Impact of variable generation in generation resource planning models

Long-term power system planning is beset by a
trade-off between detail and scope: The chosen approach usually lies somewhere between modeling a great many generation
portfolios coarsely and very few in a more detailed manner. This paper seeks to argue that the performance of generation
portfolios is influenced by a sufficiently large number of variables, of varying uncertainties, such that the long-term investment problem can only be effectively tackled with very many runs of computationally light models that capture the most essential
features of the problem. Taking a linear optimization program as the intended computational core, this paper describes two
algorithms to build constraints for the linear program which capture many of the effects that are difficult or impossible to
capture directly in non-chronological models, namely: unit starts, unit ramping, unit average output and adequate total system
capacity. An application of these methods is also presented.