Abstract
With increasing concern related to carbon dioxide emission, the concept of Zero Energy Building (ZEB) appears. Electric Vehicles are also considered environment friendly automobiles to reduce greenhouse gas emission. With global trends above, Building energy system should consider ZEB concept and electricity demand for EVs charging. Therefore, this thesis suggests a reasonable problem-solving method to find the best optimal energy system design to meet ZEB condition. EVs charging demand is predicted from the fittest Machine Learning regression model. This result is combined with Hourly Building Energy demand from the EnergyPlus simulator. Last, Genetic Algorithm and PROBID method will be applied to meet two objective functions: Annual Total Cost and Self-Energy Sufficiency ratio.
Introduction
Buildings are currently responsible for approximately 33% of global greenhouse gas emissions, making the transition to Zero Energy Buildings (ZEB) a critical climate goal. However, designing these systems is inherently complex due to high capital costs and fluctuating energy demands. A significant emerging challenge in this field is the integration of electric vehicle (EV) charging infrastructure. While traditional building energy models focus on electricity, heating, and cooling as primary loads, they often overlook the substantial and variable load added by EVs. This research addresses this gap by creating a holistic design tool that predicts EV demand and optimizes the capacity of various energy technologies to meet ZEB standards economically.
Methodology
The study employs a comprehensive four-step methodology to integrate EV infrastructure into building energy systems:
- EV Demand Prediction: Five machine learning regression models—Multi-Linear Regression (MLR), Artificial Neural Networks (ANN), Bagging ANN, Gradient Boosting Machine (GBM), and Random Forest (RF)—were tested to predict electricity demand for EV charging. Data was sourced from the Korean Government, accounting for variables such as day of the week, vehicle counts, and ambient temperature.
- Building Simulation: Using EnergyPlus, the researchers simulated hourly heating, cooling, and electricity demands for 16 prototype building models. Benchmark models were utilized to convert monthly demand data into granular hourly scales for higher precision.
- System Integration: The predicted EV charging demand is combined with the simulated building energy demand to create a unified load profile.
- Multi-Objective Optimization: The NSGA-II (Non-dominated Sorting Genetic Algorithm II) was used to identify the "Pareto frontier," balancing the competing goals of minimizing Annual Total Cost (ATC) and maximizing the Self-Energy Sufficiency Ratio (SSR). The energy system considered includes Photovoltaics (PV), Energy Storage Systems (ESS), Fuel Cells (FC), Electric Heat Pumps (EHP), Gas Heat Pumps (GHP), and Absorption Chillers.
Finally, the PROBID method was applied to select the single most robust design from the Pareto set for practical implementation.
Key Insights and Results
Model Performance
Among the machine learning models tested, the Random Forest model achieved the highest accuracy, yielding the lowest Mean Absolute Error (MAE) for both office (6.90) and residential (4.20) scenarios. This highlights the effectiveness of ensemble learning in capturing the human variability inherent in EV charging behavior.
Optimization and Economic Impact
The framework successfully identified system configurations that met "nZEB Level 5" criteria, requiring at least 20% annual energy self-sufficiency. In a case study of a large hotel on Jeju Island, South Korea, the optimized Annual Total Cost was approximately $918,624 for the office area and $956,596 for the residence area. The research also noted that government subsidies could further reduce these annual costs by roughly $12,000 to $14,000, significantly improving financial viability.
Technology Selection
The study demonstrated a clear trade-off between SSR and cost; as self-sufficiency increases, costs rise sharply. Interestingly, local policy plays a major role in technology selection. In Korea, Gas Heat Pumps (GHP) are often incentivized to mitigate the massive electricity spikes caused by summer air conditioning, a factor the optimization model successfully accounted for.
Conclusion
This research demonstrates that EV charging is no longer a negligible load and must be integrated into initial building energy balance equations to achieve true ZEB status. While the current model provides a robust framework for optimization, future work should explore dynamic "Smart Charging" or Vehicle-to-Grid (V2G) technologies and account for the long-term performance degradation of components like PV panels and batteries over their full lifecycle.
Summary updated