Evaluation of prediction choice in economic dispatch with uncertainty

The unavoidable wind power forecast error results in the pervasive uncertainty in power system economic dispatch. In order to solve optimization problem such as economic dispatch with random variables, the commonest and simplest method is using a deterministic value to replace this random variable. Several kinds of predictions are used such as maximum-likelihood approximation, expectation and median. This study analyses the effect of these prediction choices on uncertainty of power system via comparing their statistical consistency. It is shown that when the wind power output is very volatile, the quality of these prediction choices is not guaranteed. Additionally, a simplified three-unit thermal system and a ten-unit system are studied for illustrating the consequence of statistical inconsistency.