基于传播模型与神经网络的输变电工程造价分析与预测方法研究

doi:10.3969/j.issn.2097-0706.2025.04.003

摘要/Abstract

摘要：

在现代电力系统中，准确预测输变电工程的造价对项目规划和实施至关重要。传统预测方法在处理时间序列和结构分析等定量预测问题时存在精度低和自适应能力差的问题。为了改进预测精度，提出了一种基于易感者-感染者-治愈者（SIR）传染病模型和神经网络的输变电工程造价预测方法。该方法利用SIR模型对可变费用进行动态建模，并通过非线性最小二乘法拟合模型参数。将历史数据和模型参数输入前馈神经网络（FNN），通过训练和计算得到预测结果。最终，采用贝叶斯优化算法（BOA）对FNN的超参数进行优化，完成BOA-FNN模型训练。研究结果表明，该预测方法的平均绝对百分比误差（MAPE）低至0.430 7%，稳定可靠地提高了预测精度。

关键词: 输变电工程, 工程造价, 传染病模型, SIR模型, 前馈神经网络, 贝叶斯优化算法, 工程投资预测

Abstract:

Accurate prediction on the costs of transmission and substation projects is crucial for the planning and implementation of modern power systems. Traditional prediction methods often suffer from low accuracy and poor adaptability while handling quantitative prediction problems such as time series and structural analyses. To improve prediction accuracy， a cost prediction method for transmission and substation projects was proposed based on the Susceptible-Infected-Removed （SIR） epidemic model and neural networks. This method utilized the SIR model for dynamic modeling of variable costs， and fitted the model parameters with nonlinear least squares. Historical data and model parameters were then input into a Feedforward Neural Network（FNN）， and predictions were obtained through training and computation. Finally， Bayesian optimization algorithm （BOA） was employed to optimize the hyperparameters of the FNN， completing the BOA-FNN model training. The study results indicated that this prediction method achieved a mean absolute percentage error （MAPE） as low as 0.430 7%， significantly enhancing prediction accuracy with stability and reliability.

Key words: transmission and transformation engineering, project cost, epidemic model, SIR model, feedforward neural network （FNN）, Bayesian optimization algorithm（BOA）, project investment prediction

中图分类号:

TK011：TM926

陆汉东, 方明, 刘刚刚, 周妍. 基于传播模型与神经网络的输变电工程造价分析与预测方法研究[J]. 综合智慧能源, 2025, 47(4): 33-40.

LU Handong, FANG Ming, LIU Ganggang, ZHOU Yan. Research on cost analysis and prediction methods for power transmission and transformation projects based on propagation models and neural networks[J]. Integrated Intelligent Energy, 2025, 47(4): 33-40.

图/表 7

图1

表1

输变电工程造价历史数据样本

项目序号	动态总投资/ 万元	固定费用/万元	可变费用/万元	$β$	$γ$
1	6 661.00	3 730.16	2 930.84	0.51	0.60
2	4 016.74	2 209.21	1 807.53	0.48	0.57
3	4 777.53	2 627.64	2 149.89	0.51	0.60
4	8 282.74	4 721.16	3 561.58	0.51	0.60
5	5 647.46	3 106.10	2 541.36	0.51	0.60
6	7 325.89	4 102.50	3 223.39	0.51	0.60
7	5 823.97	3 203.18	2 620.79	0.51	0.60
8	7 773.03	4 430.63	3 342.40	0.51	0.60
9	7 001.40	3 920.78	3 080.62	0.51	0.60
10	7 011.41	3 926.39	3 085.02	0.51	0.60
11	7 735.36	4 409.15	3 326.20	0.51	0.60
12	7 361.03	4 122.18	3 238.85	0.51	0.60
13	8 181.84	4 663.65	3 518.19	0.51	0.60
14	5 479.16	3 013.54	2 465.62	0.51	0.60
15	7 786.15	4 438.10	3 348.04	0.51	0.60
16	5 681.04	3 124.57	2 556.47	0.51	0.60
17	9 613.33	5 575.73	4 037.60	0.51	0.60
18	3 568.85	1 962.87	1 605.98	0.48	0.57
19	3 623.77	1 993.07	1 630.70	0.48	0.57
20	7 386.62	4 136.51	3 250.11	0.51	0.60
21	5 560.26	3 058.14	2 502.12	0.51	0.60
22	7 789.83	4 440.20	3 349.63	0.51	0.60
23	8 241.64	4 697.73	3 543.91	0.51	0.60
24	6 017.76	3 369.95	2 647.81	0.51	0.60

表1

图2

表2

图3

图4

表3

参考文献 27

[1]	阎涛, 贺胜利. 造价精益化管理在火电基建工程中的落实[J]. 华电技术, 2019, 41(4): 76-80.
	YAN Tao, HE Shengli. Implementation of lean cost management in thermal power infrastructure projects[J]. Huadian Technology, 2019, 41(4): 76-80.
[2]	沈晨姝. 广东电网输变电工程造价指数体系构建与计算方法研究[D]. 北京: 华北电力大学, 2019.
	SHEN Chenshu. Research on the construction and calculation method of transmission and transformation project cost index system of guangdong power Grid[D]. Beijing: North China Electric Power University, 2019.
[3]	LIN T Y, YI T, ZHANG C, et al. Intelligent prediction of the construction cost of substation projects using support vector machine optimized by particle swarm optimization[J]. Mathematical Problems in Engineering, 2019, 2019(1): 1-10.
[4]	MA L, HE L F, ZHANG X F, et al. Research on substation project prediction method in power transmission and transformation by improved neural network intelligent model[J]. Journal of Physics: Conference Series, 2021, 1952(3):032052.
[5]	WANG S S, QIAO W X, WANG L, et al. Cost management of power grid transmission and substation project based on BIM technology[J]. Applied Mathematics and Nonlinear Sciences, 2023, 8(2): 1433-1446.
[6]	中国人民银行广州分行课题组. 基于复杂网络传染病模型的金融风险防控研究[J]. 南方金融, 2021(7):29-39.
	Research Group of Guangzhou Branch of People's Bank of China. Research on financial risk prevention and control based on complex network epidemic model[J]. South China Finance, 2021(7):29-39.
[7]	杨以恒. SIR传染病模型在网络信息传播中的应用[J]. 计算机时代, 2023(11):68-70.
	YANG Yiheng. Application of the SIR epidemic model in network information dissemination[J]. Computer Era, 2023(11): 68-70.
[8]	ZHONG X, PANG B, DENF F, et al. Hybrid stochastic control strategy by two-layer networks for dissipating urban traffic congestion[J]. Science China Information Sciences, 2024, 67(4): 140204.
[9]	白世贞, 吴文娅, 鄢章华, 等. 基于马尔可夫与传染病模型的供应链网络中断风险传播研究[J]. 管理工程学报, 2023, 37(4):206-221.
	BAI Shizhen, WU Wenya, YAN Zhanghua, et al. Research on the propagation of supply chain network disruption risk based on Markov and epidemic models[J]. Journal of Management Engineering, 2023, 37(4): 206-221.
[10]	程乐峰, 余涛, 张孝顺, 等. 机器学习在能源与电力系统领域的应用和展望[J]. 电力系统自动化, 2019, 43(1): 15-43.
	CHENG Lefeng, YU Tao, ZHANG Xiaoshun, et al. Machine learning for energy and electric power systems: State of the art and prospects[J]. Automation of Electric Power Systems, 2019, 43(1): 15-43.
[11]	余涛, 程乐峰, 张孝顺. 基于信息-物理-社会系统融合和群体机器学习的弱中心化微元网: 理论研究与关键科学问题分析[J]. 中国科学:技术科学, 2019, 49(12): 1541-1569.
	YU Tao, CHENG Lefeng, ZHANG Xiaoshun. The weakly-centralized web-of-cells based on cyber-physical-social systems integration and group machine learning: Theoretical investigations and key scientific issues analysis[J]. SCIENTIA SINICA Technologica, 2019, 49(12): 1541-1569.
[12]	安磊, 张洁, 齐霞, 等. 基于随机森林输变电线路工程造价估算研究[J]. 控制工程, 2016, 23(11):1841-1844.
	AN Lei, ZHANG Jie, QI Xia, et al. Research on cost estimation of transmission and transformation line engineering based on random forest[J]. Control Engineering, 2016, 23(11):1841-1844.
[13]	彭云, 叶文昊. 基于神经网络模型的输变电工程投资计划决策研究[J]. 中国勘察设计, 2023(S1):82-85.
	PENG Yun, YE Wenhao. Research on investment plan decision-making of transmission and transformation engineering based on neural network model[J]. China Survey and Design, 2023(S1):82-85.
[14]	李靖飞, 袁敏. 基于SVM模型的输变电工程概预算预测研究[J]. 电工电气, 2020,(8):69-70.
	LI Jingfei, YUAN Min. Research on budget prediction of transmission and transformation engineering based on SVM Model[J]. Electrical Engineering, 2020(8):69-70.
[15]	李迎. 基于神经网络的输变电工程造价研究[J]. 电气技术与经济, 2023 (6):40-42.
	LI Ying. Research on cost of transmission and transformation engineering based on neural network[J]. Electrical Technology and Economy, 2023(6):40-42.
[16]	张恒武, 吴小忠, 沈晓隶, 等. 基于BP神经网络优化算法的输变电工程造价预测模型[J]. 沈阳工业大学学报, 2023, 45(4):381-386. doi: 10.7688/j.issn.1000-1646.2023.04.05
	ZHANG Hengwu, WU Xiaoping, SHEN Xiaoli, et al. Transmission and transformation project cost prediction model based on BP neural network optimization algorithm[J]. Journal of Shenyang University of Technology, 2023, 45(4):381-386. doi: 10.7688/j.issn.1000-1646.2023.04.05
[17]	KERMACK W, MCKENDRICK A. A contribution to the mathematical theory of epidemics[J]. Proceedings of the Royal Society of London:Series A, 1927, 115(772): 700-721.
[18]	WANG X, WANG Z, SHEN H. Dynamical analysis of a discrete-time SIS epidemic model on complex networks[J]. Applied Mathematics Letters, 2019, 94: 292-299.
[19]	TIBERIU H, FRANCISCO S M L, Mark M K. Exact analytical solutions of the susceptible infected recovered epidemic model and of the SIR model with equal death and birth rates[J]. Applied Mathematics and Computation, 2014, 236: 184-194.
[20]	ZHAO Y, JIANG D. The threshold of a stochastic SIRS epidemic model with saturated incidence[J]. Applied Mathematics Letters, 2014, 34: 90-93.
[21]	XIA L L, JIANG G P, SONG B, et al. Rumor spreading model considering hesitating mechanism in complex social networks[J]. Physica A: Statistical Mechanics and its Applications, 2015, 437: 295-303.
[22]	LIU G Y, ZHANG J Z, ZHONG X J, et al. Hybrid optimal control for malware propagation in UAV-WSN System: A stacking ensemble learning control algorithm[J]. IEEE Internet of Things Journal, 2024, 34:1110-1113.
[23]	CHAI R, TSOURDOS A, SAVVARIS A, et al. Design and implementation of deep neural network-based control for automatic parking maneuver process[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022, 33:1400-1413.
[24]	GOODFELLOW I, BENGIO Y, COURVILLE A. Deep Learning[M]. Cambridge, Massachusetts, USA: MIT Press, 2016.
[25]	JASPER S, HUGO L, RYAN P A. Practical Bayesian optimization of machine learning algorithms[C]// Annual Conference on Neural Information Processing Systems, 2012:2951-2959.
[26]	YANG Y Q, LIU G Y, LIANG Z W, et al. Hybrid control for malware propagation in rechargeable WUSN and WASN: From knowledge-driven to data-driven[J]. Chaos, Solitons & Fractals, 2023, 113703.
[27]	武小琳, 栾凌, 潘连武, 等. 基于LM-CNN的输变电工程造价自动计算模型[J]. 中国电力, 2023, 56(2):157-163. doi: 10.11930/j.issn.1004-9649.202103063
	WU Xiaolin, LUAN Ling, PAN Lianwu, et al. Automatic cost calculation model for power transmission and transformation projects based on LM-CNN[J]. Electric Power, 2023, 56(2): 157-163. doi: 10.11930/j.issn.1004-9649.202103063

评估次数	第1层隐藏层大小	第2层隐藏层大小	学习率	RMSE
1	142	43	0.000 2	177.62
2	104	146	0.051 6	142.09
3	73	10	0.085 6	190.65
4	46	50	0.000 2	267.54
5	121	81	0.099 5	53.17
6	130	14	0.099 9	158.65
7	150	94	0.098 9	366.07
8	118	145	0.049 3	83.76
9	116	105	0.099 7	671.31
10	19	141	0.040 3	28.00
11	25	137	0.047 8	110.37
12	86	75	0.091 7	614.90
13	25	146	0.039 7	81.24
14	10	144	0.000 2	294.34
15	11	143	0.067 7	80.85
16	133	78	0.096 7	81.40
17	17	142	0.011 1	99.20
18	40	142	0.095 1	155.07
19	128	81	0.001 1	144.03
20	127	150	0.000 4	163.19
21	120	138	0.000 8	141.39
22	141	142	0.097 0	2 945.59
23	123	53	0.066 6	44.89
24	10	105	0.096 1	90.08
25	23	71	0.093 5	31.98
26	11	52	0.094 1	75.72
27	24	29	0.041 8	86.07
28	10	10	0.000 4	417.95
29	36	88	0.059 9	81.64
30	52	109	0.059 3	99.11
31	58	150	0.000 2	175.15
32	32	52	0.000 2	318.94
33	111	11	0.098 8	180.85
34	90	150	0.000 2	139.16
35	150	63	0.094 3	345.21
36	20	99	0.098 6	152.94
37	133	53	0.051 7	194.09
38	115	150	0.000 1	179.85
39	62	68	0.095 2	387.44
40	114	53	0.082 9	283.39
41	123	16	0.000 1	315.45
42	74	150	0.069 8	328.53
43	43	108	0.000 1	269.39
44	144	10	0.026 3	107.01
45	94	10	0.000 7	181.98
46	56	10	0.080 6	33.39
47	44	10	0.059 2	122.36
48	60	10	0.000 1	447.62
49	37	49	0.099 4	305.43
50	51	11	0.093 4	219.58

样本序号	实际可变费用/万元	预测可变费用/万元		相对误差/%
样本序号	实际可变费用/万元	BOA-FNN	BP	BOA-FNN	BP
1	1 630.70	1 630.27	1 540.93	0.026 5	5.504 8
2	3 250.11	3 224.50	3 153.09	0.788 0	2.985 2
3	2 502.12	2 488.92	2 426.59	0.527 5	3.018 3
4	3 349.63	3 366.06	3 148.88	0.490 5	5.993 2
5	3 543.91	3 545.90	3 256.61	0.056 1	8.106 9
6	2 647.81	2 666.22	2 597.97	0.695 4	1.882 3
MAPE				0.430 7	4.581 8