基于LM-BP神经网络的农机总动力预测

Prediction of Total Power in Agriculture Machinery Based on LM-BP Neural Network

利用黑龙江省1983-2013年农机总动力数据,运用标准BP神经网络和LM-BP神经网络对黑龙江省未来5年的农机总动力进行预测。预测结果表明:在达到相同的误差目标值(即计算期望精度),LM-BP神经网络与标准PB相比,具有更快的收敛速度。如果需进一步减小误差目标值(即提高计算期望精度)时,标准BP神经网络在16h内都无法满足给定的精度要求;而LM-BP神经网络在20s内即可满足给定的精度要求。此时,LMBP神经网络的收敛速度优势非常明显,而拟合的精度也进一步提高,表明LM-BP神经网络具有较高的预测精度。准确的预测黑龙江省农机总动力,可为黑龙江省农业机械化发展规划的制定和近阶段农业机械化的发展水平提供参考依据。

Taking advantage of the total power of agriculture machinery data in Heilongjiang Province from 1983 to 2013, using BP neural network and LM-BP network to predict the total power of agriculture machinery in Heilongjiang Province over the next 5 years. The result of prediction show that in the same error target values （ expected accuracy of compute） , compared with BP neural network, LM-BP neural network has faster convergence speed. To reduce the error of target （ improve the expected accuracy of compute） , the BP neural network cannot satisfy the requirement of the given accuracy within 16 hours, while the LM-BP neural network can satisfy the requirement of accuracy in 20 seconds. At this point, the convergence speed of LM-BP neural network is obvious, and the fitting precision is also improved, which show that the LM-BP neural network has higher prediction accuracy. Accurate prediction of the total power of agriculture machiner- y in Heilongjiang Province, which can make the development plan for the agricultural mechanization in Heilongjiang Prov- ince and provide references for the level of agricultural mechanization development.