AI Enhances Bridge Arm Reactor Efficiency Via Genetic Algorithms

When a 1,733-kilogram reactor generates over 60 kilowatts of power loss during operation, with hotspot temperatures climbing to 71.1°C, it represents not just significant energy waste but also accelerated degradation of insulation materials. In power systems, bridge-arm reactors serve as critical current-limiting and filtering components, where structural design directly impacts operational stability and economic efficiency. With increasingly stringent energy conservation requirements, engineers face the challenge of balancing weight, power loss, and temperature rise while optimizing performance.

Core Technical Architecture: Equivalent Electromagnetic-Thermal Field Modeling

Modern bridge-arm reactors typically employ a multi-layer encapsulated structure, with internal windings secured through epoxy resin casting. Insulation spacers between layers form cooling air channels. This parallel configuration allows each encapsulation layer to be treated as an equivalent solenoid winding. For electromagnetic parameter calculations, researchers apply Kirchhoff's voltage and current laws at fundamental frequencies, simplifying complex winding mutual inductance into equivalent coaxial solenoids.

The resulting electromagnetic model accounts for both DC components and harmonic influences. Through coupled thermal-fluid simulations, winding losses transform into heat source terms, accurately mapping temperature field evolution from startup to steady-state operation.

Optimization Strategy: Multi-Objective Genetic Algorithm Implementation

The research team implemented NSGA-II (Non-dominated Sorting Genetic Algorithm II) to achieve optimal balance between low loss, minimal temperature rise, and lightweight design. Four key variables were selected: branch count (n), winding average radius (R), turn count (N), and conductor diameter (d).

A novel triple-constraint strategy was developed, maintaining equal resistance voltage drop, equivalent temperature rise, and uniform loss density. The optimization process employed Latin Hypercube Sampling (LHS) to construct a four-dimensional design space, while Kriging response surface modeling enhanced temperature rise prediction accuracy, ensuring scientific rigor and convergence.

Performance Milestones: Before and After Optimization

Testing a decommissioned 21.25 mH bridge-arm reactor from a power plant, the team generated a uniformly distributed Pareto optimal solution set after 150 generations of genetic evolution. The optimized prototype demonstrated comprehensive improvements while maintaining rated inductance at 21.82 mH:

• Energy Efficiency: Total losses reduced from 61.65 kW to 34.21 kW (44.51% reduction)
• Thermal Management: Hotspot temperature decreased from 71.1°C to 42.9°C (39.66% reduction), significantly extending insulation lifespan
• Structural Optimization: Total winding mass reduced from 1,733.2 kg to 1,302.7 kg (24.83% reduction)

This design breakthrough demonstrates the potential of multi-objective optimization in power equipment development, establishing a technical framework for future high-reliability, energy-efficient reactor designs.