Data driven prediction based reliability assessment of solar energy systems incorporating uncertainties for generation planning.

In the era of renewable energy integration, precise solar energy modeling in power systems is crucial for optimized generation planning and facilitating sustainable energy transitions. The present research proposes a comprehensive framework for assessing the operational reliability of solar integrated systems, validated using the IEEE RTS 96 test system. A robust uncertainty model has been developed to characterize variations in solar irradiance to address the uncertainties in solar panel output, followed by a multi-state modeling approach to account for the dynamic nature of solar panel output. The research introduces a time series-based 'non-linear autoregressive neural network' (NAR-Net) to forecast the solar irradiance levels five days ahead to optimize solar power efficiency. A comparative analysis has been conducted of three other state-of-the-art approaches, such as auto-regressive (AR), auto-regressive with moving average, and multi-layer perceptron, for predicting solar irradiance. Performance metrics, including mean square error, regression, and computational time, were evaluated to demonstrate the efficacy of the NAR-Net. The proposed prediction-based approach enhances the reliability of power generation planning by integrating modeling, which is based on forecasting. It is found that the proposed method achieves an accuracy of 98% w.r.t its counterpart. Moreover, the assessment to optimize the operational reliability of solar-integrated systems and improve generation planning for a sustainable energy future is achieved.