Multi-Objective Optimization for Wireless Power Delivery

The patent replaces traditional gradient-based optimization methods with a neural network-based system. The neural network is trained offline to learn the mapping from transmission parameters to power delivery performance, eliminating the need for real-time gradient calculations and objective function evaluations. This substitution of mechanical/computational optimization processes with an intelligent system achieves both high accuracy and computational efficiency.

The patent performs preliminary training of the neural network offline using simulated or measured data before actual wireless power delivery operations. By pre-computing the optimization mappings and storing them in the neural network, the system avoids time-consuming real-time optimization calculations during actual power delivery, thus improving computational efficiency while maintaining optimization accuracy.

The patent substitutes traditional iterative optimization algorithms that search for Pareto optimal solutions with a pre-trained neural network that directly outputs optimal transmission parameters. The neural network has learned the Pareto front relationships during offline training, enabling it to provide high-quality solutions instantaneously without time-consuming iterative searches.

The patent creates a computational model (neural network) that copies and represents the complex multi-objective optimization landscape. By training the neural network on simulated or measured data that captures the Pareto optimal relationships, the system creates a simplified surrogate model that can quickly provide high-quality solutions without repeating the expensive optimization process in real-time.