The invention relates to an 
ethylene raw material optimization method based on a mechanism and data 
hybrid modeling and 
linear programming technology, which utilizes a large-scale 
linear programming method to carry out 
ethylene raw material selection optimization in a constraint range of a whole 
plant so as to maximize profits of the whole 
plant. The method comprises the following steps: firstly,establishing a mechanism model for an 
ethylene cracking furnace, simulating and generating massive data samples based on the mechanism model, constructing a neural 
network data model, and establishinga 
hybrid model for predicting the 
cracking product yield; secondly, aiming at the nonlinear relation that the yield of the 
cracking product changes along with the 
raw material attributes and the process parameters, adopting a multi-section linear 
processing mode, and constructing a reference-increment 
database between the yield and the raw material attribute key parameters and between the yield and the key process parameters through a 
mixed model; and writing the reference-increment multi-segment linear structure 
database of the yield into a plan optimization 
linear programming model in combination with other constraint information, and solving the model in a distribution 
recursion mode to obtain a raw 
material selection optimization result. In order to ensure the accuracy of a yield reference-increment structure and the prediction precision of a linear 
programming model, the 
hybrid model can automatically correct 
model parameters and quickly update a yield reference-increment multi-segment linear structure 
database. The method provided by the invention can provide a quantitative basis for raw material 
purchasing and planned production scheduling of ethylene enterprises, so that the 
economic benefits and the raw material 
utilization rate of the enterprises are improved.