摘要
本文使用IPPC(intergovernmental panel on climate change,IPPC)方法对物流运输业2005-2019 年消耗的汽油、煤油、柴油等五种主要能源的碳排放量进行测算汇总,通过模糊认知图(fuzzy cognitive graph,FCM)的构建方法选取八个影响因素作为模糊认知图的节点,在搭建好的FCM结构以及数据的基础上,将实数编码遗传算法(real-coded genetic algorithm,RCGA)与神经网络(neural networks,NN)相结合,以NN→RCGA→NN的顺序来进行优化,先使用神经网络训练出FCM的初始权值矩阵,再通过RCGA迭代出局部最优的权值矩阵,然后再由NN进行参数的优化更新,以此循环直到得到最优权值矩阵,最后利用模糊认知图的推理算法对碳排放进行预测.结果表明,基于GA-ANN-FCM算法(genetic algorithm-artifical neural network-fuzzy cog-nitive map,遗传算法-人工神经网络-模糊认知图混合算法)的碳排放量预测结果是有效的,并且相比于人工神经网络和FCM两种模型较好.
Abstract
This paper uses the IPPC(intergovernmental panel on climate change)method to measure and summarize the carbon emissions of five major energy sources,such as gaso-line,kerosene and diesel,consumed in the logistics and transportation industry from 2005 to 2019;Eight influencing factors are selected as nodes of the fuzzy cognitive graph by the construction method of Fuzzy Cognitive Maps;On the basis of the built FCM structure and data,the Real-coded Genetic Algorithm(RCGA)is combined with Neural Networks(NN)and optimized in the order of NN→RCGA→NN.The initial weight matrix of the FCM is trained by the neural network,and then the locally optimal weight matrix is iterated by the RCGA,and then optimized and updated by the NN,until the optimal weight matrix is obtained.Finally,the inference algorithm of Fuzzy Cognitive Maps is used to predict car-bon emissions.The results of carbon emission prediction based on GA-ANN-FCM algorithm are effective and better compared with both artificial neural network and FCM models.
维普
万方数据