A general framework algorithm is proposed for energy minimization control for a discrete chaotic system. A quadratic performance function is first given and the chaotic system is decomposed into a linear and a nonlinear parts. Then, the twolevel algorithm is presented to solve the nonlinear optimal control problem: The first level predicates the nonlinear part of the chaos system; the second level solves a nonlinear quadratic optimization control problem by dynamic programming. The solution is fed back into the first level. The first level re-estimates the nonlinear part according to the solution from the second level. The information has been exchanged between the two levels by this means such that the optimal control law is obtained eventually. This method not only can make the control of chaos system be realized but also makes the energy consumed minimal during the whole control process.Simulations show the effectiveness of this algorithms.