Simulation is the major approach in power system studies for prototyping new devices, evaluating new scenarios, and implementing new controls. Such a simulator relies on a framework for the dynamic memory management of internal variables. Our work presents a C-based framework that significantly reduces overhead with reduced object conversion and improved matrix storage techniques. Optimized sparse matrix add and set methods take advantage of indexing characteristics of power system equations, avoiding reconstruction of matrices within the framework if possible. Evaluation of the proposed methods substantiates the efficiency and utility of our framework in a differential algebraic equation solver.