This dissertation consists of three studies related to goal-based asset management tailored to individual needs. Three studies consist of a stochastic programming model that uses client friendly inputs for the model, a stochastic programming model that makes decisions for large number of time stages, and a study of generating scenario trees for use in stochastic models.
First, we propose a multistage stochastic model to achieve the client’s consumption goals. In this model, unlike the existing models, the optimization proceeds with input values that are easily understood by the general public.
Second, we propose a multistage stochastic programming model that make decisions on investment and consumption for a long period over 50 years. We propose to use a decomposition method to solve the curse of the dimensionality, which approximate the optimal solution efficiently.
Finally, we propose a method to generate the scenario tree, which is used for solving stochastic programming problems. We propose to use a meta-heuristic algorithm to generate a scenario tree that matches its statistical moments.
Through these three studies, we expect that we will be able to perform customized asset management more efficiently according to individual consumption goals.