Decision on optimal selection of interconnected river system network schemes is an analysis process which combines deterministic evaluation index and evaluation criteria with uncertain evaluation factors and their weight changes. The evaluation indexes involved include resource criteria, economic criteria, social criteria, environmental criteria, engineering criteria and so on. Intermediary transition of these evaluation indexes leads to uncertainty in the results of different multi-attribute decision-making methods. How to evaluate the stability of these decision-making methods is worth studying. According to the theorem of the unity of opposites of variable sets, the principle and method of variable sets for optimal selection of interconnected river system network schemes were proposed, and the sensitivity analysis method of variable sets decision results based on cloud model was proposed. Taking Xinanjiang (S1), Fuchunjiang (S2) and Taihu-Fuchunjiang (S3) water diversion projects in northern Zhejiang Province as examples, the ranking of the three schemes was evaluated and analyzed in terms of comprehensive evaluation and different criteria evaluation, and the robustness of the evaluation results of variable model under change of single index and multiple index weights was analyzed. The main conclusions were as follows: using the rand feature values and relative difference degrees to analyze the results of multi-criteria comprehensive evaluation, it was indicated that S3 was superior to S1 and S2;variable model (α=1, p=1) and (α=2, p=1) comprehensive evaluation of the scheme ranking stability was better from the index sensitivity analysis. It can be seen from the weight sensitivity analysis that the ranking of the comprehensive evaluation schemes, especially S2 and S3, of the four kinds of variable models with different parameter combinations was easily affected by the change of index weight. According to the analysis of index sensitivity and index weight sensitivity combined with the comprehensive evaluation and the evaluation of different criteria, the evaluation results were affected by the index system to some extent, which indicated that it was needed to make specific analysis on the research objects to choose the appropriate evaluation method.