The growing demand of DG (diesel electric generators) has led to air pollution and green house gas emissions, especially CO2 (Carbon-di-oxide). Hence, it is necessary to predict the level of CO2 released from the DG, to ensure the minimum level of emission. Forecasting the CO/CO2 ratio, flue gas temperature (T-F) and gross efficiency (eta), ensures the effective and smooth operations of DGs. Keeping this in view, in this paper, ANN (artificial neural network) models are proposed for the prediction of CO2, CO/CO2 ratio, T-F and (eta) of DG. The training and testing data required to develop the ANN were obtained through a number of experiments in 3 phase, 415 V, DG of different capacities operated at different loads, speed and torques. Three different capacities of DGs such as 180, 250, and 380 kVA have been investigated. Back propagation algorithm was used for training the ANN. The application of the newly developed models shows better results in terms of accuracy and percentage error. The co-efficient of multiple determination values are found to be above 0.99 for all the models. It is evident that the ANN models are reliable tools for the prediction of the performance and exhaust emissions of DGs. (C) 2015 Elsevier Ltd. All rights reserved.