资源科学 ›› 2017, Vol. 39 ›› Issue (6): 1048-1058.doi: 10.18402/resci.2017.06.06

• • 上一篇    下一篇

基于FI-GA-NN融合的区域能源安全外生警源分级预警研究

胡健(), 孙金花   

  1. 重庆理工大学管理学院,重庆 400054
  • 收稿日期:2016-09-12 修回日期:2017-03-06 出版日期:2017-06-20 发布日期:2017-06-20
  • 作者简介:

    作者简介:胡健,男,黑龙江望奎县人,博士,副教授,硕士生导师,研究方向为能源安全。E-mail:jianhu-hit@163.com

  • 基金资助:
    国家自然科学基金项目(71301181);重庆市社会科学规划项目(2015YBSH051);重庆市教委人文社科项目(15SKG134)

Early classification warning for regional energy security exogenous sources based on FI-GA-NN Model

Jian HU(), Jinhua SUN   

  1. School of Management,Chongqing University of Technology,Chongqing 400054,China
  • Received:2016-09-12 Revised:2017-03-06 Online:2017-06-20 Published:2017-06-20

摘要:

区域能源安全预警研究对于解决中国现阶段区域能源安全突发事件频现问题,保障区域经济与区域安全协调发展具有重要现实意义。本文以区域能源安全外生警源为研究对象,通过对区域能源安全事件案例收集及整理,构建了能源安全外生警源预警指标和数据集。融合模糊积分(Fuzzy Integral)、遗传算法(Genetic Algorithm)和神经网络(Neural Network)等方法的基本原理,设计了区域能源安全外生警源分级预警的FI-GA-NN模型,该模型首先利用模糊积分方法评估出区域能源安全外生警源样本分级预警的期望值,然后利用训练样本对遗传神经网络进行训练,最后对外生警源测试样本进行分级预警。实验测试结果表明:①利用FI-GA-NN模型对外生警源训练样本(1999-2006年)进行拟合训练,模型收敛速度快,训练到第717步时,模型误差平方和小于期望值。经过大约60代的搜索后模型的拟合度趋于稳定,模型训练的实际输出值与期望输出值较接近;②利用FI-GA-NN模型对能源安全外生警源测试样本(2007-2015年)进行分级预警,预警准确率较高,能有效提高区域能源安全外生警源预警的正确性,降低预警风险,模型体现出了较强的应用价值。

关键词: 区域能源安全, 外生警源, 分级预警, 模糊积分, 遗传算法, 神经网络

Abstract:

In order to solve the problem of frequent occurrence of energy security incidents in China,we took the regional energy security exogenous source as focal research to develop effective early warning. We defined the concept of regional energy security exogenous source. From an external point of view we divided the root cause of regional energy events into three kinds of exogenous sources using analysis of typical regional energy security incidents,including energy price fluctuations,energy policy intervention and external environmental changes. Then,the attribute set and data set of regional energy security exogenous sources are constructed by extracting different cases. At the premise of considering exogenous source characteristics,such as suddenness nonlinear and complex,we designed a classification early warning model,known as the FI-GA-NN model,an exogenous source warning model with fuzzy integral (FI),genetic algorithm (GA)and neural network (NN). In this model,we have first of all designed the fuzzy integral method to make an exploration of the sample grade of exogenous source. Secondly,we adopted BP neural network methods as the early warning method. At the same time,a genetic algorithm was used to optimize the weights of BP neural network to improve the accuracy of the model. To verify the validity of the FI-GA-NN model,we have selected 30 typical cases of energy security exogenous source to make the classification early warning report during the years from 1999 to 2015. The training results showed that the FI-GA-NN model had fast convergence speed. The model error was less than the expected value when the model was trained to the 717th step. After approximately 60 generations of searching, the actual output value of the model training was closer to the expected output value. The testing results show that the classification warning accuracy rate of FI- GA- NN was higher. The classification early warning results of the regional energy security exogenous source prove to be well in agreement with actual conditions. Therefore, it is reasonable and effective to apply the FI- GA- NN model to the monitoring the regional energy security situations.

Key words: regional energy security, exogenous source, classification early warning, fuzzy integral, genetic algorithm, neural network