Abstract With the rapid growth in the number of electric vehicles (EVs), the construction and load optimization management of charging infrastructure in airport transportation hubs, as critical public service facilities, have become increasingly important. This study focuses on the load characteristics of airport charging facilities, establishing a multi-objective optimization model that considers charging demand, power grid carrying capacity, and user service quality. By introducing spatio-temporal distribution feature analysis and load forecasting methods, operational data from an international airport’s charging infrastructure is modeled and analyzed. A load optimization strategy based on time-of-use (TOU) pricing and intelligent scheduling is proposed. The results show that the optimized scheme reduces the peak-valley difference in charging load by 37.9%, lowers power supply costs by 13.6%, and achieves a charging service availability rate of 99.95%.
Keywords: electric vehicle; airport; charging infrastructure; load optimization; intelligent scheduling
1. Introduction
The thriving development of the electric vehicle industry has propelled the rapid expansion of charging infrastructure. Airports, as major transportation hubs, exhibit distinct spatio-temporal aggregation and randomness in EV charging demand within their parking areas. Unordered use of large-scale charging facilities may lead to overloading of the distribution network and deterioration of power supply quality. Therefore, in-depth research on the load characteristics of airport charging facilities and the establishment of optimization strategies adaptable to dynamic changes in charging demand are of significant importance for improving the operational efficiency of charging infrastructure and power supply reliability.
Existing studies primarily focus on load optimization for urban public charging stations, with relatively insufficient systematic research on special scenarios such as airports. This study aims to fill this gap by analyzing the unique load patterns in airport environments and developing tailored optimization solutions.
2. Load Characteristics of Airport Charging Infrastructure
2.1 Spatio-Temporal Volatility and Aggregation
Airport charging facility loads demonstrate pronounced spatio-temporal fluctuations. Peak periods concentrate between 7:00–9:00 and 18:00–20:00, closely correlated with flight takeoff/landing peaks and business travel patterns. Short-term parking areas prioritize fast charging, with an average single charging duration of 45 minutes, while long-term parking areas favor slow charging, with durations ranging from 4 to 6 hours.
Spatially, charging facilities near terminal buildings exhibit the highest utilization rate, with a load density of up to 4.5 kW/m², while remote parking areas have more dispersed loads, with densities fluctuating between 1.2–2.0 kW/m². Seasonal variations are also significant: summer sees a 20–30% increase in charging peaks due to air conditioning loads, while winter extends charging times by 15–25% due to reduced charging efficiency.
2.2 Key Metrics Summary
| Parameter | Short-Term Parking | Long-Term Parking | Terminal Area Load Density | Remote Area Load Density | Summer Peak Increase | Winter Charging Time Extension |
|---|---|---|---|---|---|---|
| Average Charging Duration | 45 min | 4–6 h | 4.5 kW/m² | 1.2–2.0 kW/m² | 20–30% | 15–25% |
| Table 1. Summary of Charging Load Characteristics |
3. Composition of the Load Optimization System
3.1 Infrastructure and Functional Layers
The load optimization system is structured into four layers: perception, network, platform, and application:
- Perception Layer: Comprises hardware devices such as charging piles, power metering units, and environmental sensors, enabling real-time monitoring of key parameters during charging.
- Network Layer: Utilizes 5G and industrial Ethernet to establish a highly reliable data transmission network for real-time communication of device status and control commands.
- Platform Layer: Serves as the system core, integrating a data processing center, intelligent scheduling module, and security management module for data analysis, strategy optimization, and system coordination.
- Application Layer: Provides visual monitoring, intelligent booking, and statistical analysis for operators and users.
3.2 Data Collection and Processing
Data collection systems use distributed sensors to continuously sample electrical parameters (charging power, voltage, current, duration) at a 100 ms interval and environmental parameters (temperature, humidity, light) at a 5-minute interval. Preprocessing includes digital filtering and outlier removal, followed by feature extraction and data fusion at minute, hourly, and daily scales.
Wavelet transform is applied for multi-scale decomposition of load curves to identify variation features and periodic patterns. Principal component analysis (PCA) reduces data dimensionality, extracting key factors influencing load changes to establish a load feature index system.
3.3 Optimization Model Formulation
The multi-objective optimization model aims to minimize distribution network operation costs and maximize charging service quality. The objective function F is expressed as:\(F = \min \left( C_{\text{loss}} + C_{\text{dep}} + C_{\text{wait}} \right)\) where:
- \(C_{\text{loss}}\): Power loss cost
- \(C_{\text{dep}}\): Equipment depreciation cost
- \(C_{\text{wait}}\): User waiting cost
Constraints include:
- Charging power constraint: \(P_{\text{min}} \leq P(t) \leq P_{\text{max}}\)
- Transformer capacity constraint: \(\sum P(t) \leq S_{\text{rated}}\)
- Voltage deviation constraint: \(|V(t) – V_{\text{nom}}| \leq \Delta V_{\text{max}}\)
- Charging completion time constraint: \(T_{\text{start}} + t_{\text{charge}} \leq T_{\text{deadline}}\)
The improved Non-dominated Sorting Genetic Algorithm (NSGA-II) is employed for model solving, with adaptive crossover and mutation operators to enhance convergence. Fuzzy decision-making filters the Pareto optimal solution set to obtain the optimal scheduling plan, incorporating robustness constraints to address load forecasting errors.
4. Design of Load Optimization Strategies
4.1 Time-of-Use (TOU) Pricing Parameters
TOU pricing divides the day into peak, flat, and valley periods based on load characteristics and power supply costs, with the goal of maximizing peak shaving and valley filling. Machine learning analyzes historical data to determine optimal price ratios. The peak-valley price difference is set at 0.8–1.2 元 /kWh, achieving a 15–20% load transfer rate. Seasonal differentiation is applied:
| Time Interval | Season | Price (¥/kWh) | Description |
|---|---|---|---|
| 7:00–9:00, 18:00–20:00 | Summer | 1.50 | Peak period (+50% 基准价) |
| Winter | 1.35 | Peak period (+50% 基准价) | |
| 9:00–18:00, 20:00–23:00 | All | 1.00/0.90 | Flat period (基准价) |
| 23:00–5:00 | Summer | 0.70 | Valley period (-30% 基准价) |
| Winter | 0.63 | Valley period (-30% 基准价) | |
| 5:00–7:00 | All | 0.85/0.77 | Special period (mid-price) |
| Table 2. TOU Pricing Parameters for Airport Charging Infrastructure |
4.2 Intelligent Scheduling System Architecture
The hierarchical intelligent scheduling system includes a master control center and field controllers:
- Master Control Center: Employs deep reinforcement learning for dynamic power allocation and load forecasting, achieving 92% prediction accuracy with a response time <100 ms.
- Field Controllers: Implement differentiated strategies: fast charging priority for temporary parking areas and economy-oriented scheduling for long-term parking areas.
Key performance metrics:
- Peak load reduction rate: 25%
- System response time: <100 ms
- Function modules: Load forecasting, real-time monitoring, fault diagnosis
4.3 Safety Operation Assurance System
Composed of three subsystems:
- Monitoring and Early Warning: Real-time monitoring of charging equipment status with 12 alarm indicators (e.g., voltage over-limit, temperature exceedance).
- Emergency Handling: Includes load emergency shutdown and power supply switching within 20 ms.
- Safety Assessment: Quantifies operational risks using a state evaluation model across electrical, environmental, and operational safety dimensions, with a reliability index of 99.99% and annual failure rate <0.1%.
Key Equipment Redundancy Configuration
| Equipment Type | Redundancy Ratio | Monitoring Frequency | Fault Switching Time |
|---|---|---|---|
| Charging Piles | 15% | Continuous | <20 ms |
| Transformers | 20% | 10-minute intervals | <30 ms |
| Communication Devices | 25% | 5-minute intervals | <10 ms |
| Table 3. Redundancy Configuration for Critical Equipment |
5. Evaluation of Optimization Effects
5.1 Load Curve Smoothness Analysis
Data from September 2023 to February 2024 (6 months) shows:
- Load volatility reduced from 32.5% to 18.2%
- Peak-valley difference decreased from 580 kW to 360 kW
- Standard deviation reduced from 156 kW to 92 kW
- Peak load reductions: 25.3% (morning) and 23.8% (evening)
- Load shifted to 10:00–16:00 and 22:00–4:00 (18.6% of total charge volume)
- Smoothness index improved by 42.3%
5.2 Power Supply Reliability
| Metric | Pre-Optimization | Post-Optimization | Improvement |
|---|---|---|---|
| Voltage Qualification Rate | 92.3% | 97.8% | +5.5% |
| Voltage Deviation | ±8% | ±5% | -37.5% |
| Harmonic Content | N/A | Reduced by 28.5% | -28.5% |
| Power Factor | <0.90 | ≥0.95 | +5.6% |
| Mean Time Between Failures (MTBF) | 1250 h | 1860 h | +48.8% |
| Service Availability | N/A | 99.95% | – |
| Table 4. Power Supply Reliability Metrics |
5.3 Economic Indicators
- Annual operation cost: Reduced from ¥2.875 million to ¥2.483 million (13.6% savings)
- Power loss: Reduced by 15.8% (¥426,000 annual savings)
- Maintenance cost: Reduced by 23.4%
- Labor cost: Reduced by 18.2%
- Charging service revenue: Increased by 12.5% (¥853,000 annual growth)
- Static payback period: 2.3 years
- Dynamic payback period: 2.8 years
- Net present value (NPV): ¥5.236 million
- Internal rate of return (IRR): 28.4%
6. Conclusion
This study presents a comprehensive load optimization strategy for EV charging infrastructure in airport hubs, integrating TOU pricing and intelligent scheduling. Key findings include significant reductions in peak-valley load differences, operational cost savings, and enhanced power supply reliability. The proposed model and strategies provide a practical framework for managing charging infrastructure in high-traffic, spatio-temporally dynamic environments, supporting both grid stability and user demand satisfaction.