Modeling and solving resource allocation and path planning problems often requires extensive computational resources and hardcoded constraint. In this work, with the help of synthetically generated graphs to emulate real-world data, we explore a preliminary framework that uses neural networks and reinforcement learning for optimization in the context of urban planning operations. We show that this model scales better than Google's OR-Tools on large instances (n > 50), and can be trained without the use of distance matrices or explicitly defined constraints. By training on just synthetic data, our model can be applied to different scenarios without the need for retraining,
In this paper, a new method for estimation and removal of decaying dc offset from fault current signal using the Least Median of Squares Regression (LMSR) method is proposed. The fault currents include the decaying component and the mentioned component reduces the speed response and validity of grid protection relays. The proposed algorithm can foresee the dc decaying component and disassemble it from the fault current wave in a period of the wave cycle. as well as it has used from conventional fast Fourier equations to measure the phasor magnitude and fault current wave in grid protection relays which are mean the dc decaying magnitude and time constant is estimating by integrating fault current during one period cycle. The dc decaying component by deducing the ever sampling of a cycle wave in one period. About the fact that the mentioned proposed method creates the best performance, the simulation has been applying in PSCAD/EMTDC. the results of this work illustrate that the proposed method can foresee the component from fault current wave, as well as this method, could apply as phasor derivation in digital protection relays.