Analysis and Modeling of Photovoltaic Arrays for Sustaining Power Generation in Geostationary Satellite Solar Panels using Machine Learning

Geostationary satellite solar panels are vital energy sources for space-borne systems. Understanding their power generation and accurately modeling performance is crucial for satellite design, manufacturing, and operation optimization. This study explores how solar panel power fluctuates in response to varying conditions on geostationary satellites. We present a method employing neural networks to model this power variability over time effectively. To achieve this, we employ non-linear autoregressive neural networks with exogenous inputs, utilizing both single-input and six-input configurations with feedback. Our comprehensive analysis yields a Mean Squared Error (MSE) of 0.0477 and a regression value of 0.9999, indicating exceptional performance. These results validate a strong correlation between predicted and actual power values, underscoring the accuracy of our neural networkbased approach in capturing the dynamics of solar panel power generation on geostationary satellites. Satellite operators can employ this technique for effective monitoring and forecasting of solar panel-generated power.