基于组合模型的农村环境污染治理成本预测研究  

Prediction of rural environmental pollution control cost based on combination model

作　　者：刘刚 Liu Gang(Linyi Ecological Environment Bureau Yishui County Branch,Linyi Shandong 276400,China)

机构地区：[1]临沂市生态环境局沂水县分局,山东临沂276400

出　　处：《环境与发展》2024年第6期91-97,共7页Environment & Development

摘　　要：针对环境治理过程中存在的资金浪费、成本缺乏合理规划等问题,提出基于组合模型的农村环境污染治理成本预测研究。从资源使用、环境破坏、环境治理三方面确立成本核算体系,优化成本函数;利用前窥式神经网络设定不同训练规则,分析污染治理过程中投入与产出的关系;使用大量随时间变化的数据建立差分自回归平均模型,结合过去值与过去误差,获得变量未来值;设计小波神经网络学习过程,引入尺度与平移因子,计算每个连接层权重,经过反复学习,确保预测误差最小;整合两个模型预测结果,运算单一模型权重,输出最终成本预测值。在实际区域的实验表明,所提组合模型不仅能够准确预测单项指标的治理成本,还能从长期角度出发,提高整体成本预测精度。Aiming at the problems of waste of funds and lack of reasonable cost planning in the process of environmental governance,a study on the prediction of rural environmental pollution governance cost based on combination model is proposed.Establish the cost accounting system and optimize the cost function from three aspects:resource use,environmental damage and environmental governance;The relationship between input and output in the process of pollution control is analyzed by setting different training rules using forward looking neural network;A large number of time-varying data are used to establish a differential autoregressive average model,and the future value of the variable is obtained by combining the past value and the past error;The learning process of wavelet neural network is designed,the scale and translation factors are introduced,and the weight of each connection layer is calculated.After repeated learning,the prediction error is minimized;The first mock exam is based on two models,and the single model weight is calculated to output the final cost prediction value.Experiments in real areas show that the proposed combined model can not only accurately predict the governance cost of single index,but also improve the overall cost prediction accuracy from a long-term perspective.

关 键 词：组合模型 小波神经网络 农村环境污染 治理成本 投入与产出 

分 类 号：X32[环境科学与工程—环境工程]