All-Electric Ship Energy System Design Using Classifier-Guided Sampling

The addition of power-intensive electrical systems on the U.S. Navy´s next-generation all-electric ships (AES) creates significant new challenges in the area of total-ship energy management. Power intensive assets are likely to compete for available generation capacity, and thermal loads are expected to greatly exceed current heat removal capacity. To address this challenge, a total-ship zonal distribution model that includes electric power, chilled water (CW), and refrigerated air (RA) systems is developed. Classifier-guided sampling (CGS), a population-based optimization algorithm for solving problems with discrete variables and discontinuous responses, is used to identify high-performance configurations with respect to fuel consumption. This modeling approach supports early-stage design decisions and performance analyses of notional systems in response to changing operating modes and damage scenarios. A set of configurations that enhance survivability is identified. Results of a comparison study demonstrate that CGS improves the rate of convergence toward superior solutions, on average, when compared to genetic algorithms (GAs).