Scorpionfly Optimization Algorithm

To address the multi-peak structure, local extremum traps, and high-dimensional coupling characteristics inherent in complex non-convex optimization problems, this paper constructs a scorpionfly optimization algorithm model with an energy memory mechanism and topological adaptive structure. Building upon the existing curvature-sensing and information entropy adjustment framework, a non-equilibrium energy system, an adaptive topological phase transition mechanism, and an anisotropic attack tensor are further introduced to establish a unified discrete and continuous-time dynamical system expression. By introducing individual energy variables to characterize stagnation and burst behaviors, a population structure phase transition model is constructed, and dimensionality-level enhancement is achieved by constructing an attack tensor using dimensional variance. The stability, complexity, and dynamic characteristics of the system are analyzed. The results show that the model effectively enhances adaptability to complex search spaces while maintaining linear complexity.