University of Mohaghegh Ardabili Journal of Operation and Automation in Power Engineering 2322-4576 2025 11 27 Next-Gen Solar Forecasting: PSO-Optimized Bayesian LSTM for Enhanced Accuracy 4267 10.22098/joape.2025.16360.2266 EN S. J. Sadheesh Kumar Department of Electrical and Electronics Engineering, National Institute of Technology Puducherry, Karaikal, India. Navin Sam K Department of Electrical and Electronics Engineering, National Institute of Technology Puducherry, Karaikal, India. Journal Article 2024 12 13 Accurate solar photovoltaic power (SPVP) generation forecasting is vital for integrating solar energy into the power grid. This paper presents an advanced forecasting model using a Bayesian enhanced Long Short-Term Memory neural network (BLSTM NN) model optimized by the Particle Swarm Optimization (PSO) algorithm to elevate the accuracy and reliability of SPVP generation forecasting. The hyperparameters of the BLSTM NN are optimized using the PSO algorithm, resulting in improved forecasting performance. The model is evaluated using a comprehensive dataset comprising five years of historical data from a 1 MW SPVP plant in South India, sampled at 15-minute intervals. Key performance indicators, including Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and R-squared, are used for various analyses. The performance of the proposed model is evaluated through accuracy, uncertainty, scalability, and sensitivity analyses. Experimental results highlight a 16.02% reduction in RMSE, a 22.84% reduction in MAPE, and a 24\% improvement in R² over the conventional baseline DB model. The outcomes underscore the capability of the proposed model to deliver superior forecasting accuracy. The approach helps integrate reliable and efficient solar power into grid planning.         https://joape.uma.ac.ir/article_4267_1854790e2603fe669f6cffe03cfa6c69.pdf