MILP Optimization of Nonlinear Energy Systems by Multivariate Piecewise-Affine Surrogate Modeling
This work presents a methodological framework for MILP modeling and optimization of distributed energy supply systems(DESS)with nonlinear system characteristics.The required MILP linearization is based on a multivariate piecewise-affine(PWA)surrogate modeling approach.A machine learning algorithm using Model Trees is employed to enable adjustments of model accuracy and to ensure an efficient discretization.The MILP component models are assembled to a DESS superstructure,which is optimized with respect to system structure,equipment sizing and load dispatch.During optimization,an iterative model refinement procedure is used to adapt the multivariate linear discretization to realize a prescribed prediction accuracy.The features of the presented modeling framework are demonstrated for the optimal synthesis of a refrigeration system.
Distributed Energy Supply Systems MILP Optimization Surrogate Modeling Multivariate Regression Piecewise-Affine Modeling
Benjamin Meyer Philip Voll Stefan Kirschbaum André Bardow
Institute of Technical Thermodynamics,RWTH Aachen University,Aachen,Germany Society for the Promotion of Applied Computer Sciences,Berlin,Germany
国际会议
桂林
英文
1-12
2013-07-16（万方平台首次上网日期，不代表论文的发表时间）