Model development and validation of a large-scale photovoltaic plant with a dual-axis tracking system: case of the Free State, South Africa

Abstract

The production of outdoor photovoltaic (PV) modules is subject to a broad range of parameters that directly influence their energy output. These parameters can generally be categorised into two groups: module-specific features and external environmental conditions. The former pertains to the physical characteristics of the solar modules themselves, including their efficiency, material composition, and the specific technology employed. In contrast, external parameters encompass environmental factors such as wind speed, solar irradiance, ambient temperatures, and geographical location, which may vary significantly between different locations. For instance, elevated temperatures can diminish the efficiency of solar cells, while low wind speeds may result in inadequate cooling of the modules. Additionally, factors such as solar angles, cloud cover, and dirt accumulation (soiling) further complicate the task of accurately predicting the output of PV systems. The majority of contemporary research on PV systems has focused on static systems, wherein solar modules are fixed at a predetermined tilt and orientation, and models have been developed to forecast energy generation based on a static set of parameters. However, these models frequently fail to accurately represent the realities of dual-axis tracking systems, which possess higher energy output potential by adjusting the angle of panels to follow the sun’s horizontal and vertical movement. Although some studies have investigated dual-axis tracking systems, there is a relative paucity of research conducted within the South African context, particularly regarding the unique challenges posed by the country’s climate, which includes semi-arid conditions and significant diurnal temperature variations. The primary objective of this study was to address the research gap by developing and validating an extremely accurate and dependable predictive model for the energy output and performance of dual-axis PV tracking systems in South Africa, with specific application to the CUT installation in Bloemfontein. This research utilised real-world data gathered from a dual-axis PV tracking power plant at the Central University of Technology, Free State (CUT) in the Free State province. An average tracking system was employed, and the data were interpolated to provide a global aggregate for the entire plant. By incorporating data collected from meteorological stations and the Sunny Portal monitoring system platform, this research offers a comprehensive approach to modelling PV system performance under South African environmental conditions. A notable aspect of this study was the utilisation of machine learning (ML) techniques to enhance the accuracy and reliability of the predictive model. Various popular ML methods, including linear regression, decision trees, support vector machines, and artificial neural networks, have been explored using MATLAB’s Regression Learner App. These methods were evaluated based on their capacity to predict the power output of the PV component from the tracking system, considering environmental parameters such as global horizontal irradiance, direct normal irradiance, diffuse horizontal irradiance, wind speed, temperature, and relative humidity. Among the models assessed, the ensemble tree model was found to be the most effective, as it achieved a coefficient of determination (R²) of 0.99338 and a root mean square error of 0.39487. This model successfully characterised the complex, non-linear interactions between environmental parameters, thus demonstrating its potential for accurately forecasting the output of the PV component under various operational conditions and time periods. The newly developed predictive model not only provides exceptional accuracy but also generalises well across a range of conditions, which suggests its applicability for year-round operation. A sensitivity analysis performed on the model highlighted the significant impact of wind speed in reducing the temperature of the PV cells, which enhanced conversion efficiency. This factor has often been overlooked in other research but was identified through this modelling process as a crucial element that contributes to real-world output variability in South Africa’s semi-arid climate. The model also effectively accounted for the dynamic behaviour of solar angles, which change throughout the day and year, thus influencing the system’s total energy output. These findings further underscore the necessity of incorporating local environmental conditions into PV performance models for more reliable predictions. In addition to performance prediction, the study addressed the degradation of PV systems over time, which is an essential consideration for long-term system planning and maintenance. An analysis of data spanning from 2019 to 2023 revealed a gradual decline in the annual energy output of the CUT’s PV plant, from 41 689.72 kWh in 2019 to 40 038.13 kWh in 2023, which resulted in a performance loss of approximately 3.96%. This degradation was attributed to various factors, including environmental soiling, component ageing, and maintenance delays. The degradation analysis emphasises the importance of ongoing monitoring and maintenance to mitigate the impact of these factors on long-term system efficiency. The predictive model’s ability to reflect real-world changes in system performance adds significant practical value for both energy planners and maintenance teams, as it enables them to anticipate performance issues and to proactively implement corrective measures. Moreover, this research conducted a comprehensive cost analysis of the CUT’s PV system to demonstrate whether it constitutes a sustainable and financially viable investment. The system accrued substantial savings during its first five years of operation, amounting to approximately R714 000. The investment has proven to be highly rewarding, with a return on investment of 2.22. This financial analysis, coupled with the environmental benefits of reduced carbon dioxide emissions and diminished reliance on fossil fuels, further strengthens the case for increased investment in solar energy systems at the CUT and other South African universities. This research makes a significant contribution to the expanding body of literature on South African PV system performance modelling, specifically for dual-axis tracking systems. The validated performance prediction model represents a meaningful advancement towards facilitating decision making in the planning, installation, and operational management of solar PV systems in South Africa. Given South Africa’s increasing focus on the adoption of renewable energy sources, particularly through initiatives such as the Renewable Energy Independent Power Producer Procurement Programme, the results and methodologies presented in this study can inform future site selection, investment appraisal, and energy yield forecasting for PV installations. In addition to its intellectual contributions, this research provides valuable methodologies for optimising solar power in South Africa. Future investigations may extend the model to other provinces with varying climatic conditions and incorporate economic metrics for a more comprehensive long-term perspective on the benefits of PV systems. Additionally, the development of real-time system performance dashboards for optimisation could further enhance the model’s forecasting capabilities. The research also advocates for the implementation of soiling and shading models to improve accuracy in areas with high levels of dust accumulation, as these factors can significantly reduce the efficiency of PV systems. Overall, this study presents a validated and robust tool for assessing and optimising the performance of dual-axis PV trackers in South Africa, thereby contributing to both the body of knowledge through research and to practical energy planning. The findings of this research hold significance for the advancement of renewable energy technologies and the transition towards a cleaner, more reliable energy grid in South Africa and beyond. By integrating advanced modelling techniques, real data, and sustainability considerations, this study establishes a solid foundation for the future optimisation and forecasting of solar energy, thereby promoting a more sustainable energy future.