Diverse forms of nonlinear optimization problems can be recast to the special form of second-order cone problems (SOCPs), permitting a wider variety of highly effective solvers to be applied. Popular solvers assume, however, that the necessary transformations to required canonical forms have already been identified and carried out. We describe a general approach to the construction of algorithms that automatically detect equivalent to SOCPs and apply the necessary transformations. To test this approach, the algorithms are implemented in the context of the AMPL modeling language and various solvers. The automated transformations are seen to allow for more effective and reliable modeling, while in some cases transforming problems to forms that are much easier to solve.
Technical report, Department of Industrial Engineering and Management Sciences, Northwestern University, Evanston, IL, U.S.A. Posted to Optimization Online (May 2019).