地下水环境质量的准确评价可降低由水质不确定性导致的农业水管理决策风险，对于指导区域发展绿色农业、清洁生产等均具有重要的实践意义。针对ELM会随机产生输入层与隐含层的连接权值和隐含层阈值，导致网络泛化能力降低以及过度拟合引起的评价结果失真问题，提出一种基于粒子群算法优化极限学习机的地下水水质综合评价模型。选取黑龙江省建三江地区为研究区域，通过典型农业机井中地下水体的实地采样，利用构建的PSOELM模型对地下水水质进行综合评价。结果表明：PSOELM模型拟合性高于传统的ELM模型和RBF模型，并提高了区域地下水水质的模拟评价效果。建三江地区下辖的15个农场，整体地下水水质较好。研究区范围内地下水水质等级呈现出集中式分布特点，地下水综合水质较差的农场集中在研究区中东部，水质较好的农场集中在研究区的东西两侧。建三江地区地下水水质地域性规律主要与化肥施用量有关。为此，在保障粮食安全与用水安全的同时实现清洁生产应注意控制化肥的使用量，积极探索作物高产新举措。

With great and practical significance to green agricultural development and clean production,accurate evaluation of groundwater environmental quality may reduce the risk,caused by water quality uncertainty,in decisionmaking for agricultural water management.As ELM may randomly generate the connection weight and hidden layer threshold of input layer and hidden layer,resulting in decrease in network generalization ability and evaluation distortion caused by over-fitting,a comprehensive evaluation model of groundwater quality is developed based on PSO-ELM.In the case study area,Jiansanjiang in Heilongjiang Province,by field sampling of groundwater in the typical irrigation wells,the PSO-ELM model is applied for comprehensive evaluation of groundwater quality.The results show that the PSO-ELM model presents higher goodness of fit than the traditional ELM model and RBF model,thus improving the simulation evaluation of regional groundwater quality.Covering 15 farms,Jansanjiang area has good groundwater quality in general.The groundwater quality grades in the study area are characterized by concentrated areal distribution,as observing the farms with poorer groundwater quality located in the middle-east part and those with better groundwater quality located in the east and west parts.The pattern of regional distribution of groundwater quality in Jiansanjiang area is mainly related to fertilizer application.While ensuring food and water securities,managing fertilizer application should be considered and new measures adopted for high crop yield.

国家重点研发计划（2017YFC0406002）；国家自然科学基金（51579044;41071053;51479032）；黑龙江省自然科学基金（2017007）;黑龙江省水利科技项目（201319;201501;201503）