In recent years, many types of research focused on optimizing, modernizing, and developing the fuels for VVER-1000s. A new generation and developed fuel are TVS-2M. Using TVS-2M has many benefits, such as burnup and cycle length increase compared to former fuel assemblies. However, the main issue is to find the best arrangement of the fuel assemblies to increase burnup and safety features. An artificial neural network (ANN) is a powerful tool to predict core burnup for different reactor core configurations in nuclear power plants. One can use ANNs to find the best Fuel arrangement. This article uses a multi-layer neural network to determine the optimum TVS-2M fuel arrangement, which can calculate the burnup of different fuel arrangements. The method utilizes some core parameter data to train a neural network to set up a real-time predicting system for burnup and actinides concentration predictions for different fuel assembly arrangements. The MCNPX code output applies as training data on a multi-layer perceptron neural network. After ensuring the high accuracy of the designed neural network, 20 different arrangements are inserted into the neural network to calculate the burnup parameter for them and select the best arrangement. At last, trained ANN can calculate the actinides concentration and other desired parameters. The results showed that replacing UTVS with TVS-2M FAs, in addition to increasing the length of the operation cycle from 289.6 days to 338.7 days, can increase the cycle's burnup, thus increasing the economic efficiency of the power plant. Also, the results confirm the power of the developed neural networks and approve the use of these networks instead of executing time-consuming codes like MCNPX.