Development of Machine Learning based Technique for Solar Energy Estimation

Abstract

The Republic of India, which has the second-largest population in the world, has challenges in meeting its rising energy demands by reducing fossil fuels. Also, the energy produced by fossil fuels contributes to global warming as the energy sector is explicitly responsible for generating harmful substances during the production, distribution, and consumption of energy. Without posing any environmental risks, the Sun, a limitless energy source, can be used as an alternative to meet this rising need. Further, the advancement of technology in chemistry, material science, and solid-state physics improves the efficiency of photovoltaic (PV) modules, has resulted in various topologies with different performance characteristics, and added sun-fuelled plants to a portfolio of the electricity market. Despite their many advantages and relative popularity as a renewable energy source, Even the greatest solar panels eventually lose their effectiveness. Inspections are necessary to maintain cell performance levels and minimize financial losses since solar cells are susceptible to damage from weather- related incidents, temperature changes, and UV exposure over time. How can real-time panel inspection be done in a way that is both economical and quick? This research work exploits the possibility of real-time estimation of solar power in PV systems. A new method was brought to light: Utilizing historical weather data, the Clustering-based Computation Rate (CCDR) calculates performance ratios and degradation rates. This method also allows for off-site inspection. Most meth- ods on the market base do their calculations via on-site physical assessment of PV installations. These methods are not preferred for real-time degradation investigation because it is time-consuming, expensive, and labor-intensive. The suggested model provides a real-time estimation of the performance ratio. The degradation effect in terms of performance ratio is incorporated in estimation results. The degradation is calculated in real-time using the clustering-based technique without a physical on-site inspection, therefore, can be used for the real-time estimation of solar power. Two important research goals are to maximize the power output from PV sys- tems and further reduce economic losses. This study proposes a novel technique that uses a clustering-based technique to evaluate the degradation of PV panels with dif- vi ferent topologies. The proposed method calculates the degradation in output solar power in terms of PR for panels with three different topologies, namely, amorphous silicon (a-Si), polycrystalline silicon (p-Si), and heterojunction with an intrinsic thin layer (HIT), over a period of three years. The degradation rate produced by a-Si technology was the lowest, and it was highest for HIT technology. The degradation is calculated in real-time using the clustering-based technique is used further to fine- tune the estimation of solar power. Initially, the power is estimated using the Support Vector Regression (SVR) model with the meteorological parameters. The estimation is further fine-tuned in sync with the degradation rate. The model is validated on the actual data (Meteorological parameters and Solar power) procured from the solar plant. After refinement, the estimation results show significant improvement in terms of statistical measures.