Construction and verification of dual-fuel engine combustion model

In order to achieve real-time mapping and online optimization of the combustion process of a dual-fuel engine that is in prolonged operation, this paper is the first to combine the Wiebe function with a deep learning neural network to propose a zero-dimensional (0-D) combustion prediction model for a biodiesel-diesel dual-fuel engine. First, the parameters of the double Wiebe functions are calculated by the Pelican Optimization Algorithm (POA), and the operating parameters and Wiebe parameters are used as input and output parameters of neural networks, respectively, to construct parameter identification models. Then, the combustion process is simplified and reconstructed by combining the Wiebe function with the deep learning neural network, and a 0-D prediction model based on the hybrid model-driven and data-driven method is established, which can further obtain combustion results such as cylinder pressure curve and indicated mean effective pressure (IMEP). The results show that the coefficient of determination (R-2) value of the dual-fuel engine 0-D prediction model based on the double Wiebe function combined with POA-CNN-Bi-LSTM is 0.9827, and the model has good prediction accuracy and generalization. The development of the combustion model provides reliable numerical model support for the online evaluation and optimization of dual-fuel engine performance.