The invention provides a two-stage energy-saving scheduling optimization method based on a geometric adaptive
recovery mechanism, which comprises the following steps of: A, initializing core parameters such as a weight vector, a neighborhood size and iteration times, and giving a problem boundary, an operation
machine set and a
model parameter; b, generating an initial weight vector set, and calculating a neighborhood index of each weight; c, generating an initial
population; d, calculating an individual objective function value and an ideal point, and initializing an external file as a non-dominated solution set; e, constructing a
resource pool, calling a weight adaptive and
population synchronization algorithm to adjust a weight vector, and reconstructing a
population; f, generating a new solution in each sub-problem neighborhood by
differential evolution, updating an ideal point, and updating a neighborhood solution according to a scaling function value; g, constructing a local
resource pool, calling a
geometric matching and
recovery algorithm, and rescuing non-dominated solutions with position advantages to a candidate set; h, merging the candidate set to an archive candidate set, and performing non-dominated screening; if the archive capacity is exceeded, invoking an archive updating
algorithm to update the archive; i, performing periodic weight adaptive adjustment every specified algebra interval; and J, if the termination condition is not met, returning to the step F, otherwise, outputting the external archive as a final non-dominated solution set. The method has the advantages that multiple constraints of the two-stage energy-saving flexible flow shop can be met, the defects of a traditional
algorithm are overcome through a weight self-adaption and solution rescue mechanism, a scheduling scheme with multi-direction
parallel search and search direction self-adaption adjustment, uniform output distribution and excellent performance is achieved, and multiple feasible bases are provided for enterprise production decisions.