Advancements in Solar Radiation Data Analysis for Improved Solar Energy Applications

Solar energy, has witnessed significant advancements in recent years. These advancements are not limited to solar panel efficiency improvements or energy storage technologies; they also extend to the way researchers analyze and utilize solar radiation data.  Indeed, solar radiation data plays a pivotal role in assessing the energy yield capability of solar energy applications. It’s crucial to recognize that global solar radiation (I_global) consists of two components: beam radiation (I_beam) and diffuse radiation (I_diffuse). While measuring global radiation is less expensive since it only requires a single pyranometer, determining the performance of concentrating solar energy systems necessitates knowledge of the hourly solar diffuse fraction (d), defined as I_diffuse/I_global. To estimate the solar diffuse fraction, various empirical models have been developed using global radiation and other meteorological factors. One of the early models, known as the Liu-Jordan type, estimates d using the hourly sky clearness index (k_t), where k_t = I_global / H₀, with H₀ being the hourly extraterrestrial radiation determined by site latitude, day of the year, and time of day. These models provide a simple approach but may lack precision. In contrast, multiple-predictor models like the Boland-Ridley-Lauret (BRL) model, which incorporates predictors such as k_t, solar altitude (α), persistence of global radiation (Ψ), apparent solar time (AST), and clearness indices (k_t and K_T), offer improved accuracy but are less suitable for real-time predictions due to their reliance on multiple predictors.

Selecting the right dataset is crucial when developing and testing correlation models to avoid weather bias. Existing approaches include random sampling, using out-of-sample data for testing, and covering various seasons. However, constructing a dataset that represents long-term typical weather conditions for a year is considered the most suitable approach. This is where the concept of a Typical Meteorological Year (TMY) becomes invaluable. The TMY method involves selecting typical meteorological months (TMMs) from different years and combining them to create a TMY dataset. The selection process relies on statistical comparisons between long-term and short-term cumulative distribution functions for various weather parameters.

In a new study published in the peer-reviewed Journal Renewable Energy by Chun-Tin Lin, Professor Keh-Chin Chang, and Kung-Ming Chung from the National Cheng Kung University, they developed innovative approaches to harnessing solar radiation data. The research team used ten years (2011–2020) of solar radiation data measured at the Kuei-Jen campus of the National Cheng Kung University in Taiwan. They employed the TMY3 method to generate a TMY dataset and construct the training dataset using all twelve TMMs.

Two multiple-predictor correlation models were developed: a modified BRL model and a piece-wise linear model, alongside a single-predictor (Liu-Jordan-type) model. These models aim to predict the hourly (d) using a combination of predictors, including k_t, α, AST, daily clearness index (K_T), and Ψ.

To evaluate the models’ performance, the researchers employed several statistical indicators, including the root-mean-square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), standard deviation (SD), and coefficient of determination (R2). These metrics help assess the models’ accuracy, precision, and reliability. Additionally, a global performance indicator (GPI) was introduced, which combines the results from various statistical indicators to provide an overall assessment of each model’s performance.

The authors conducted a comprehensive comparative analysis between the models developed using the TMY dataset and those based on one or two years of data from the same location. The results consistently demonstrated that models developed using the TMY dataset outperform those based on limited datasets. This demonstrates the importance of using representative, long-term data for model development. Furthermore, when the authors compared the developed models with existing models from different regions, it becomes evident that no single model is universally applicable to all geographical regions and climates. Each model performs best in its respective region, highlighting the need for region-specific modeling.

In summary, the study by Professor Keh-Chin Chang and colleagues highlighted the significance of representative datasets and advanced modeling techniques for solar energy. The development of correlation models for solar radiation data using a TMY dataset has proven to yield superior results, emphasizing the need for long-term, region-specific data in model development. The authors’ findings also underscore the diversity of solar radiation patterns across different geographical regions and climates, highlighting the importance of tailoring modeling approaches to specific locales. As solar energy continues to play a pivotal role in the transition to sustainable energy sources, these advancements in data analysis techniques contribute significantly to optimizing energy yield and efficiency in solar applications.