Simulation of Daily Diffuse Solar Radiation Based on Three Machine Learning Models

Solar radiation is an important parameter in the fields of computer modeling,engineering technology and energy development.This paper evaluated the ability of three machine learning models,i.e.,Extreme Gradient Boosting(XGBoost),Support Vector Machine(SVM)and Multivariate Adaptive Regression Splines(MARS),to estimate the daily diffuse solar radiation(Rd).The regular meteorological data of 1966-2015 at five stations in China were taken as the input parameters(including mean average temperature(Ta),theoretical sunshine duration(N),actual sunshine duration(n),daily average air relative humidity(RH),and extra-terrestrial solar radiation(Ra)).And their estimation accuracies were subjected to comparative analysis.The three models were first trained using meteorological data from 1966 to 2000.Then,the 2001-2015 data was used to test the trained machine learning model.The results show that the XGBoost had better accuracy than the other two models in coefficient of determination(R2),root mean square error(RMSE),mean bias error(MBE)and normalized root mean square error(NRMSE).The MARS performed better in the training phase than the testing phase,but became less accurate in the testing phase,with the R2 value falling by 2.7-16.9%on average.By contrast,the R2 values of SVM and XGBoost increased by 2.9-12.2%and 1.9-14.3%,respectively.Despite trailing slightly behind the SVM at the Beijing station,the XGBoost showed good performance at the rest of the stations in the two phases.In the training phase,the accuracy growth is small but observable.In addition,the XGBoost had a slightly lower RMSE than the SVM,a signal of its edge in stability.Therefore,the three machine learning models can estimate the daily Rd based on local inputs and the XGBoost stands out for its excellent performance and stability.

Prediction of the Diffuse Solar Energy on Horizontal at Different Selected Locations

The main objective of this paper is to predict the diffuse solar energy on a horizontal surface by using data of global solar energy (H) and diffuse solar energy (Hd) at different selected geographical locations in Saudi Arabia during the period time from 1980 to 2019. The low values of the root mean square error RMSE for all correlations indicated a good agreement between the measured and calculated values of Hd. The negative values of mean percentage error MPE % for all models show that for all locations, the proposed correlations slightly overestimate Hd, and the absolute values of MPE never reach 1.35%. The first, second and third order correlations between the diffuse solar fraction Hd/H and the clearness index Kt and between the diffuse transmittance Hd/H0 and the sunshine hours have been proposed for the selected locations using the method of regression analysis. The differences between the measured and calculated values of Hd show that a first order correlation between Hd/H and Kt can be used for estimating Hd at the present locations with good accuracy. However, second order correlations between Hd/H or Hd/H0 and S/So are recommended for estimating Hd at these locations. The average annual differences between measured and calculated values of diffuse solar energy Hd on horizontal at selected sites in the present research are discussed.

Study on the evaluation method of diffuse radiation models——taking 5 typical cities in China as examples

Diffuse solar radiation models play an extremely important role in solar photothermal utilization,resource assess-ment and energy consumption simulation,etc.The accuracy of these diffuse solar radiation models usually need to be evaluated by various statistical parameters.Among these statistical parameters,the Global Performance In-dex(GPI)has been extensively employed in recent years because of its comprehensiveness and wide applicability.This paper takes five cities in China as representatives of 5 typical climate regions,and 12 solar scattered radia-tion models are fitted with the meteorological data of 5 cities.Based on the comparative analysis of the existing GPI calculation methods,this paper points out that there are some defects in the existing GPI,and modifies the existing GPI based on the comprehensive consideration of statistical parameters,normalization preprocessing of statistical parameters,unified evaluation direction of parameters,weight redistribution of statistical parameters,and adjustment of extreme coefficient.12 types of new GPI are established in this paper,and the performance of diffuse solar radiation models are compared based on these GPI.The rationality of GPI corrective measures is analyzed by means of the method reasonable index(MRI).The results show that the GPI calculation method(N10)which takes five corrective measures has the best performance,and the accuracy of the existing GPI can be improved by 13.33 to 65%.

Distribution and temporal variability of the solar resource at a site in south-east Norway

Globally, solar energy is expected to play a significant role in the changing face of energy economies in the near future. However, the variability of this resource has been the main barrier for solar energy development in most locations around the world. This paper investigated the distribution and variability of solar radiation using the a 10-year （2006 to 2015） data collected at Sorhs meteor- ological station located at latitude 59° 39＇ N and longitude 10° 47＇E, about 93.3 m above sea level （about 30 km from Oslo）, in south-eastern part of Norway. It is found that on annual basis, the total number of days with a global solar radiation of less than 1 kWh/（m2.d） is 120 days while the total number of days with an expected global solar radiation greater than 3 kWh/（m2.d） is 156 days （42.74%） per year. The potential energy output from a horizontally placed solar collector in these 156 days is approximately 75% of the estimated annual energy output. In addition, it is found that the inter-annual coefficient of variation of the global solar radiation is 4.28%, while that of diffuse radiation is 4.96%.

Research exploring greenhouse environment control over the last 50 years

Environments do not exist in isolation.Their main components in greenhouse systems are plants.Without consideration of plants,analysis of greenhouse environments and environmental control of greenhouses can be accomplished,although it is not simple to achieve.Initial attempts were undertaken to analyze greenhouse environments and then reproduce them.Ventilation rate effects on plant photosynthesis in a growth chamber were reported in 1966.Computer simulations then became a main subject of research.The first dynamic computer simulation of a greenhouse environment including plants was published in 1971.According to innovations of computer technology,the use of minicomputers and microcomputers spread in many areas.By measuring the net photosynthesis of lettuce plants grown under artificial lighting,air temperature was optimized using a minicomputer with the hill-climbing method.The method was designated as the Speaking Plant Approach to environment control(SPA).After the author developed the first reported environmental control system in Japan,systems using microcomputers spread widely for greenhouse environmental control.Knowledge-based expert systems were tested for plant management.Also,a machine vision system was developed to detect critical moments for watering of muskmelon plants.The first feed-forward control method for greenhouses with a large heat mass was reported.Then space farming was tested in 1996 to assess gravity effects on plants.Energy-saving aspects such as solar sterilization,ground heat storage system,and storage using phase change material(PCM)have been reported.Defects of ordinary solarimeters were reported in 2008 along with an approach to estimate evapotranspiration in a greenhouse without the effect of so-called cosine law.Later,this technique was expanded to estimate photosynthesis of the plant canopy in a greenhouse using newly developed sensor units.

A unified model of bidirectional reflectance distribution function for the vegetation canopy

An accurate and operational bidirectional reflectance distribution function （BDRF） canopy model is the basis of quantitative vegetation remote sensing. The canopy reflectance should be approximated as the sum of the single scattering reflectance arising from the sun, pl, and the multiple scattering reflectance arising from the canopy, fin, as their directional characteristics are dramatically different. Based on the existing BRDF model, we obtain a new analytical expression of ρ1 and ρm in this paper, which is suitable for different illumination conditions and different vegetation canopies. According to the geometrical optic model at the leaf scale, the anisotropy of ρ1 can be ascribed to the geometry of the object, sun and the sensor, multiple scale clumping, and the fraction of direct solar radiation and diffuse sky radiation. Then, we parameterize the area ratios of four components： the sunlit foliage, sunlit ground, shadow foliage and shadow ground based on a Poisson distribution, and develop a new approximate analytical single scattering reflectance model. Assuming G=0.5, a recollision probability theory based scattering model is developed which considers the effects of diffuse sky radiation, scattering inside the canopy and rebounds between the canopy and soil. Validation using ground measurements of maize and black spruce forest proves the reliability of the model.