Enhanced deep learning models for photovoltaic generation forecasting 

Supervisors 

Dr. Richardt Wilkinson (RMIT) 

The project 

The global transition to sustainable energy is imperative to meet decarbonisation targets by 2050. Solar photovoltaic (PV) systems, a major contributor to this transition, face challenges due to their dependency on weather conditions, leading to significant uncertainty in power output. Accurate forecasting of PV generation is crucial to mitigate this uncertainty and ensure reliable integration into the energy grid. This research identifies key gaps in current forecasting models, including limitations in handling diverse environmental conditions and integrating historical data effectively. Existing methods, such as statistical models and traditional machine learning approaches, fall short of addressing these complexities.  

This study aims to develop a hybrid deep learning model that combines the strengths of individual models to improve forecasting accuracy for PV generation. By incorporating the research learnings to weather factors and historical PV data, the proposed model seeks to enhance prediction reliability under varying environmental conditions. Preliminary results indicate potential improvements over conventional methods, paving the way for more robust and adaptive forecasting techniques. This research contributes to the advancement of renewable energy integration, supporting a stable and sustainable energy future.