The invention relates to the technical field of
computational chemistry and computational
electrochemistry, in particular to a material performance analysis and
system prediction method. According to the material performance analysis and prediction method provided by the invention, a constant potential
message passing model is adopted to analyze atomic configuration; wherein the multi-head attention mechanism layer combines atomic features and initial features of over-
electron generation atomic configuration nodes, and the multiple atomic features form fusion features carrying over-
electron information through linear fusion; and the
message passing layer performs information interaction between each central atom and neighbor atoms by using a geometric
tensor product, performs atomic feature and
electron reconstruction, and finally performs replacement-invariant summation operation on messages of all neighbor nodes to obtain giant
potential energy, atomic force and a
work function of atomic configuration. The invention realizes an electrochemical calculation method with potential control capability, physical consistency and cross-scale
simulation efficiency so as to support interface
process research under real electrochemical conditions. The problems that in the prior art, electrochemical reaction research cannot give consideration to scale and precision, potential control precision is lacked, and physical consistency is insufficient are solved.