The role of future information in control system design for shipboard power systems

Abstract

Both naval and commercial ships are incorporating new power and energy system technologies to improve fuel economy and performance while servicing high power pulsed loads. These assets can be best utilized with load demand forecasting and/or prediction, especially when considering limits on generator ramp rates, distribution lines, and energy storage capacity. Obtaining future load demand data and designing a controller to accommodate it can be challenging, but with potentially large payoff. However, this information is not useful in all cases. This paper develops a method to quantify the potential value of future information depending on the specific power system characteristics. This quantitative approach aids designers in deciding how and when to deploy future forecasting in controller design, and provides insight into the potential benefits of these more complex controllers. To quantify this trade off, two optimization-based control methods are developed. One uses only current information, while the other has an exact forecast of the future. As examples, the method is applied to a notional naval ship and drill platform service vessel with representative power and energy system architectures under indicative operational load demands.

Upper Bound Performance of Shipboard Power and Energy Systems for Early-Stage Design

“…The energy storage and generator ramping limits add system dynamics such that optimal control decisions now depend on future events. Considering future loads improves performance in many conditions [23].…”

“…Allowing such a large prediction horizon is an optimistic assumption, but it does provide a rigorous upper bound on system performance and avoids the difficulties of selecting among thousands of possible controller designs. The results can be compared against the performance of a controller with no future knowledge [19], [20], [23].…”

“…Although the actual controller can never have perfect future information, it may preview load demands several seconds into the future through feedback from loads or by imposing activation delays on load requests. The quality of such a controller's preview information dictates its proximity to this upper bound benchmark [23].…”

Artificial Intelligence-based short-term forecasting of vessel performance parameters

“…To ensure improved energy efficiency and safety, various strategies and methods are employed to control a vessel's dynamic power systems by utilising real-time feedback and state predictions or a forecast of the system's future behaviour (Xie et al, 2022). For example, Model Predictive Control (MPC) has gained popularity due to the incorporation of a forecast of system parameters for a set time horizon within its structure, which has been shown to increase performance compared to controllers that lack advanced knowledge of changes in power load (Opila et al, 2018). In this sense, an MPC strategy relies on a forecasting method or model that can accurately identify ensuing operational shifts (Camacho and Alba, 2013;Schwenzer et al, 2021;Dahl et al, 2018).…”