Predictive optimization using long short-term memory for solar PV and EV integration in relatively cold climate...

Area of Science:

• Sustainable Energy Systems
• Renewable Energy Integration
• Electric Mobility

Background:

• Growing adoption of sustainable energy and electric mobility faces challenges in integrating renewables, especially in cold climates with variable resource availability.
• Ensuring consistent energy performance in dynamic conditions requires innovative solutions for managing supply and demand.
• Cold regions present unique complexities for energy reliability due to unpredictable renewable resource availability.

Purpose of the Study:

• To explore energy dynamics in a cold-climate residential community, considering building energy needs and electric vehicle (EV) charging demands.
• To evaluate the potential of solar energy generation from photovoltaic (PV) systems.
• To optimize energy supply and demand management using a machine learning approach.

Main Methods:

• Estimation of individual building energy requirements and additional EV charging demand.
• Assessment of solar energy generation potential using PV systems.
• Application of Long Short-Term Memory (LSTM) machine learning models for energy demand forecasting and management.

Main Results:

• Heating demands in cold climates significantly exceed cooling needs.
• Solar energy covers approximately 32.1% of energy needs during warmer months, with grid support necessary in colder seasons.
• LSTM models achieved over 93% accuracy in predicting EV charging patterns, improving energy demand forecasting and load management.

Conclusions:

• Optimizing renewable energy use is crucial for reducing grid dependency in cold climates.
• Effective production-demand management strategies, aided by machine learning, can enhance overall energy efficiency.
• The study demonstrates a viable approach for managing renewable energy integration in challenging climatic conditions.