Photovoltaic (PV) power generation exhibits considerable sensitivity to both weather variability and fluctuations in solar irradiance. Consequently, precise forecasting of PV power is crucial for ensuring grid reliability, load balancing, and the effective functioning of energy markets within a grid-connected solar plant. Conventional forecasting methodologies frequently prove inadequate in accurately capturing the nonlinear and intricate temporal patterns present within PV datasets. To address these shortcomings, this research presents a hybrid short-term PV power forecasting model. This model integrates Neighborhood Component Analysis (NCA) for dimensionality reduction with a Convolutional Neural Network (CNN)-Long Short-Term Memory (LSTM) framework. NCA helps reduce computational complexity while still preserving the important features of high-dimensional PV data. CNN layers are designed to extract spatial features that are localized. In contrast, the LSTM component is adept at capturing temporal dependencies. This combination allows the model to effectively process short-term dynamics. The hybrid model under consideration was evaluated using empirical data obtained from a solar power plant, and its performance was compared to that of conventional machine learning models and individual deep learning (DL) models. The evaluation of performance revealed a MAE of 0.1945, a MAPE of 7.1118, a RMSE of 0.3645, and an R² value of 0.976, thereby confirming its enhanced predictive abilities. These results indicate that the hybrid model improves accuracy, precision, and stability in the forecasting of PV power. Consequently, this enhancement underscores the potential of a hybrid DL model to optimize real-time energy management within renewable power systems.

The reliability and efficiency of photovoltaic (PV) systems are essential for sustainable energy production, requiring accurate fault detection to minimize energy losses. This study proposes a hybrid model integrating Neighborhood Components Analysis (NCA) with a Convolutional Neural Network (CNN) to improve fault detection and diagnosis. Unlike Principal Component Analysis (PCA), which may compromise class relationships during feature extraction, NCA preserves these relationships, enhancing classification performance. The hybrid model combines NCA with CNN, a fundamental deep learning architecture, to enhance fault detection and diagnosis capabilities. The performance of the proposed NCA-CNN model was evaluated against other models. The experimental evaluation demonstrates that the NCA-CNN model outperforms existing methods, achieving 100% fault detection accuracy and 99% fault diagnosis accuracy. These findings underscore the model’s potential in improving PV system reliability and efficiency.