A Dynamic Multi-Stage Design Framework for Staged Deployment Optimization of Highly Stochastic Systems
A Dynamic Multi-Stage Design Framework for Staged Deployment Optimization of Highly Stochastic Systems
Embedded Files
The need for staged design optimization for multidisciplinary systems with strong, cross-system links and complex systems has been acknowledged in various contexts. This is prominent in fields where decisions between subsystems are dependant, as well as in cases where tactical decisions need to be made in uncertain environments. The flexibility gained by incorporating evolutionary design options has been analyzed by discretizing the time-variant uncertainties into scenarios and considering the flexible decision variables in each scenario separately. However, these problems use existing information at the decision time step. This paper presents a dynamic multi-staged design framework to solve problems that dynamically incorporate updated system information and reformulate the problem to account for the updated parameters. The importance of considering staged decisions is studied, and the benefit of the model is evaluated in cases where the stochasticity of the parameters decreases with time. The impact of considering staged deployment for highly stochastic, large-scale systems is investigated through a numerical case study as well as a case study for the IEEE 30 bus system. The case studies presented in this paper investigate multi-disciplinary design problems for large-scale complex systems as well as operational planning for highly stochastic systems. The importance of considering staged deployment for multi-disciplinary systems that have decreasing variability of their parameters with time is highlighted and demonstrated through the results of numerical and engineering case studies.
The need for staged design optimization for multidisciplinary systems with strong, cross-system links and complex systems has been acknowledged in various contexts. This is prominent in fields where decisions between subsystems are dependant, as well as in cases where tactical decisions need to be made in uncertain environments. The flexibility gained by incorporating evolutionary design options has been analyzed by discretizing the time-variant uncertainties into scenarios and considering the flexible decision variables in each scenario separately. However, these problems use existing information at the decision time step. This paper presents a dynamic multi-staged design framework to solve problems that dynamically incorporate updated system information and reformulate the problem to account for the updated parameters. The importance of considering staged decisions is studied, and the benefit of the model is evaluated in cases where the stochasticity of the parameters decreases with time. The impact of considering staged deployment for highly stochastic, large-scale systems is investigated through a numerical case study as well as a case study for the IEEE 30 bus system. The case studies presented in this paper investigate multi-disciplinary design problems for large-scale complex systems as well as operational planning for highly stochastic systems. The importance of considering staged deployment for multi-disciplinary systems that have decreasing variability of their parameters with time is highlighted and demonstrated through the results of numerical and engineering case studies.
Published 1 paper as the 2nd author on Structural and Multidisciplinary Optimization
Example of scenario discretization with three decision points and possible evolution paths (i.e., scenarios)
Non-anticipativity constraints for the stochastic decision variables are formulated based on the scenario tree
Possible updated scenario schemes at the next decision step
The DMSD solution framework
The variability reduction and change in expected value over time when estimating the force at time T
The results of the production of the 6 generators per hour for the dynamic model
The benefit of a Dynamic model for stochastic variable scenario realizations: a Shows the available knowledge of the stochastic parameter at each time period. b Shows the information available for the multistage model at time t=1 and c, d Show how the model is updated at t = 10 and t = 20, respectively
Battery pack immersion cooling system
The framework for the first stage solution process
The optimized coolant flow rate over 720 s with decisions made over four stages. The shaded area represents the possible flow rates and the points represent the optimization results. The energy consumption cost of each of the stages is also presented
Hourly load profile behavior divided into stages. Adapted from Monterrat et al. (2017)
IEEE 30 bus system as presented in Data (1961)
Average daily wind speed ALP (2009)
Electric power produced based on wind speed Wind (2015)
Page updated
Google Sites
Report abuse