Power System Economic Dispatch Under Line Flow Constraints

The patent replaces traditional mechanical iterative optimization algorithms with a neural network-based computational system. The neural network uses parallel processing and continuous relaxation techniques to solve the economic dispatch problem, substituting the sequential mechanical iteration with a more efficient computational model that converges faster while maintaining distributed architecture.

The patent transforms the discrete optimization problem into a continuous parameter optimization problem by using neural network parameters (weights and biases) as decision variables. This parameter transformation allows the system to escape from the slow convergence of discrete iterative methods and achieve real-time solving through continuous gradient-based optimization.

The patent segments the global inequality constraints into local inequality constraints that can be processed by individual neural networks at each node. By decomposing the system into independent nodes that each handle their own constraints and communicate only necessary information, the system maintains real-time solving capability while incorporating global constraints through distributed coordination.

The patent introduces Lagrange multipliers as intermediary variables that mediate between local node decisions and global constraint satisfaction. These multipliers act as messengers that transmit constraint violation information back to the nodes, enabling the system to process global inequation constraints without compromising the distributed real-time solving architecture.