Bio:
Email: liaohongqu1212@163.com
廖红蕖(1994—),女,硕士,工程师,主要从事生态环保、植物营养施肥方向的研究;liaohongqu1212@163.com
针对上海市崇明滧东地区传统施肥方式不能满足水稻需肥规律的问题,以水稻“3414”肥效试验结果为数据来源,以N、P2O5、K2O施用量为优化目标,建立3-7-1拓扑结构BP神经网络模型,通过遗传算法得到最优产量下的最优施肥配比。预测结果表明:当地较优的施肥配比是N、P2O5、K2O的施用量分别为24.94、0.87、4.27 kg/亩(1亩=667 m2),预计最高产量为531.5 kg/亩。通过验证试验,在较优的施肥配比条件下,水稻的实际产量为548.7 kg/亩,验证了BP神经网络模型预测结果的准确性。
In connection with the problem that traditional fertilization method cannot meet the rule of fertilizer requirement of rice in Yaodong area of Chongming District, Shanghai, a 3-7-1 topological structure BP neural network model is established based on the results of rice "3414" fertilizer efficiency experiment as the data source and the application amounts of N, P2O5 and K2O as the optimization objective. The optimal fertilization ratio under the optimal yield is obtained by genetic algorithm. The predicted results show that the optimal fertilization ratio of N, P2O5 and K2O in local area is 24.94, 0.87 and 4.27 kg/mu (1 mu=667 m2), respectively, and the highest yield is expected to be 531.5 kg/mu. Through validation experiment, the actual yield of rice is 548.7 kg/mu under the optimal fertilization ratio, which verifies the accuracy of the predicted results of the BP neural network model.