Ultra-Short-Term Solar PV Power Forecasting for Smart Grids Using a Hybrid CNN-BiLSTM-Attention Model with Real-Time Weather Data Integration
Ultra-Short-Term Solar PV Power Forecasting for Smart Grids Using a Hybrid CNN-BiLSTM-Attention Model with Real-Time Weather Data Integration
Authors - Sudhakar kumar and Mrs.Ritu Singh
Sudhakar kumar- M.Tech student, Department Of Electrical Engineering, K.K University, Bihar Sharif, Nalanda, Bihar-803115
Mrs. Ritu Singh- Assistant Professor, Department Of Electrical Engineering, K.K University, Bihar Sharif, Nalanda, Bihar-803115
Abstract
Accurate forecasting of solar photovoltaic power is essential for smart grid operation because solar output changes quickly with weather conditions such as irradiance, cloud cover, and temperature. This paper presents a hybrid forecasting framework that combines convolutional neural networks, bidirectional long short-term memory, and attention to improve ultra-short-term PV power prediction. The model is designed to use real-time weather information so that it can respond better to rapid changes in generation. Compared with traditional statistical and machine learning approaches, the proposed architecture is more suitable for nonlinear and time-dependent solar data. The study supports more reliable grid balancing, reserve scheduling, and energy management in smart grid systems.