神经网络模型在脱氮预报中的应用

Application of Neural Network to Predicting Nitrogen Content

讨论了人工神经网络方法在含氮量预报上的应用策略 ,并建立了一个 6 - 7- 1结构的三层 BP网络模型 ,进而分析了 BP网络模型在实际应用中存在的问题 ,对 BP网络算法进行了改进 ,在基于改进的神经网络算法基础上 ,使用 C语言实现了程序设计 ,采用收集的 6 7组实验数据进行了离线学习 ,完成了对网络的训练 ,并用训练好的网络模型对 1 2组样本进行测试 ,预测值误差在± 1 0× 1 0 - 6范围内时命中率为 74% .

Artificial neural network method is applied to predict the nitrogen content. A three-layer BP(Background Propagation) neural network model is set up. Its limitations and problems are analyzed in the practical application and the BP network algorithm is improved. Based upon the improved algorithm, a program is designed with C language. On the basis of the 67 collected data, the model's training is completed by off-line studying. Then 12 samples are tested, and percentage of hits is 74% when error of predicting value range is from -10×10\+\{-6\} to 10×10\+\{-6\}.