Abstract:
This thesis investigates the use of Benders’ decomposition as a tool for the construction of schedules for the New Zealand hydro-thermal electricity generation network. This system is highly reliant on hydro production, has reservoirs that are small and inflows that have large variance. In order to obtain an effective representation of the stochastic nature of the problem the current ECNZ model, which uses stochastic dynamic programming, aggregates all reservoirs in the country into two reservoirs. In this thesis a model is developed that reflects the stochastic process of the inflows and incorporates the network structure of the problem. The problem is solved using a multistage Benders’ decomposition of up to six decision stages. Inflows are modelled using both independent historical data and generated using a Iag-I model. A future cost function is used in the Iast stage to minimise the end effects. The model was tested in a case study of the Waikato river chain and computational results obtained.