R&D Project 3 – The solar resource tool from landsat vs ground data project

We have different types of models to calculate the direct radiation of solar from ground data and the other one is the data of satellite that we can use for solar radiation measurement. In this report we have compared the different tools from ground data and LandSat. 

Ground Data

One of the ways of getting accurate and reliable global radiation data for alternative energy system design is by ground measurements at the location of interest. Ground measurements are typically pinpoint measurements which are temporally integrated. This involves installation of solar sensor like pyranometer for continuous, long-term measurements of solar data. Compared to the measurements of other meteorological parameters, the equipment for radiation measurements is extremely expensive and requires experts for operation and maintenance. Although ground measure data are said to be accurate and reliable, the price implication and technicality involved have made such data unavailable in many locations. This has led to the rummage around for alternative means of getting solar data for research and development of alternative energy systems. 

GHI=DNI∗(cos(zenith))+DHI 

Global Horizontal Irradiance (GHI) is usually measured by (i) thermocouple based pyranometers or (ii) silicon photodiode cells. For development and monitoring of solar power plants, it’s advised to use high-standard meteorological pyranometers to attain the very best possible accuracy and stability of measurements. Direct Normal Irradiance (DNI) is measured by pyrheliometers, where the instrument always aims directly at the sun by continuously sun tracking mechanism. Optionally DNI are often measured by Rotating Shadow band Radiometer (RSR) or by integrated pyranometer such Sunshine Pyranometer (e.g. SPN1). Diffuse Horizontal Irradiance (DIF) is measured by (i) pyranometers, which obscure the direct radiation with a sun tracking disk or an adjustable fixed shadow ring, or (ii) by RSR equipped with rotating shadow band. Diffuse radiation can be also calculated as a difference between global and direct components. However, this method isn’t ideal because it increases uncertainty compared to a zealous measurement and doesn’t allow full quality testing of measurements should be considered, appropriate correction techniques must be applied to obtain accurate results. 

Landsat Data

Reliable solar models supported the employment of satellite and atmospheric data exist today. an honest description of the current approaches may be consulted in. In brief, the 

state-of-the-art high-accuracy modelling approaches have the following features: 

  • Use of recent models supported sound theoretical grounds, which are regionally and temporally consistent and computationally stable. 
  • Use of the state-of-the-art input data: satellite, aerosols, water vapor, etc. These computer files are systematically quality controlled and validated. 
  • Models and computer file are integrated and regionally adapted to perform reliably at a large range of geographical conditions. 
𝑝𝑃 = 𝜋2𝐸𝑆𝑈𝑁𝜆 𝑐𝑜𝑠𝜃𝑠 

Where: 

𝑃
= Unitless planetary reflectance 
= spectral radiance at the sensor’s aperture 
= Earth-Sun distance I astronomical units from an Excel file or interpolated from values listed I table 11.4 

𝐸𝑆𝑈𝑁𝜆
=Mean solar exoatmospheric irradiances from Table 11.3 

𝜃       𝑠 = Solar zenith angle in degrees 

Due to their complexity and diverse fine tunings, it is no surprise that models supported the identical theoretical principles perform very differently. The accuracy of the model depends on the underlying algorithms and quality of the input data and their optimized interaction. Advanced models are usually adaptive to numerous non-standard situations (e.g. snow, desert areas or extreme aerosol situations). within the modern computing approaches, the input data come from global monitoring systems: 

Satellite data from geostationary satellites are used for monitoring of clouds. Currently at minimum five operational satellite missions are needed to hide the Earth’s surface between latitudes 60 degrees North and South. Spatial resolution of satellite data is approximately 3 to five km, reckoning on the placement. Each satellite produces data within the range of three to 12 spectral bands at frequency of 15 and half-hour. Combined use of several spectral bands is that the optimum approach for accurate detection of cloud properties. 

Aerosol and vapor are used for modelling atmospheric properties. Data from global meteorological models are used. Spatial resolution of the modeled data is approx. 22 km to 125 km and their frequency of update is 3 and 6 hours. In case of aerosols, alternative to the models are data processed from satellite missions (e.g. MODIS, MISR, MERIS). Spatial resolution of those satellite-computed products is higher, but their availability and geographical coverage is irregular. 

High-resolution digital terrain model is employed for dealing with terrain shading and elevation effects. At a world scale, digital terrain model with spatial resolution up to 90 m (at the equator) is routinely used (SRTM-3). 

Other databases like land cover, temperature and snow are used as a support for modelling. 

Validation statistics are good for measuring the accuracy of model estimates at individual sites. However, the solar industry prefers working with uncertainty, the character of which is probabilistic. We present an approach for an uncertainty estimate of long-term annual GHI and DNI that is based on multiplying values for normal deviation of bias with different levels of confidence. The use of top-quality measurements for a model evaluation is general practice.

Uncertainty of such measurements, after rigorous internal control, is low and affects the user’s total uncertainty only marginally. Therefore, model uncertainty is commonly considered an equivalent to the user’s uncertainty. In some cases, ground measurements of dubious quality are used, which ends up in lack of correspondence with the solar model. The resulting discrepancies are then unfairly attributed to the satellite-based models.

Such practice does not result in objective findings and hence use of inferiority ground measurements for the model validation is strongly discouraged. We demonstrate that the accuracy of well-designed satellite based solar model are often stable over various geographies. Uncertainty of annual DNI is about 2-times higher compared to GHI. These findings are per an independent evaluation 

Specific Milestone Objectives

The project comprises 3 milestones, where: 

Milestone 1 is divided into 4 milestones: 

Milestone 1A – Identify potential data sources, evaluate and determine best dataset 

Milestone 1B – Setup storage platform and develop access and disaster recovery protocols 

Milestone 1C – Download global solar irradiation data at minimum 1 hour intervals 

Milestone 1D – Perform data cleansing activities on dataset 

Milestone 2 is divided into 4 milestones: 

Milestone 2A – Develop solar azimuth and elevation model 

Milestone 2B – Integration of relief into solar path model 

Milestone 2C – Visualisations and graphical representations based on ANE prototype 

Milestone 2D – Complete Monte Carlo simulation selectable by 1km/1km location 

Milestone 3 is divided into 4 milestones: 

Milestone 3A – Develop Monte Carlo simulation based on minimum of 20 years of daily data for analysing global dataset. 

Milestone 3B – Calculate standard error from theoretical sine wave-based maximum, based on specific location 10 minute dataset. Reduce standard error to log normal distribution 

Milestone 3C – Programme bespoke statistical distribution based on historic data  Milestone 3D – Finalise tool and integration into Google Earth Engine with reporting suite