As for comprehensive evaluation of alternative schemes which can not confirm its goal attribute weight or membership, a support vector machine learning algorithm is presented. Based on water supply scheme as reclaimed water source for the power plant, the learning algorithm sets up a model to synthesize attribute optimization utilizing the support vector machine. The result shows that the comprehensive evaluation value of three schemes were 0.24, 0.55 and 0.61, which shows that the third scheme is reasonable. Contrasted with actual choice scheme and AHP to determine scheme, the result is the same as them. The effect of comprehensive evaluation is feasible for the selection of water supply scheme by support vector machine method.