关键词: 水资源安全; 水资源预警; 临界状态; 主成分分析; 支持向量机;

Abstract: The early warning of water resources security is an effective way to achieve efficient and orderly management of water resources in the new situation, the key of which is to determine the critical state of water resources security.From the perspective of water resources quantity, three critical states of water resources security, namely, critical drought, critical security, and critical flood, are divided, and a critical state analysis method based on principal component analysis and support vector machine is proposed.The results show that the principal component analysis can be used to eliminate the correlation between the indicators, and the critical surface of the water resources system can be divided by the support vector machine, so that the early warning of water resources can be realized according to the development trend of the critical state.This method was applied to flood early warning, drought early warning and water resource security early warning, and the critical hyperplane equations in three cases were obtained to give basis for judging the early warning objects.The research results have certain reference significance.

Key words: water resources security; water resources early warning; critical state; principal component analysis; support vector machine;