1. Recording age and variability
A long-term (20-50 year) solar resource dataset is relatively desirable because it can provide a good indication of long-term changes in solar resource and can ensure that the data studied adequately include captured volcanic eruptions and climatology dynamic influence.
Although a volcanic eruption is an exceptional event, it can have a significant impact on system performance, so the solar resource at the time of an eruption is defined as a “worst-case scenario.” In assessing solar resources, two major volcanic eruptions in the past 30 years are particularly important: the 1991 eruption of Mount Pinatubo in the Philippines and the 1982 eruption of El Chichon in Mexico. During the Pinatubo eruption, GHI decreased by 10% and DNI decreased by 15% to 20% in some areas. The eruption also severely curtailed production at the CSP facility of the California Solar Power System in 1992.
Some analysts do not list volcanic eruptions as a factor to consider in assessing long-term solar energy resources, arguing that volcanic eruptions do not represent naturally occurring changes in solar energy resources. While this is an obvious fact in meteorological concepts, no one can predict when the next volcanic eruption will occur and how much it will affect the performance of the solar system. For lenders, all possible events should be considered comprehensively when conducting project analysis, as this will have an impact on their financial position.
Beginning in 1998, resource assessments in North America have only used satellite-based datasets due to their high availability and high resolution. But there are problems with these reasons: (1) 2000-2010 were “good” years for solar energy resources for most sunny regions; and (2) these datasets did not include any volcanic eruptions.
2. Comparing and calibrating solar datasets across time and space
In addition to a long-term data set, data information obtained from multiple sources that overlap in time records is also beneficial for resource risk assessment. When coordinated, this type of information also provides an opportunity to identify biases that exist in one or even multiple datasets, while also enhancing the reliability of the data and avoiding major biases.
Under clear-sky conditions, anthropogenic radiation can be estimated fairly accurately through a range of models. Under variable or cloudy conditions, modeling solar resources is fraught with uncertainty. In cloudy months, the proportion of this uncertainty is much larger than in sunny months. Since there are more solar resources in sunny days than other time periods, the annual uncertainty is largely determined by the deviation of the sunny days. Therefore, when using the two solar datasets for comparative calibration, it is important to correct the data bias for sunny periods.
To effectively identify site-specific data deviations and quality issues, data information from nearby and regional measurement sites can be compared. Often, the sites with the longest records do not have solar installed. Therefore, in order to confirm the solar resource of a site, the long-term variability of the resource is included in the assessment, it is necessary to compare with the neighboring site with the longest record, and to develop a method to incorporate the variability of the long-term solar resource dataset into the In a dataset of actual solar installations, this may use simple ratios or complex statistical methods.
3. Data Uncertainty
From the perspective of solar resource risk assessment, the uncertainty of resource data is difficult to quantify. One of the most important aspects of project lending is the assessment of potential negative bias errors in the data. However, equity investors will also be interested in positive bias, as this type of information helps to understand the potential benefits of a project.
Assessing uncertainty in long-term data sets compiled for specific engineering sites is a meticulous process that requires detailed analysis and engineering evaluation. A resource evaluation for a particular site is likely to involve data from multiple sources, each with its own uncertainties. Furthermore, the uncertainties associated with each data source (ground measurements, ground modelling or satellite retrievals) can vary significantly over different time periods (time scales) or under specific meteorological conditions. Therefore, it is not possible to propose a standard method here.