Considering worst-case conditional value-at-risk(WCVaR) as a risk metric, an optimization model of multi-
confidence levels minimization WCVaR is built. Under the assumption of discrete box distribution of random variables, the transformation from complex min-max optimization model to traditional single-target linear program is done with the combination of multi-objective decision and duality theory. As an application, the power allocation of generators in power markets is set up. Monte Carlo simulation shows the effectiveness of the model.