Direct Normal Irradiance Prediction Research Articles

Intra-hour solar irradiance forecasting using topology data analysis and physics-driven deep learning

Real-time Direct Normal Irradiance (DNI) prediction is crucial for reliable and economic operation of Concentrated photothermal Solar Power (CSP) system in arid desert areas. However, the stochastic characteristics of short-term multidimensional meteorological time series make intra-hour DNI prediction a challenging task. In this study, we have proposed a deep learning model called TLD, which is combined with topological features captured by Topology Data Analysis (TDA) and temporal features captured by LSTM to address this challenge. Experimental results demonstrated that TLD outperformed the five latest models (Ridge, RF, C_GRU, BiLSTM, and GBRT) on seven solar radiation datasets in arid desert areas. Further analysis revealed that the proportion of cloudy days is a key factor affecting the model's performance. To enhance the forecast ability of TLD, we developed a physics-informed hybrid model named TLDP based on TLD and a smart persistence model, which fully combines the DNI prediction ability of TLD under cloudy conditions and that of the smart persistence model under sunny conditions. Experimental results of eight datasets collected from real-world solar photothermal power stations indicated that TLDP outperformed existing models, which may lay a foundation for more economical and stable operation of CSP plants in arid desert areas.

DNI Forecast Tool for the Smart Operation of a Parabolic Trough Collector System With Concrete Thermal Energy Storage

This work presents a basic forecast tool for predicting direct normal irradiance (DNI) in hourly resolution, which the Solar-Institut Jülich (SIJ) is developing within a research project. The DNI forecast data shall be used for a parabolic trough collector (PTC) system with a concrete thermal energy storage (C-TES) located at the company KEAN Soft Drinks Ltd in Limassol, Cyprus. On a daily basis, 24-hour DNI prediction data in hourly resolution shall be automatically produced using free or very low-cost weather forecast data as input. The purpose of the DNI forecast tool is to automatically transfer the DNI forecast data on a daily basis to a main control unit (MCU). The MCU automatically makes a smart decision on the operation mode of the PTC system such as steam production mode and/or C-TES charging mode. The DNI forecast tool was evaluated using historical data of measured DNI from an on-site weather station, which was compared to the DNI forecast data. The DNI forecast tool was tested using data from 56 days between January and March 2022, which included days with a strong variation in DNI due to cloud passages. For the evaluation of the DNI forecast reliability, three categories were created and the forecast data was sorted accordingly. The result was that the DNI forecast tool has a reliability of 71.4 % based on the tested days. The result fulfils SIJ’s aim to achieve a reliability of around 70 %, but SIJ aims to still improve the DNI forecast quality.

A Gridded Solar Irradiance Ensemble Prediction System Based on WRF-Solar EPS and the Analog Ensemble

The WRF-Solar Ensemble Prediction System (WRF-Solar EPS) and a calibration method, the analog ensemble (AnEn), are used to generate calibrated gridded ensemble forecasts of solar irradiance over the contiguous United States (CONUS). Global horizontal irradiance (GHI) and direct normal irradiance (DNI) retrievals, based on geostationary satellites from the National Solar Radiation Database (NSRDB) are used for both calibrating and verifying the day-ahead GHI and DNI predictions (GDIP). A 10-member ensemble of WRF-Solar EPS is run in a re-forecast mode to generate day-ahead GDIP for three years. The AnEn is used to calibrate GDIP at each grid point independently using the NSRDB as the “ground truth”. Performance evaluations of deterministic and probabilistic attributes are carried out over the whole CONUS. The results demonstrate that using the AnEn calibrated ensemble forecast from WRF-Solar EPS contributes to improving the overall quality of the GHI predictions with respect to an AnEn calibrated system based only on the deterministic run of WRF-Solar. In fact, the calibrated WRF-Solar EPS’s mean exhibits a lower bias and RMSE than the calibrated deterministic WRF-Solar. Moreover, using the ensemble mean and spread as predictors for the AnEn allows a more effective calibration than using variables only from the deterministic runs. Finally, it has been shown that the recently introduced algorithm of correction for rare events is of paramount importance to obtain the lowest values of GHI from the calibrated ensemble (WRF-Solar EPS AnEn), qualitatively consistent with those observed from the NSRDB.

Optimizing direct normal irradiance prediction of BP neural network based on Logistic Sparrow search algorithm

With the wide application of photothermal power stations in power systems, Direct Normal Irradiance(DNI) prediction is very important to improve the economic benefits of photothermal power stations and the utilization of solar energy. The short-term prediction of wind power by traditional Sparrow search algorithm (SSA) optimized BP(SSA-BP) neural network is prone to fall into local optimum, slow convergence rate, and low prediction accuracy. A short-term DNI prediction method based on the Logistic Sparrow search algorithm (Logistic-SSA) optimized BP(Logistic-SSA-BP) neural network is proposed. This paper takes a photothermal power station in Northwest China as the research object. Firstly, the Pearson correlation coefficient is introduced to analyze the environmental data set with a strong correlation with the direct illumination radiation index as the input of the model, so as to avoid redundant data affecting the prediction of the direct illumination radiation index. Secondly, the SSA algorithm and Logistic-SSA algorithm are used to improve BP neural network. Finally, the measured historical data of the photothermal power station is used to simulate the prediction models. The simulation results show that the prediction model based on Logistic-SSA-BP has better prediction accuracy than SSA-BP, and is more in line with the production and operation requirements of the photothermal power station.

Fuzzy inference systems based on multi-type features fusion for intra-hour solar irradiance forecasts

The cloud cover information in ground-based cloud (GBC) images is important to direct normal irradiance (DNI) prediction. In order to obtain better performance, the DNI forecasting model needs to incorporate the features from GBC images and numerical time series. In this paper, to resolve the issues of feature fusion with various types, three novel fuzzy inference systems (FIS) models are proposed for intra-hour DNI prediction. The features of the GBC image and the numerical time series are firstly fuzzified by clustering and grid partition, respectively. Then, the hybrid fuzzification and optimization method lead to the presented two hierarchical fuzzy inference systems (HFIS) models and an adaptive neuro-fuzzy inference system (ANFIS) model for DNI prediction. The performance of the proposed models is validated with the data from the National Renewable Energy Laboratory (NREL) from January 1, 2018 to December 31, 2018. Experiments demonstrate that the proposed models outperform the reference model, and the best model is capable of achieving 19.66% improvement over the reference model for 10-minute ahead DNI prediction.

Ensemble model output statistics for the separation of direct and diffuse components from 1-min global irradiance

Separation models split diffuse and direct components of solar radiation from the global horizontal radiation. At the moment, all separation models only issue predictions that are deterministic (as opposed to probabilistic). Since the best prediction is necessarily probabilistic, a parametric post-processing framework called the ensemble model output statistics (EMOS) is introduced in this paper, to make probabilistic predictions. EMOS takes the diffuse fractions predicted by an ensemble of existing 1-min separation models, and outputs a predictive distribution, with parameters optimized by maximum likelihood estimation. Clearly, the EMOS-based separation modeling goes beyond the current literature, in terms of uncertainty quantification.Eight popular separation models from the literature, with different architectures, are used to demonstrate the predictive power of EMOS. Using 1-min high-quality radiometric data from seven stations in the USA and four stations in Europe, it is found that Yang2 is the best stand-alone model with an average RMSE of 21.8%, in terms of direct normal irradiance prediction, contrasting the 26.3% of the previously reported best model, namely, Engerer2. On the other hand, the EMOS post-processed predictions have an average RMSE of 20.8%, which is lower than that of the best stand-alone model. Moreover, EMOS is shown superior to simple model averaging, in terms of continuous ranked probability score and ignorance score.

Assessment of Direct Normal Irradiance Forecasts Based on IFS/ECMWF Data and Observations in the South of Portugal

Direct Normal Irradiance (DNI) predictions obtained from the Integrated Forecasting System of the European Centre for Medium-Range Weather Forecast (IFS/ECMWF) were compared against ground-based observational data for one location at the south of Portugal (Évora). Hourly and daily DNI values were analyzed for different temporal forecast horizons (1 to 3 days ahead) and results show that the IFS/ECMWF slightly overestimates DNI for the period of analysis (1 August 2018 until 31 July 2019) with a fairly good agreement between model and observations. Hourly basis evaluation shows relatively high errors, independently of the forecast day. Root mean square error increases as the forecast time increases with a relative error of ~45% between the first and the last forecast. Similar patterns are observed in the daily analysis with comparable magnitude errors. The correlation coefficients between forecast and observed data are above 0.7 for both hourly and daily data. A methodology based on a new DNI attenuation Index (DAI) was developed to estimate cloud fraction from hourly values integrated over a day and, with that, to correlate the accuracy of the forecast with sky conditions. This correlation with DAI reveals that in IFS/ECMWF model, the atmosphere as being more transparent than reality since cloud cover is underestimated in the majority of the months of the year, taking the ground-based measurements as a reference. The use of the DAI estimator confirms that the errors in IFS/ECMWF are larger under cloudy skies than under clear sky. The development and application of a post-processing methodology improves the DNI predictions from the IFS/ECMWF outputs, with a decrease of error of the order of ~30%, when compared with raw data.

Direct normal irradiance predictions using broadband models for Indian stations

In this paper three parametric models CPCR2, MLWT2 and REST have been used to compute direct normal irradiance (DNI) for five Indian stations, namely New Delhi, Pune, Jaipur, Kolkata and Mumbai. Computed values of DNI have been compared with measured values in terms of percentage root mean square error (RMSE) and percentage mean bias error (MBE). It is observed that the average percentage RMSE for the year is the minimum for MLWT2 model for Jaipur followed by New Delhi and Pune and the values are 1.67%, 2.33% and 2.49%, respectively. But for coastal stations Kolkata and Mumbai the corresponding minimum values occur for CPCR2 model and REST model respectively, and the values are 2.62% and 2.76%, respectively. It shows that MLWT2 model performs better for most of the Indian stations except the coastal stations.

3D-CNN-based feature extraction of ground-based cloud images for direct normal irradiance prediction

Cloud cover and cloud motion have a large impact on solar irradiance. One of the effective ways for direct normal irradiance (DNI) prediction is to use cloud features, which has been extensively studied. A Convolutional Neural Network (CNN) has the advantage of automatic features extraction by using strong computing capabilities. In this paper, a novel 3D-CNN method is proposed by processing multiple consecutive ground-based cloud (GBC) images in order to extract cloud features including texture and temporal information. The resulting features and the DNI data are then used to establish a DNI forecasting model. The experiments are carried out to evaluate the performance of the proposed forecasting method by using the data from January 1, 2013 to December 31, 2014. The experimental results show that the proposed method coupled with the multilayer perceptron (MLP) model achieves forecast skill of 17.06% for 10-minute ahead DNI prediction.

Real-time Clear-sky Model and Cloud Cover for Direct Normal Irradiance Prediction

Direct normal irradiance (DNI) prediction is of great significance for the concentrated solar power (CSP) generation and grid work. The clouds are the main cause of DNI intensity reduction and drastic changes. Therefore, the real-time clear-sky DNI model coupled with cloud cover is proposed for short-term DNI prediction in this study. The Linke turbidity coefficient is used to develop a clear-sky model, and real-time adjustment of the coefficients is achieved by identifying clear-sky period. Then, the theoretical clear-sky value of current day can be obtained. By the combination of the historical cloud cover and theoretical clear-sky value, Auto Regressive Moving Average (ARMA) and Artificial Neural Network (ANN) are used to develop linear and nonlinear models. Experiment with the data in the National Renewable Energy Laboratory (NREL) database, the simulation results show that this approach using cloud cover and clear-sky information can improve the forecasting accuracy.

Intra-hour direct normal irradiance forecasting through adaptive clear-sky modelling and cloud tracking

Hybrid fossil-solar power cycles possess high solar-to-electric conversion efficiency and the potential to provide stable power output without the need for costly storage systems. However, the power output of the concentrating solar power (CSP) component of the plant can fluctuate in sync with intermittent direct normal irradiance (DNI). For hybrid plants, the controllable fossil-fuel unit can be used to compensate for shortfalls in CSP output during periods of light DNI intermittency or provide the entire plant output during periods of high DNI intermittency. This fossil-fuel unit requires 5–10 min to ramp its power output up or down to perform these balancing functions. Thus, by accurately predicting future DNI, and hence CSP output, the fossil-fuel unit can guarantee stable plant-wide power output during periods of intermittency.This paper develops an intra-hour DNI prediction system using a ground-based cloud motion vector (CMV) framework and real-time DNI measurements. This system comprises a clear-sky DNI model and a cloud fraction prediction algorithm. The presented CSM is based on the Ineichen model (Ineichen, 2008), where the model parameters are adaptively estimated from identified clear-sky DNI measurements over a moving window. For the cloud fraction prediction model, this paper presents an enhanced “sector-ladder” method (Quesada-Ruiz et al., 2014) that uses the weighted mean of circular quantities (Fisher, 1995) and autoregressive filtering to improve cloud flow predictions. Furthermore, a method to forewarn against periods of high DNI intermittency using the generated DNI predictions is presented.The proposed DNI prediction system is evaluated using 37 days of sky-camera images and DNI data collected over the summer of 2014/2015 at the University of Queensland. Over all test days, the adaptive CSM has an average root mean square error of 3.06%, which represents a 19% improvement over a CSM that uses the optimal model parameters from the previous day’s data. Additionally, the modifications to the cloud flow prediction algorithm (the sector-ladder method) are shown to improve the cloud velocity prediction accuracy by a factor of seven over a period of visually determined constant cloud velocity. We find the overall prediction accuracy of the DNI prediction system to be statistically similar to the accepted short-term benchmark of persistence; however, it performs more consistently over a range of weather conditions and is able to forewarn against periods of impending intermittency with 93% accuracy. The latency from data collection to prediction is less than 30 s, making the method eminently suitable for real-time applications.

Use of Image-Based Direct Normal Irradiance Forecasts in the Model Predictive Control of a Solar-Thermal Reactor

A model predictive control (MPC) system for a solar-thermal reactor was developed and applied to the solar-thermal steam-gasification of carbon. The controller aims at rejecting the disturbances in solar irradiance, caused by the presence of clouds. Changes in solar irradiance are anticipated using direct normal irradiance (DNI) forecasts generated using images acquired through a Total Sky Imager (TSI). The DNI predictor provides an estimation of the disturbances for the control algorithm, for a time horizon of 1 min. The proposed predictor utilizes information obtained through the analysis of sky images, in combination with current atmospheric measurements, to produce the DNI forecast. The predictions of the disturbances are used, in combination with a dynamic model of the process, to determine the required control moves at every time step. The performance of the proposed DNI predictor-controller scheme was compared to the performance of an equivalent MPC that does not use DNI forecasts in the calculation of the control signals. In addition, the performance of a controller fed with perfect DNI predictions was also evaluated.