Deep learning based solar forecasting for optimal PV BESS sizing in ultra fast charging stations

Area of Science:

• Electrical Engineering
• Renewable Energy Systems
• Optimization Techniques

Background:

• Ultra-fast charging stations (UFCS) face challenges with high power demands and grid dependency.
• Integrating renewable energy sources is crucial for sustainable electric vehicle infrastructure.

Purpose of the Study:

• To develop an optimization framework for sizing photovoltaic (PV) and battery energy storage systems (BESS) for UFCS.
• To enhance the techno-economic feasibility and grid independence of UFCS through renewable integration.

Main Methods:

• Utilized a Gated Recurrent Unit (GRU) deep learning model for accurate solar PV output forecasting.
• Employed a Genetic Algorithm (GA) to optimize PV and BESS sizes, maximizing Net Present Value (NPV).
• Analyzed weekday and weekend demand profiles for tailored system sizing.

Main Results:

• PV integration alone improved NPV by €6.19 million.
• Combined PV and BESS integration increased NPV to €33.97 million, reaching €34.05 million with projected cost reductions.
• Reliability metrics (ESR, AR) confirmed enhanced self-sufficiency and reduced grid dependency.

Conclusions:

• The proposed framework demonstrates the significant techno-economic potential of hybrid renewable-powered UFCS.
• Intelligent forecasting and evolutionary optimization are key to maximizing the benefits of PV and BESS integration.
• This approach offers a viable pathway to sustainable and grid-independent ultra-fast charging infrastructure.