Based on gain optimization methods, superdirective beamformers can achieve high beam directivity with small aperture array. However, the extreme sensitivity to array uncertainty is a main obstacle to engineering application. In this work, a robust gain optimization algorithm under uncertainty set constraint is proposed. Considering steering vector mismatch is the result of combined effect of various array errors, and it is a measurable indicator especially in receiving system. We apply its uncertainty as a constraint on gain optimization method, which is more intuitive in physical sense. Different from existing solutions, it makes a better tradeoff among directive gain, robustness and radiation efficiency. Experimental analysis verifies its good performance in engineering.