[1]夏晓靖,高尚,陈虹丽.基于神经网络的燃气日负荷智能预测方法[J].应用科技,2020,47(2):23-28.[doi:10.11991/yykj.201904019]
 XIA Xiaojing,GAO Shang,CHEN Hongli.Research on combination forecasting method of gas daily load based on neural network[J].Applied science and technology,2020,47(2):23-28.[doi:10.11991/yykj.201904019]
点击复制

基于神经网络的燃气日负荷智能预测方法(/HTML)
分享到:

《应用科技》[ISSN:1009-671X/CN:23-1191/U]

卷:
第47卷
期数:
2020年2期
页码:
23-28
栏目:
智能科学与技术
出版日期:
2020-03-05

文章信息/Info

Title:
Research on combination forecasting method of gas daily load based on neural network
作者:
夏晓靖1 高尚2 陈虹丽2
1. 上海机电工程研究所,上海 201109;
2. 哈尔滨工程大学 自动化学院,黑龙江 哈尔滨 150001
Author(s):
XIA Xiaojing1 GAO Shang2 CHEN Hongli2
1. Shanghai Electro-Mechanical Engineering Institute, Shanghai 201109, China;
2. College of Automation, Harbin Engineering University, Harbin 150001, China
关键词:
城市燃气日负荷随机性不确定性广义回归神经网络灰色理论梯度神经网络法广义动态模糊神经网络组合预测
Keywords:
urban gas daily loadrandomnessuncertaintyGRNNgrey theorygradient neural networkGD-FNNcombination forecast
分类号:
TE01
DOI:
10.11991/yykj.201904019
文献标志码:
A
摘要:
为解决城市燃气日负荷具有随机性和不确定性问题,首先分别采用广义回归神经网络(GRNN)、灰色-GRNN和梯度-GRNN对燃气日负荷数据进行预测,通过MATLAB仿真表明,得到的预测误差大部分都在20%以内,说明这3种预测模型总体上是可行的,但预测精度并不是很高;然后针对城市燃气日负荷可能发生的预测模型故障现象,将GRNN、灰色-GRNN和梯度-GRNN的预测数据作为组合预测模型的数据基础,采用广义动态模糊神经网络(GD-FNN)进行智能组合预测,仿真结果表明:GD-FNN的平均预测精度为93.639%,平均每组预测用时为7.668 s,从预测精度上看,组合预测模型的预测精度要明显高于单一预测模型的预测精度,尤其是在预测过程中发生故障现象时更显其优势。
Abstract:
In order to solve the problem of randomness and uncertainty of urban gas daily load, firstly, the data of urban gas daily load are forecasted by using generalized regression neural network (GRNN), grey-GRNN and gradient-GRNN respectively. The simulation results of MATLAB show that most of the forecasting errors are within 20%, which shows that the three forecasting models are generally feasible, but the forecasting accuracy is not very high, Then, aiming at the fault phenomenon of forecast model of urban gas daily gas load, which is likely to occur, the forecast data of GRNN, grey-GRNN and gradient-GRNN are taken as the data base of combination forecast model. Intelligent combination forecast is carried out by using the generalized dynamic fuzzy neural network (GD-FNN). The simulation results show that the average prediction accuracy of GD-FNN is 92.725%, and the average prediction time of each group is 7.668 seconds. The prediction accuracy of the combined forecasting model is much higher than that of the single forecasting model, especially when fault occurs in the forecasting process.

参考文献/References:

[1] 贾承造, 张永峰, 赵霞. 中国天然气工业发展前景与挑战[J]. 天然气工业, 2014, 34(2): 1–11
[2] 马新华. 天然气与能源革命—以川渝地区为例[J]. 天然气工业, 2017, 37(1): 1–8
[3] 郝迎鹏, 孙春良, 赵忠德, 等. 天然气市场短期负荷预测技术特点与发展方向探讨[J]. 城市燃气, 2015(6): 14–19
[4] 魏黎明, 李晓龙, 赵征, 等. 神经网络在吸气式高超声速导弹突防弹道设计上的应用[J]. 空天防御, 2018, 1(2): 14–17
[5] 陈芳, 楼文高. 基于广义回归神经网络的蔬菜市场日价格预测[J]. 浙江农业学报, 2015, 25(7): 1253–1258
[6] 王晓梅. 神经网络导论[M]. 北京: 科学出版社, 2017.
[7] 杨祺煊, 张广楠, 荀烨. 广义回归神经网络在装甲部队油料消耗预测中的应用[J]. 军事交通学院学报, 2016, 18(6): 82–85, 89
[8] 胡浩, 闫伟, 李泓明. 基于组合预测方法的城市道路短时交通流预测[J]. 工业工程与管理, 2019, 24(3): 107–115
[9] BENNETT C J, STEWART R A, LU Junwei. Forecasting low voltage distribution network demand profiles using a pattern recognition based expert system[J]. Energy, 2014, 67: 200–212.
[10] ULMKE M, ERDINC O, WILLETT P. Gaussian mixture cardinalized PHD filter for ground moving target tracking[C]//Proceedings of 2007 10th International Conference on Information Fusion. Quebec, Canada, 2007: 1–8.
[11] 伍世虔, 徐军. 动态模糊神经网络[M]. 北京: 清华大学出版社, 2008.
[12] 叶淑霞, 郭裕顺, 沈建国. 基于规则自动生成的模糊神经网络建模[J]. 杭州电子科技大学学报, 2008, 28(2): 9–12
[13] 沈丹. 船舶纵向运动姿态预测及控制[D]. 哈尔滨: 哈尔滨工程大学, 2016: 39-46.
[14] KALAMPALIKAS N G, PILAVACHI P A. A model for the development of a power production system in Greece, Part II: where RES meet EU targets[J]. Energy policy, 2010, 38(11): 6514–6528.

备注/Memo

备注/Memo:
收稿日期:2019-04-26。
基金项目:国家自然科学基金项目(61371172);黑龙江省自然科学基金项目(F2017008);黑龙江省博士后科研启动基金项目(LBH-Q10140)
作者简介:夏晓靖,女,高级工程师;高尚,男,硕士研究生
通讯作者:高尚,E-mail:chenli401401@163.com
更新日期/Last Update: 2020-04-21