Simulation of a gas-steam combined cycle power plant using artificial neural networks

A neural network (NN) is implemented to predict the plant performance of a gas-steam combined cycle power plant. The Multi-Layer Perceptron neural network with one hidden layer and back propagation learning algorithm is used in the present study. Two NN prototypes were designed and produced with reference to the gas turbine and steam section of the combined cycle power plant. Data from physical simulations of plant components are used to train the NN. Results from the system simulation technique are compared with those based on the NN approach. The results indicate it is feasible to use NN to accurately predict plant-operating conditions. The NN gives a good time response and performance prediction capability with change of boundary conditions. Significantly lower computation times are obtained with the NN compared to the physical simulations. The accuracy of the NN output and its suitability for on-line monitoring of a gas-steam combined plant are discussed.