Increasingly stringent government regulations regarding vehicular emissions and rising concerns over fossil fuel depletion encourage the use of biofuels. In general, biofuels are used together with traditional fuels at different specific concentrations. As the fuel blend ratios can sensitively affect the performance and emission characteristics of automotive engines, it is desirable to determine the most optimal blend ratio that can result in fewer emissions and better performance with reasonable fuel economy. Moreover, to fit the properties of the optimal fuel blend and to maximize the combustion performance, it is essential to optimize the parameters of the engine control systems. Furthermore, to account for the engine transient response due to abrupt changes in fuel mixture, as well as any effects that may be caused by fuel variation and other disturbances, it is also necessary to design a fuel controller that can maintain the drivability and stability of biofuel engines. In order to accomplish all these tasks, development of accurate and reliable models for predicting the behavior of biofuel engines is of great importance. Traditional physically-based engine models may not be sufficiently robust and are usually very labour- and computational-intensive for optimization and control purposes. Therefore, this thesis attempts to use emerging machine learning methods to deal with the engine modeling problem. Several machine learning methods have been reviewed in this thesis, and extreme learning machine (ELM) is selected to construct the biofuel engine models. Three biofuel engine related tasks, including (i) optimization of biodiesel engine performance, (ii) calibration of dual-fuel engine system and (iii) adaptive air-fuel ratio control of dual-injection engine system, are presented and solved with the use of ELM-based biofuel engine models. Different novel techniques have been proposed in each of the tasks to further improve the performance of the iii ELM models, and an engine has been retrofitted for experimental implementation and verification of the optimization and control results. The results from all the three presented tasks suggest that ELM is promising and effective for modeling of biofuel engine systems.