基于人工神经网络的输变电土地复垦应用探讨

Discussion on land reclamation in power transmission and transformation based on BP neural network

针对传统的土地复垦适宜性评价方法受人为主观影响较大的问题,该文通过分析神经网络结构、计算流程及误差传递规律,建立了基于人工神经网络的输变电土地复垦适宜性评价模型;然后根据土壤等级7个评价参数的特点确定输入层神经元数为7,根据土地复垦成4类的特点确定输出层神经元数为4,根据经验公式确定隐含层神经元数为9;配置成7-9-4的网络结构,采用Levenberg-Marquardt快速学习算法对网络进行训练、测试及矫正;最后采用输变电线路具有代表性实测样本对该算法进行验证。实验表明经过神经网络识别的土壤等级与实际综合评估土壤等级相符,说明了BP神经网络在输变电土地复垦项目中的可行性和实用性。

This paper analyzed the neural network structure,process flow,error transmission and built the land reclamation suitability evaluation model based on BP neural network.The input layer was configured 7according to soil parameters,the output layer was configured 4according to soil degrees and the hidden layer was configured 9according to experience.The network was trained,tested and validated by using Levenberg-Marquardt algorithm and configured 7-9-4network structure.Experiments showed that the soil degree recognized by BP neural network was equated with the actual soil degree.It proved the feasibility of BP neural network in Land Reclamation of Transmission and Transformation.