Abstract Generation reserves are needed to maintain the real time balance between power supply and power demand. Because power is noninventoriable, power generation follows power demand. Demand for power varies considerably depending on the time of day, day of the week and season. The predictable portion of power demand is met by purchasing firm energy on a day ahead or real time market. The random unpredictable portion of demand is met by purchasing a set of online and offline generation reserves on an ancillary market. The total energy purchasing cost includes payments for firm energy and payments for generation reserves. The latter include fixed capacity payments for reserve generation capacities and variable payments for the random energy produced from these reserves. The main contribution of this paper is to present an optimization model that captures the dynamism in the selection of the dispatch interval to determine the amount of firm energy and reserve capacities given a set of market prices. This is done by explicitly including in the model the duration of the dispatch period and the frequency this decision is reevaluated. In this model the randomness of the demand is captured by using a Doubly Truncated Normal Distribution. The cost incurred to activate generation reserves is modeled as a Poisson process. The total model captures the price differences from using different reserve sources. An empirical example is presented to illustrate the cost benefits of using the method proposed in this research with two different strategies: a static strategy and a dynamic strategy. It is shown that dynamically setting generation reserves results in cost savings.