Abstract:
In order to model and analyze gas networks, several methods have already been developed and presented. Nevertheless, all these methods have their own specific applications and most of them are very complex and usually contain some errors. In this paper, in an attempt to resolve these problems, an Artificial Neural Network (ANN) has been used to model a gas distribution network. The algorithms utilized for ANN training, such as the gradient descent algorithm, are usually subjected to local minima; in this regard, the new Cuckoo Optimization Algorithm (COA) is used in training the weights of the neural network. However, gas networks are often very large and operate a multitude of distant points, which explains why time delays in these networks are inevitable. Accordingly, in order for all points of the output (pressure) to achieve the desired value, a Model Predictive Controller was used. According to the results achieved, it can be said that the Artificial Neural Network Cuckoo Optimization Algorithm (ANN_COA), in comparison to regular ANN, yields a more suitable performance and is less prone to error. In addition, the MPC controller is faster and suffers from fewer errors compared to the Proportional-Integral-Derivative (PID) controller while also preventing fluctuations in gas system input.