论文部分内容阅读
储粮环境是一个由多种因素构成的复杂系统,粮情状况与环境中的微生物活性、温度、湿度和CO_2浓度等因素密切相关,常规粮情预测方法已经很难满足当代储粮监控的高度和精度。基于BP神经网络的灵活性建立了一种新型粮情监控模型。该模型针对BP神经网络在训练过程中存在的学习速度慢、精度低、易于陷入局部最小等缺点,分别采用动量法对粮情因子的权值进行调整,采用快速动量法对粮情学习效率进行调整,采用L-M算法对粮情监控网络进行综合改进。通过采集储粮环境中的温度、湿度、CO_2浓度等信息,对粮情样本进行训练和预测,并与常规粮情预测方法进行了效果对比。实验结果表明,综合改进后的粮情监控模型应用于粮情预测效果显著,很好的满足了当前粮情的监控需求。
Grain storage environment is a complicated system composed of many factors. Grain condition is closely related to the activity of microorganisms, temperature, humidity and CO 2 concentration in the environment. Conventional grain condition prediction method has been difficult to meet the requirements of contemporary grain storage monitoring And precision. Based on the flexibility of BP neural network, a new grain food monitoring model was established. The model aims at the slow learning speed, low precision and easy to fall into the local minimum in the training process of BP neural network. We use the momentum method to adjust the weight of the grain factor, and use the fast momentum method to study the grain learning efficiency Adjust and adopt LM algorithm to comprehensively improve the monitoring network of grain situation. By collecting the information of temperature, humidity and CO 2 concentration in grain storage environment, the grain food samples were trained and predicted, and compared with the conventional grain estimation methods. The experimental results show that the comprehensively improved grain condition monitoring model applied to the prediction of grain condition has a significant effect and satisfactorily meets the monitoring needs of the current grain condition.