A multi-step ahead global solar radiation prediction method using an attention-based transformer model with an interpretable mechanism

The conventional multi-step ahead solar radiation prediction method ignores the time-dependence of a future solar radiation time series. Therefore, according to sequence-to-sequence (seq2seq) model theory, this paper proposes the seq2seq long- and short-term memory model (seq2seq-LSTM), the seq2seq-LSTM model with an attention mechanism (seq2seq-at-LSTM), and a transformer model, which consists only of the attention mechanism. The hourly global solar radiation data between 2016 and 2018 from Shaanxi, China, is used to train and validate the models. The results show that the introduction of the attention mechanism can effectively improve the prediction accuracy of the seq2seq-LSTM model. However, the model is still not very good at capturing the long-distance dependence of the solar radiation time series due to the inherent properties of LSTM. In comparison, the transformer model, which is based entirely on the attention mechanism, performs much better at capturing the long-distance dependence of the solar radiation time series. Furthermore, as the number of time-steps increases, the performance of the solar radiation prediction decreases relatively smoothly and slowly. The obtained average coefficient of determination, root mean square error (RMSE), relative RMSE, and mean bias error are 0.9788, 72.91 W/m², 25.25%, and 38.35 W/m², respectively. In addition, the average skill score of the transformer model is around 44.9%, which is 20.54% higher than that of the seq2seq-at-LSTM model and about 40.84% higher than that of the seq2seq-LSTM model. Besides, the use of the attention mechanism can explain the improved prediction compared to other models. This model developed in this study could also be used for predictions in other fields, such as wind energy predictions and building energy predictions.     

Prediction of solar radiation with genetic approach combing multi-model framework

In this paper, a genetic approach combing multi-model framework for solar radiation time series prediction is proposed. The framework starts with the assumption that there exists several different patterns in the stochastic component of the solar radiation series. To uncover the underlying pattern, a genetic algorithm is used to segment the solar series dynamically, and the subsequences are further grouped into different clusters. For each cluster, a prediction model is trained to represent that specific pattern. In the prediction phase, identifying the pattern for current period is of great importance. Thus a procedure for the pattern identification is performed to identify the proper pattern for the series belong to. The prediction result of the proposed framework is then compared to other algorithms. It shows that the proposed framework could provide superior performance compared to others.     

Testing for nonlinearity in solar radiation time series by a fast surrogate data test method

Recently, various time series models have been proposed to predict solar radiation, for instance the ARIMA (autoregressive integrated moving average) model and neural networks. Before building a model for the data, however, it is advisable to check whether the data suggest this type of modeling. More specifically, a nonlinearity test is suggested before further analysis with the linear or nonlinear tools are to be applied. In this paper, we test the presence of nonlinearity in the solar radiation time series by the method of surrogate data. The surrogate test method used in this paper is based on evaluation of the differences between the original time series and the linear model that best approximates it. Nonlinearity tests are carried out for four data sets including 5-min, hourly, daily and monthly global solar radiation time series from the UO (University of Oregon) Solar Radiation Monitoring Laboratory. The test statistics show that the 5-min, hourly, daily global solar radiation time series exhibit apparently nonlinearity while the monthly time series does not.     

Effects of air pollution for estimating global solar radiation in India

Solar energy is the primary resource for all biological, chemical and physical processes. The amount of global solar radiation is an important parameter for solar energy applications. It is common to estimate a monthly average of daily global solar radiation using different regression models. These models in turn exploit the correlation between solar radiation and various atmospheric factors. These factors are commonly derived from meteorological, geographical and climatological data that are readily available for majority of weather stations across the world. In this paper, a novel regression model that can predict location-independent daily global solar radiation is presented. The proposed exponential quadratic model captures the correlation between measured global solar radiation values, sunshine hour and Air Pollution Index for Indian cities. In addition to this, an extended study of several other regression models (e.g. linear, quadratic, exp.-linear and exp.-quadratic) is also presented. This analysis with real data from Indian cities suggests that air pollution is a more significant factor than location when predicting solar radiation. Finally, the model parameters (regression coefficients) for each model are listed out. Additionally, the generalised model equation for the best performing model is also presented.     

Smart model for accurate estimation of solar radiation

Prediction of solar radiation has drawn increasing attention in the recent years. This is because of the lack of solar radiation measurement stations. In the present work, 14 solar radiation models have been used to assess monthly global solar radiation on a horizontal surface as function of three parameters: extraterrestrial solar irradiance (G₀), duration sunshine (S) and daylight hours (S₀). Since it has been observed that each model is adequate for some months of the year, one model cannot be used for the prediction of the whole year. Therefore, a smart hybrid system is proposed which selects, based on the intelligent rules, the most suitable prediction model of the 14 models listed in this study. For the test and evaluation of the proposed models, Tamanrasset city, which is located in the south of Algeria, is selected for this study. The meteorological data sets of five years (2000–2004) have been collected from the Algerian National Office of Meteorology (NOM), and two spatial databases. The results indicate that the new hybrid model is capable of predicting the monthly global solar radiation, which offers an excellent measuring accuracy of R² values ranging from 93% to 97% in this location.     

Prediction of monthly global solar radiation using adaptive neuro fuzzy inference system (ANFIS) technique over the State of Tamilnadu (India): a comparative study

Enormous potential of solar energy as a clean and pollution free source enrich the global power generation. India, being a tropical country, has high solar radiation and it lies to the north of equator between 8°4′ &; 37°6′ North latitude and 68°7′, and 97°5′ East longitude. In southindia, Tamilnadu is located in the extreme south east with an average temperature of gerater than 27.5° (> 81.5 F). In this study, an adaptive neuro-fuzzy inference system (ANFIS) based modelling approach to predict the monthly global solar radiation(MGSR) in Tamilnadu is presented using the real meteorological solar radiation data from the 31 districts of Tamilnadu with different latitude and longitude. The purpose of the study is to compare the accuracy of ANFIS and other soft computing models as found in literature to assess the solar radiation. The performance of the proposed model was tested and compared with other earth region in a case study. The statistical performance parameters such as root mean square error (RMSE), mean bias error (MBE), and coefficient of determination (R2) are presented and compared to validate the performance. The comparative test results prove the ANFIS based prediction are better than other models and furthermore proves its prediction capability for any geographical area with changing meterological conditions.     

基于CNN-Bi-LSTM的太阳辐照度超短期预测

针对太阳辐射引起光伏出力的不确定性和波动性,进而造成大量光伏发电并网时对电网稳定性和安全的危害,提出一种新的太阳辐射超短期预测方法.该方法通过构建一维卷积神经网络,对多个关键气象变量进行数据融合和特征转换,然后构造双向长短期记忆网络预测模型,实现对未来15 min的太阳总辐照度的超短期预测.实验结果表明,所提出的预测模...     

Estimation of monthly solar radiation distribution for solar energy system analysis

The concept of probability density frequency, which is successfully used for analyses of wind speed and outdoor temperature distributions, is now modified and proposed for estimating solar radiation distributions for design and analysis of solar energy systems. In this study, global solar radiation distribution is comprehensively analyzed for photovoltaic (PV) panel and thermal collector systems. In this regard, a case study is conducted with actual global solar irradiation data of the last 15 years recorded by the Turkish State Meteorological Service. It is found that intensity of global solar irradiance greatly affects energy and exergy efficiencies and hence the performance of collectors.     

A grey‐box model of next‐day building thermal load prediction for energy‐efficient control

Accurate building thermal load prediction is essential to many building energy control strategies. To get reliable prediction of the hourly building load of the next day, air temperature/relative humidity and solar radiation prediction modules are integrated with a grey‐box model. The regressive solar radiation module predicts the solar radiation using the forecasted cloud amount, sky condition and extreme temperatures from on‐line weather stations, while the forecasted sky condition is used to correct the cloud amount forecast. The temperature/relative humidity prediction module uses a dynamic grey model (GM), which is specialized in the grey system with incomplete information. Both weather prediction modules are integrated into a building thermal load model for the on‐line prediction of the building thermal load in the next day. The validation of both weather prediction modules and the on‐line building thermal load prediction model are presented. Copyright © 2008 John Wiley & Sons, Ltd.     

临近空间太阳能无人机能量调度策略

针对临近空间太阳能无人机因表面太阳辐射不均匀及负荷不确定性导致的能源系统容量无法充分利用的问题,提出一种实时能量优化调度策略,旨在最小化每个光储模块在24 h周期末电量差值、太阳落山时刻最大限度地储存能量,通过构建调度模型和能量前移修正策略达到充分利用发电、储能设备装机容量的目的。利用Python软件对该模型及调度策略进行仿真验证,证实了该能量调度策略的合理性和可行性。     

Prediction of global solar irradiance based on time series analysis: Application to solar thermal power plants energy production planning

Due to strong increase of solar power generation, the predictions of incoming solar energy are acquiring more importance. Photovoltaic and solar thermal are the main sources of electricity generation from solar energy. In the case of solar thermal energy plants with storage energy system, its management and operation need reliable predictions of solar irradiance with the same temporal resolution as the temporal capacity of the back-up system. These plants can work like a conventional power plant and compete in the energy stock market avoiding intermittence in electricity production.This work presents a comparisons of statistical models based on time series applied to predict half daily values of global solar irradiance with a temporal horizon of 3 days. Half daily values consist of accumulated hourly global solar irradiance from solar raise to solar noon and from noon until dawn for each day. The dataset of ground solar radiation used belongs to stations of Spanish National Weather Service (AEMet). The models tested are autoregressive, neural networks and fuzzy logic models. Due to the fact that half daily solar irradiance time series is non-stationary, it has been necessary to transform it to two new stationary variables (clearness index and lost component) which are used as input of the predictive models. Improvement in terms of RMSD of the models essayed is compared against the model based on persistence. The validation process shows that all models essayed improve persistence. The best approach to forecast half daily values of solar irradiance is neural network models with lost component as input, except Lerida station where models based on clearness index have less uncertainty because this magnitude has a linear behaviour and it is easier to simulate by models.     

基于多维特征分析的双层协同太阳辐照度预测

为增强逐日太阳辐照度预测的准确性和普适性,提出一种基于多维特征分析的双层协同预测模型。首先,搭建一种双层协同架构,将整个模型分成基准层和提升层两部分,使用分层预测的方式追踪目标对象的多维特征和变化趋势;其次,以数值天气预报（NWP）为输入,采用LightGBM基于特征学习预测方法构建基准预测模型;然后,在前者的基础上,挖掘目标时刻太阳辐照度与历史时序数据之间的关联性,引入改进AdaBoost算法与多隐层极限学习机（MH-ELM）作为提升层主体,提高时序预测的稳定性;最后,选用中国中部地区某光伏电站实测太阳辐照度数据进行算例分析,验证了该模型的合理性和有效性。     

Using probabilistic finite automata to simulate hourly series of global radiation

A model to generate synthetic series of hourly exposure of global radiation is proposed. This model has been constructed using a machine learning approach. It is based on the use of a subclass of probabilistic finite automata which can be used for variable-order Markov processes. This model allows us to represent the different relationships and the representative information observed in the hourly series of global radiation; the variable-order Markov process can be used as a natural way to represent different types of days, and to take into account the “variable memory” of cloudiness. A method to generate new series of hourly global radiation, which incorporates the randomness observed in recorded series, is also proposed. As input data this method only uses the mean monthly value of the daily solar global radiation. We examine if the recorded and simulated series are similar. It can be concluded that both series have the same statistical properties.     

A simplified model for generating sequences of global solar radiation data for isolated sites: Using artificial neural network and a library of Markov transition matrices approach

The purpose of this work is to develop a hybrid model which will be used to predict the daily global solar radiation data by combining between an artificial neural network (ANN) and a library of Markov transition matrices (MTM) approach. Developed model can generate a sequence of global solar radiation data using a minimum of input data (latitude, longitude and altitude), especially in isolated sites. A data base of daily global solar radiation data has been collected from 60 meteorological stations in Algeria during 1991–2000. Also a typical meteorological year (TMY) has been built from this database. Firstly, a neural network block has been trained based on 60 known monthly solar radiation data from the TMY. In this way, the network was trained to accept and even handle a number of unusual cases. The neural network can generate the monthly solar radiation data. Secondly, these data have been divided by corresponding extraterrestrial value in order to obtain the monthly clearness index values. Based on these monthly clearness indexes and using a library of MTM block we can generate the sequences of daily clearness indexes. Known data were subsequently used to investigate the accuracy of the prediction. Furthermore, the unknown validation data set produced very accurate prediction; with an RMSE error not exceeding 8% between the measured and predicted data. A correlation coefficient ranging from 90% and 92% have been obtained; also this model has been compared to the traditional models AR, ARMA, Markov chain, MTM and measured data. Results obtained indicate that the proposed model can successfully be used for the estimation of the daily solar radiation data for any locations in Algeria by using as input the altitude, the longitude, and the latitude. Also, the model can be generalized for any location in the world. An application of sizing PV systems in isolated sites has been applied in order to confirm the validity of this model.     

Estimation of global solar irradiation on horizontal and inclined surfaces based on the horizontal measurements

Solar radiation data are essential in the design of solar energy conversion devices. In this regard, empirical models were selected to estimate the global solar radiation on horizontal and inclined surfaces. The hourly solar radiation data measured at the study area during the period of 2004-2007, were used to calculate solar radiations using selected models. The selected models were compared on the basis of statistical methods. Based on the results, a new model, H/H_o = 0.19490 + 0.4771(n/N) + 0.02994 exp(n/N) has been developed, based on Kadir Bakirci linear exponential model. This is highly recommended to estimate monthly mean daily global solar irradiation, on a horizontal surface. Further, a model to convert horizontal solar global radiation to that of radiation on a tilted surface is also presented. It is based upon a relatively simple model proposed by Olmo et al. which requires only measurements of horizontal solar radiation. The developed model appears to give excellent results and has the advantage of being relatively simple for applications. The present work will help to improve the state of knowledge of global solar radiation to the point where it has applications in the estimation of global solar radiation, both on horizontal and inclined surfaces.     

Prediction of hourly solar radiation using a novel hybrid model of ARMA and TDNN

In this work, a new approach that contains two phases is used to predict the hourly solar radiation series. In the detrending phase, several models are applied to remove the non-stationary trend lying in the solar radiation series. To judge the goodness of different detrending models, the Augmented Dickey-Fuller method is applied to test the stationarity of the residual. The optimal model is used to detrend the solar radiation series. In the prediction phase, the Autoregressive and Moving Average (ARMA) model is used to predict the stationary residual series. Furthermore, the controversial Time Delay Neural Network (TDNN) is applied to do the prediction. Because ARMA and TDNN have their own strength respectively, a novel hybrid model that combines both the ARMA and TDNN, is applied to produce better prediction. The simulation result shows that this hybrid model can take the advantages of both ARMA and TDNN and give excellent result.     

基于改进LSSVM的太阳辐射量预测

为准确预测太阳辐射量,提出一种基于变分模态分解和粒子群优化算法的最小二乘支持向量机组合预测模型。针对太阳辐射量序列具有不稳定性的特点,首先利用变分模态分解将历史太阳辐射量数据分解成一系列相对稳定的分量序列,再应用粒子群优化最小二乘支持向量机参数,以预测各分量序列,将各分量太阳辐射量预测值集成,从而得到最终太阳辐射量预测值。实例分析和对比研究表明,该模型预测太阳辐射量有效可行,具有较高的预测精度。研究成果可为太阳辐射量预测提供参考。     

Solar constant values for estimating solar radiation

There are many solar constant values given and adopted by researchers, leading to confusion in estimating solar radiation. In this study, some solar constant values collected from literature for estimating solar radiation with the Ångström-Prescott correlation are tested in China using the measured data between 1971 and 2000. According to the ranking method based on the t-statistic, a strategy to select the best solar constant value for estimating the monthly average daily global solar radiation with the Ångström-Prescott correlation is proposed.     

Measurements and predictions of solar radiation incident on horizontal surfaces at Santa Fe, Argentina (31° 39′S, 60° 43′W)

Measurements and predictions of solar radiation during a period of 10 years on horizontal surfaces at Santa Fe (31° 39′ S, 60° 43′ W), Argentina, reported as average daily global radiation for each month, are presented. Data are compared to those obtained with a previously published and verified model for computing solar radiation on horizontal planes at the earth's surface for cloudless sky days. Measurements show an important reduction of global radiation with respect to the cloudless sky model predictions for all months of the year. Conversely, averaged daily diffuse solar radiation calculated with Page's formula shows a small increment with respect to the predicted diffuse solar radiation for cloudless sky conditions. When direct solar radiation data, calculated from global and diffuse solar radiation values, are compared to theoretical prediction, a significant decrease is observed. This trend is similar to that obtained for global solar radiation.     

Simple model for the generation of daily global solar-radiation data in Turkey

Modelling, performance analysis, and designing of solar energy systems depend on solar radiation data. In this study, a simple model for estimating the daily global radiation is developed. The model is based on a trigonometric function, which has only one independent parameter, namely the day of the year. The model is tested for 68 locations in Turkey using the data measured during at least 10 years. It is seen that predictions from the model agree well with the long-term measured data. The predictions are also compared with the data available in literature for Turkey. It is expected that the model developed for daily global solar radiation will be useful to the designers of energy-related systems as well as to those who need to estimates of yearly variation of global solar-radiation for any specific location in Turkey.