Improvement of a DC electrical railway simulator using artificial intelligence

Abstract

Electrical railway simulators play a critical role in mass rapid transit system (MRTS) studies. In most cases, MRTSs are DC-electrified systems which include elements that exhibit different electrical states, i.e. traction substations may be in ON or OFF modes and braking trains may be in power or voltage (rheostat) modes. This adds complexity to the electrical problem to be solved by the simulator. The simulator developed by the authors in previous works includes a module in charge of determining the electrical states of all the elements in the system. The block, based on heuristic rules, demands high computation times under certain circumstances. This paper presents an upgrade of the heuristic block where artificial intelligence (AI) is used to obtain the electrical states of substations and trains. A neural network (NN) classification model is applied and compared with the previous approach by means of set of simulations. The results show that the NN approach outperforms the previous one.