Economic emission dispatching strategy considering dynamic parameter effects: A novel approach based on projection neural networks and deep learning

The economic emission dispatch (EED) problem is influenced by dynamic parameters such as power demand and climatic factors affecting renewable energy (RES) generation, which adds to the complexity of the dispatch process. Constrained by the serial iterative computing architecture, conventional optimization algorithms often face the challenges such as long computation time and computational inefficiency caused by repeated solving when dealing with EED involving continuous changes in dynamic parameters. To address the problem, this paper combines projection neural network (PNN) and deep learning to cope with the effect of dynamic parameters on the EED. First, a deep PNN (DPNN) is proposed by embedding PNN in deep learning. Then, the dynamic parameters in the EED are taken as input variables to the DPNN. Compared to PNN, DPNN do not require iterations and can respond immediately to dynamic parameter changes to directly provide predicted solutions for EED, which allows the DPNN reduce computation time and improve computational efficiency. Simulation results show that compared with PNN and convex solvers, DPNN can significantly reduce the computation time with good computational performance and can be adapted to EED problems containing dynamic parameters.