With the increasingly stringent environmental issues and regulations, there are higher requirements for improving engine performance and reducing pollution. Combining artificial neural network and particle swarm optimization algorithm to optimize the fuel consumption and emissions for micro-ignition dual-fuel engines. A model-based calibration scheme is maintained to reduce the number of experimental points by employing space-filling and V optimization design, to save the experimental cost and improve efficiency. The experimental data used to establish an RBF neural network prediction model that achieves a perfect mapping of engine input and output parameters. Controllable variables such as speed, torque, main injection timing, pilot injection timing, pilot injection quantity, rail pressure, excess air coefficient, and substitution rate limit parameters are input as neural networks. Subsequently, the combination of control parameters was optimized through PSO, thereby to achieve fuel consumption and emissions trade-off. Matching experiment results show actual emissions of NOx, THC, and CO decreased by 20.5%, 30.3%, and 43.1%, respectively, and the BSFC declined by an average of 2.1% contrasted with the original data. It achieves the optimum of emission and fuel consumption at the same time.