necessarily improve the accuracy of the forecast, but the empirical quantile mapping method provides better accuracy, which is indicated by a values (mean error ratio) lower than 1 in most stations. Information about the influence of cloud cover on model performance can be used in future application of the model output and the bias correction process carried out in this study can be applied to reduce bias in the model.

Novel Cooperative Multi-Input Multilayer Perceptron Neural Network Performance Analysis with Application of Solar Irradiance Forecasting

To forecast solar irradiance with higher accuracy and generalization capability is challenging in the photovoltaic (PV) energy system. Meteorological parameters are highly influential in solar irradiance, leading to intermittent and randomicity. Forecasting using a single neural network model does not have sufficient generalization ability to achieve the optimal forecasting of solar irradiance. This paper proposes a novel cooperative multi-input multilayer perceptron neural network (CMMLPNN) to mitigate the issues related to generalization and meteorological effects. Authors develop a proposed forecasting neural network model based on the amalgamation of two inputs, three inputs, four inputs, five inputs, and six inputs associated multilayer perceptron neural network. In the proposed forecasting model (CMMLPNN), the authors overcome the variance based on the meteorological parameters. The amalgamation of five multi-input multilayer perceptron neural networks leads to better generalization ability. Some individual multilayer perceptron neural network-based forecasting models outperform in some situations, but cannot assure generalization ability and suffer from the meteorological weather condition. The proposed CMMLPNN (cooperative multi-input multilayer perceptron neural network) achieves better forecasting accuracy with the generalization ability. Therefore, the proposed forecasting model is superior to other neural network-based forecasting models and existing models.

Variational mode decomposition based random forest model for solar radiation forecasting: New emerging machine learning technology

Forecasting of solar radiation (Radn) can provide an insight vision for the amount of green and friendly energy sources. Owing to the non-linearity and non-stationarity challenges caused by meteorological variables in forecasting Radn, a variational mode decomposition method is integrated with simulated annealing and random forest (VMD-SA-RF) for resolving this problem. Firstly, the input parameters are separated into training and testing phases​ after generating a one-day ahead significant lags at (t – 1). Secondly, the variational mode decomposition is set to factorize multivariate meteorological data of train and test sets, independently, into their band-limited signals. Thirdly, the simulate annealing based feature selection system is engaged to select the best band-limited signals. Finally, using the pertinent band-limited signals, the daily Radn is forecasted via random forest (RF) model. The outcomes are benchmarked with other comparative models. The hybrid fusion VMD-SA-RF model is tested geographically in Australia, generates reliable performance to forecast Radn. The hybrid VMD-SA-RF system combining the pertinent meteorological features, as the model predictors have substantial implications for renewable and sustainable energy resource management.

IrradianceNet: Spatiotemporal deep learning model for satellite-derived solar irradiance short-term forecasting

The presence of clouds is widely identified as the primary uncertainty in current surface solar global horizontal irradiance (GHI) forecasts. Despite a wealth of historical satellite-derived irradiance observations, only limited research has investigated this problem from a purely data-driven perspective, something that has seen tremendous success in related domains such as radar- and satellite-based precipitation short-term forecasting. This paper presents IrradianceNet, a novel satellite-based neural network for spatiotemporal forecasting of surface solar irradiance up to 4 h in the future over Europe. Our method is fully data-driven and needs no post-processing or calibration based on sparse ground-based measurements of irradiance. We demonstrate superior forecasting performance compared to several persistence models, the TV-L1 algorithm, and ERA5 reanalysis data for satellite-derived solar irradiance using the European SARAH-2.1 dataset. We also validate these results using ground-based pyranometer observations from the Baseline Surface Radiation Network. Our conclusions remain unchanged when we account for hourly and monthly seasonality. Finally, applying a simple cloud mask scheme, we demonstrate that our performance improvement arises due to a considerable reduction in cloudy pixel errors. This is initial evidence that purely data-driven methods might better approximate and infer future cloud dynamics and their impact on surface solar irradiance.