Real-time sequential convex programming for nonlinear model predictive control and application to a hydro-power plant

In this paper we propose a new algorithm for solving nonlinear optimal control problems which is called real-time sequential convex programming (RTSCP). The main difference between this approach and the previous real-time iteration algorithms is that RTSCP linearizes only the nonconvex parts of the problem while it preserves all the convex structure which can be exploited by standard convex optimization techniques. The algorithm is applied to the control of a hydro power plant with 259 states and 10 controls. The numerical results show the benefits that the proposed method offers when compared to standard ones.