Memristor-Based Circuit Optimized Gate Recurrent Unit for Wind Power Prediction

As the global energy structure shifts toward cleaner sources, wind power prediction has become increasingly important in modern energy management. Traditional prediction methods mainly rely on manual adjustment of model parameters and are constrained by hardware implementation conditions. The crossbar array structure of memristors can perform matrix operations directly in memory, offering significant energy efficiency advantages over traditional computing architectures. A hardware circuit design based on memristors for the improved red-billed blue magpie optimizer (IRBMO) and gate recurrent unit (GRU) is presented. The circuit includes the foraging module, the cooperation module, the mutation module and the GRU module. They realize the parallel computation of the prediction process. The circuit not only improves the computational efficiency, but also stores the optimal fitting value effectively. To further validate that the method is practical and effective, the simulation experiments are carried out in the short-term wind power prediction tasks. The results affirm the method high level of accuracy in short-term prediction tasks, which will provide a reference for the hardware implementation of neural network optimization.