Abstract:
In the modern world, fossil fuels meet the requirement of the transportation sector in an
extensive volume and bring several negative impacts viz., air pollution, noise, and
global warming. Besides the fast decline of below ground petroleum resources that arise
with the consumption of fossil fuels is accompanied by another grave problem for the
transportation sector. Proton Exchange Membrane Fuel Cell Electrical Vehicles possess
superior control characteristics in comparison to other types of EVs. In the past few
years, the popularity of PEMFC EVs has increased. In this research work, machine
learning techniques such as Artificial Neural Network (ANN), Adaptive Neuro Fuzzy
Inference (ANFIS), Recurrent Neural Network (RNN), Long Short-Term Memory
(LSTM) will be used . A dataset of PEMFC is acquired to predict the output voltage in
terms of input parameters which are Current (A), Temperature (T), Inlet Pressure of
hydrogen, outlet air pressure etc. The overall sample size (951 samples) is 70% for the
training data and 30% for the testing data (410 samples). The benefit to these algorithms
is that it will train a model from numeric dataset and get an optimal fit point by
incorporating efficiency parameters. Mean Squared Error (MSE) is a commonly used
metric to evaluate the precision of predictions or control outputs. The control outputs
generated by each strategy will be compared against actual system responses, and the
MSE and mean absolute error (MAE) values will be computed for each case. The
LSTM is proposed to control the performance degradation of PEM Fuel Cell .It is
evaluated that the LSTM delivers the best results. The LSTM models perform better in
degradation because their dynamical memory integrates delay parameters. The LSTM
gave the better results other than the three models. The MSE and MAE given by LSTM
model having 100 neurons in one hidden layer for 100 epochs are 0.01643 and 0.04768,
respectively.
