Abstract: Pulse Coupled Neural Network(PCNN), with the basic characteristics of coupling and pulse output, is widely implemented on image processing. PCNN model was improved as follows: the value of link load part equaled to the pulse in the last ignition action in order to reflect directly the relationship between the before and after neural cell, dynamic threshold equaled to the classification range of water resource evaluation criterion to classify the samples easily and unnecessary parameter was omitted to reduce the complexity of PCNN model. The improved PCNN model was used to evaluate agricultural water resources supply and demand in Sanjiang Plain and obtained the better results. Results show that PCNN model is feasible for evaluating agricultural water resource utilization, expanding the application areas of PCNN model and providing a new way for water resource evaluation.