Parallel Nonlinear Optimisation for Decision Making Under Uncertainty

This paper presents an overview of our current research on parallel nonlinear optimization for decision making under uncertainty at Imperial College, London. We are interested in the optimization of nonlinear systems in which some of the key parameters are subject to uncertainty characterised by continuous probability density functions with general algebraic and/or differential-algebraic models (equality constraints) and linear/nonlinear inequality constraints, both affected by the uncertainty. The research has resulted in the development of riskaverse (robust) formulations within the meanvariance optimization model capable of controlling the effects of uncertainty. The formulations enable the user to select the level of risk she/he is prepared to take as well as the feasibility requirements of the nonlinear stochastic inequality constraints. The paper also describes the use of high performance computing in the optimization process. This can be divided into parallel algorithms for spe..