A Numerically Efficient Iterative Procedure for Hybrid Power System Optimization Using Sensitivity Functions

Initiatives for all-electric ships and hybrid vehicles necessitate advanced optimal power management to coordinate between heterogeneous sources and loads for dynamic reconfiguration and efficient transient operation. Transient state trajectory optimization associated with hybrid power systems is challenging because of the high dimensional and nonlinear dynamics of the plant models involved. Existing methodologies such as Dynamic Programming (DP), Iterative Dynamic Programming (IDP), etc., incur the "curse of dimensionality" and are exponentially complex in time and/or space and need large scale computational resources. This paper, motivated by the all-electric ship initiative, proposes a new methodology for solving high dimensional, nonlinear, long time horizon trajectory optimization problems, based on the sensitivity function method. The computational complexity of the method is analyzed and compared with other standard iterative optimization procedures. The numerical advantage of the proposed method over other procedures is demonstrated using examples and analysis.