The forecast of urban water demand is the basis of plan and management of ecological city. However, it is a complex problem affected by various uncertainty factors. In order to express the integration and blending relationships of certainty and uncertainty among influence factors and between samples and historical data as a whole,the paper used an orderly clustering theory and coupling method of set pair analysis to present a forecast model of urban water demand based on connectional membership degree. In the model, the ordered historical data of urban water demand of corresponding years was first clustered according to Fisher's optimal partition method, then the identical-discrepancy-contrary analysis was carried out to express the relationship between impact factors of one year and historical data. Moreover,it structured a similar model to forecast city water demand of corresponding year.The results from the practical example and comparative analyses with other methods show it is reasonable and effective for application of the model to prediction of city water demand.