Machine learning-based prediction of coalbed methane composition and real-time optimization of liquefaction process

doi:10.16085/j.issn.1000-6613.2022-2119

Abstract

Abstract:

The skid-mounted liquefaction device is a promising way to solve the utilization problem of small, remote, and distributed coalbed methane (CBM). The CBM properties change with time, which brings a challenge to the optimal operation of liquefaction process. The research on the prediction of CBM component flowrate can provide CBM parameters required for optimization in time and make real-time optimization possible. Based on the idea of process simulation and soft measurement, the CBM liquefaction process was simulated and the CBM component flowrate prediction was carried out. A real-time optimization method for the mixed refrigerant liquefaction process was established, and three gas sources generated randomly were analyzed. The results showed that the data set obtained by the process simulation had good consistency and reliability. The parameter tuning of random forest showed that the model can obtain the optimal or near optimal accuracy when the number of decision trees was between 20 and 40. If this parameter continued to increase, the accuracy improvement was limited. The method based on prediction-optimization could obtain near-optimal operating parameters for the CBM liquefaction process, which was of great significance to the real-time optimization of industrial production.

Key words: computer simulation, process systems, coalbed methane, particle swarm, random forest

摘要：

撬装式液化装置是解决小、偏、散煤层气利用难题的途径之一。煤层气性质不断变化，给液化过程的最优操作带来挑战。研究煤层气组分流量的预测可及时提供优化所需的煤层气参数，使实时优化成为可能。本文基于流程模拟和软测量思想，开展煤层气液化过程模拟及煤层气组分流量预测的研究，建立一种混合制冷剂液化过程实时优化方法，并对随机生成的三组气源数据进行案例分析。结果表明：基于流程模拟方法获得的数据集具有较好的一致性及可信度；针对随机森林的调参显示，决策树数量处于20~40时，模型能获得最优或者接近最优的准确度，继续增大这一参数，提升的准确度极为有限；基于预测-优化的方法能够获得近似最优的运行参数，对实际生产的实时调优具有重要意义。

关键词: 计算机模拟, 过程系统, 煤层气, 粒子群, 随机森林

CLC Number:

TQ021.8

ZENG Siying, YANG Minbo, FENG Xiao. Machine learning-based prediction of coalbed methane composition and real-time optimization of liquefaction process[J]. Chemical Industry and Engineering Progress, 2023, 42(10): 5059-5066.

曾思颖, 杨敏博, 冯霄. 基于机器学习的煤层气组成预测及液化过程的实时优化[J]. 化工进展, 2023, 42(10): 5059-5066.

Figures/Tables 10

References 17

1	GAO Ting, LIN Wensheng, GU Anzhong, et al. Coalbed methane liquefaction adopting a nitrogen expansion process with propane pre-cooling[J]. Applied Energy, 2010, 87(7): 2142-2147.
2	THAKUR Pramod C, LITTLE Harold G, KARIS William G. Global coalbed methane recovery and use[J]. Energy Conversion and Management, 1996, 37(6/7/8): 789-794.
3	BIBLER Carol J, MARSHALL James S, PILCHER Raymond C. Status of worldwide coal mine methane emissions and use[J]. International Journal of Coal Geology, 1998, 35(1/2/3/4): 283-310.
4	陈仕林, 孙兆虎, 宫敬. 煤层气小型撬装液化装置的优化[J]. 油气储运, 2016, 35(5): 542-546.
	CHEN Shilin, SUN Zhaohu, GONG Jing. Optimization of small-scale skid-mounted liquefaction device for coalbed methane[J]. Oil & Gas Storage and Transportation, 2016, 35(5): 542-546.
5	WANG Meiqian, ZHANG Jian, XU Qiang, et al. Thermodynamic-analysis-based energy consumption minimization for natural gas liquefaction[J]. Industrial & Engineering Chemistry Research, 2011, 50(22): 12630-12640.
6	MORIN Alexandre, WAHL Per Eilif, Mona MØLNVIK. Using evolutionary search to optimise the energy consumption for natural gas liquefaction[J]. Chemical Engineering Research and Design, 2011, 89(11): 2428-2441.
7	KHAN Mohd Shariq, LEE Moonyong. Design optimization of single mixed refrigerant natural gas liquefaction process using the particle swarm paradigm with nonlinear constraints[J]. Energy, 2013, 49: 146-155.
8	ASPELUND A, GUNDERSEN T, MYKLEBUST J, et al. An optimization-simulation model for a simple LNG process[J]. Computers & Chemical Engineering, 2010, 34(10): 1606-1617.
9	HWANG Ji-Hyun, Myung-Il ROH, LEE Kyu-Yeul. Determination of the optimal operating conditions of the dual mixed refrigerant cycle for the LNG FPSO topside liquefaction process[J]. Computers & Chemical Engineering, 2013, 49: 25-36.
10	ALABDULKAREM Abdullah, MORTAZAVI Amir, HWANG Yunho, et al. Optimization of propane pre-cooled mixed refrigerant LNG plant[J]. Applied Thermal Engineering, 2011, 31(6/7): 1091-1098.
11	SHAO Pei, WANG Aikuan, WANG Wenfeng. Experimental simulation of biogenic coalbed gas generation from lignite and high-volatile bituminous coals[J]. Fuel, 2018, 219: 111-119.
12	YOUNG G B C. Computer modeling and simulation of coalbed methane resources[J]. International Journal of Coal Geology, 1998, 35(1/2/3/4): 369-379.
13	王安印. 小型撬装式LNG液化装置工艺设计[D]. 成都: 西南石油大学, 2014.
	WANG Anyin. Process design of small skid-mounted LNG liquefaction unit[D]. Chengdu: Southwest Petroleum University, 2014.
14	HE Tianbiao, JU Yonglin. Dynamic simulation of mixed refrigerant process for small-scale LNG plant in skid mount packages[J]. Energy, 2016, 97: 350-358.
15	WANG Mengyu, KHALILPOUR Rajab, ABBAS Ali. Thermodynamic and economic optimization of LNG mixed refrigerant processes[J]. Energy Conversion and Management, 2014, 88: 947-961.
16	Chao OU, LIN Weixing. Comparison between PSO and GA for parameters optimization of PID controller[C]//2006 International Conference on Mechatronics and Automation. Luoyang, China: IEEE, 2006: 2471-2475.
17	郑胜洁，徐余明，胡祖翰，等. 基于随机森林的城市轨道交通桥梁故障预测模型[J]. 北京交通大学学报, 2022, 46(5): 9-18.
	ZHENG Shengjie, XU Yuming, HU Zuhan, et al. Prediction model of urban rail transit bridge failure based on random forest [J]. Journal of Beijing Jiaotong University, 2022, 46(5): 9-18.

组分	物质的量分数
CH₄	0.1563
C₂H₄	0.3747
C₃H₈	0.0401
C₄H₁₀	0.3407
N₂	0.0882

组分	物质的量分数
CH₄	0.1563
C₂H₄	0.3747
C₃H₈	0.0401
C₄H₁₀	0.3407
N₂	0.0882

序号	CH₄流量 /m³·h^-1	C₂H₆流量 /m³·h^-1	C₃H₈流量 /m³·h^-1	C₄H₁₀流量 /m³·h^-1	C₅H₁₂流量 /m³·h^-1	C₆H₁₄流量 /m³·h^-1	N₂流量 /m³·h^-1	M1_N /m³·h^-1	M1_P /kPa	M2_P /kPa	N2_T /℃
1	2056.72	65.85	8.06	4.27	0.94	0.98	19.38	6002.38	237.93	3140.74	-60.07
2	1540.82	29.36	5.78	3.26	0.89	0.47	14.74	7560.17	267.08	2179.2	-60.61
3	1991	64.41	8.86	3.52	0.81	0.63	14	5541.83	230.99	3227.21	-61.13
4	1519.16	49.66	11.92	3.29	1.01	0.98	11.39	7526.84	279.78	2231.12	-60
5	1316.48	59.47	8.76	3.46	0.71	0.74	19.32	6751.24	317.55	2240.62	-61.92
…	…	…	…	…	…	…	…	…	…	…	…

序号	CH₄流量 /m³·h^-1	C₂H₆流量 /m³·h^-1	C₃H₈流量 /m³·h^-1	C₄H₁₀流量 /m³·h^-1	C₅H₁₂流量 /m³·h^-1	C₆H₁₄流量 /m³·h^-1	N₂流量 /m³·h^-1	M1_N /m³·h^-1	M1_P /kPa	M2_P /kPa	N2_T /℃
1	2056.72	65.85	8.06	4.27	0.94	0.98	19.38	6002.38	237.93	3140.74	-60.07
2	1540.82	29.36	5.78	3.26	0.89	0.47	14.74	7560.17	267.08	2179.2	-60.61
3	1991	64.41	8.86	3.52	0.81	0.63	14	5541.83	230.99	3227.21	-61.13
4	1519.16	49.66	11.92	3.29	1.01	0.98	11.39	7526.84	279.78	2231.12	-60
5	1316.48	59.47	8.76	3.46	0.71	0.74	19.32	6751.24	317.55	2240.62	-61.92
…	…	…	…	…	…	…	…	…	…	…	…

特征变量	M1_T	M1_N	M7_V	M8_P	M11_V	M12_P	M13_T	N1_N	N2_T	N4_V	N5_V
M1_T	1.0000	-0.2782	-0.2011	-0.1316	-0.1443	-0.1316	0.1161	-0.0855	-0.2275	-0.0857	-0.0427
M1_N	-0.2782	1.0000	-0.3583	0.1582	0.9336	0.1582	-0.7588	-0.3147	0.2973	-0.3128	-0.0762
M7_V	-0.2011	-0.3583	1.0000	-0.8485	-0.6690	-0.8485	0.8443	0.9883	0.1641	0.9881	-0.4216
M8_P	-0.1316	0.1582	-0.8485	1.0000	0.4517	1.0000	-0.7050	-0.9032	-0.0658	-0.9038	0.5536
M11_V	-0.1443	0.9336	-0.6690	0.4517	1.0000	0.4517	-0.9277	-0.6298	0.1758	-0.6282	0.1012
M12_P	-0.1316	0.1582	-0.8485	1.0000	0.4517	1.0000	-0.7050	-0.9032	-0.0658	-0.9038	0.5536
M13_T	0.1161	-0.7588	0.8443	-0.7050	-0.9277	-0.7050	1.0000	0.8262	0.0295	0.8252	-0.2274
N1_N	-0.0855	-0.3147	0.9883	-0.9032	-0.6298	-0.9032	0.8262	1.0000	0.1622	1.0000	-0.4785
N2_T	-0.2275	0.2973	0.1641	-0.0658	0.1758	-0.0658	0.0295	0.1622	1.0000	0.1646	-0.0959
N4_V	-0.0857	-0.3128	0.9881	-0.9038	-0.6282	-0.9038	0.8252	1.0000	0.1646	1.0000	-0.4849
N5_V	-0.0427	-0.0762	-0.4216	0.5536	0.1012	0.5536	-0.2274	-0.4785	-0.0959	-0.4849	1.0000