TY - JOUR
AU - Li,, Hu
AB - Abstract In order to investigate coalbed physical parameters and gas bearing properties, based on coal core experiments, a nonlinear comprehensive model estimating coal properties and gas content is established. Three-dimensional numerical simulation of the dual laterolog response of a coalbed fracture is carried out, and a dual laterolog fast computation model for predicting fracture porosity is established. The results show that the study area mainly developed lean coal and meagre coal. The coalbed is rich in gas. Ash content, volatile dry ash-free basis, true density and fixed carbon show good correlation. The logging responses of the coalbed have distinct characteristics; however, the responses highly depend on coal composition and gas content. The simulated annealing differential evolution (SADE) neural network is adopted to predict coal composition and gas content. Numerical analysis of the dual laterolog responses of the coal fractures shows that as coal matrix resistivity increases, the dual laterolog apparent resistivity increases accordingly. For a coalbed with low-angle fractures, deep and shallow resistivity show little difference; however, for a coalbed with high-angle fractures, its dual laterolog shows an obvious positive difference. Fracture porosity and dual laterolog apparent conductivity presents a linear relationship. According to coal matrix resistivity distribution and fracture development, a fast computation model to predict fracture porosity is established based on the dual laterolog. Field data show that the predicted coal composition and gas content from the SADE algorithm agree with that of coal core analysis, and the calculated fracture porosity based on the dual laterolog matches the coal core's macroscopic description. coalbed methane, coal composition, gas content, fracture, logging evaluation, coal core experiment, numerical simulation 1. Introduction Coalbed methane, one of the most significant unconventional energy resources, is different from conventional natural gas, and mainly composed of adsorbed gas on the surface of coal matrix particles, accompanied by a small amount of free gas and dissolved gas, and the dominant gas is methane (CH4) (Clarkson and Bustin 1997, Vishal et al2013a, Pashin et al2009). The global coalbed methane resource buried shallower than 2000 m underground is approximately 240 trillion cubic metres, more than twice that of the proved conventional gas reserves (Zou 2011). The world's major coal-producing countries have attached great importance to the development of coalbed methane (Al-Jubori et al2009). Currently, the United States has ten major coalbed methane fields, and annual coalbed methane production stabilizes at around 50 billion cubic metres, accounting for about 8% of the total gas production of the country (Zou 2011). The geological reserves of coalbed methane in China buried shallower than 2000 m underground is approximately 36.81 trillion cubic metres (Zhang 2011). The total amount ranks third after Russia and Canada in the world. Coalbed methane development and utilization in China are still at an early stage. The annual actual utilization of coalbed methane is less than 5 billion cubic metres, which is a huge gap compared with the United States. Therefore, the exploitation of coalbed methane will become one of China's strategic priorities, which is significant to ensure the sustainable development of national natural gas (Qiu 2008). Coalbed methane adsorption capacity, as the key parameter for measuring the coalbed methane production potential, is the basis for high gas production (Faiz et al2007). The coalbed fracture network system (coal cleat) serves as the major migration pathway for coalbed methane, and its development determines coalbed permeability. The system has great significance to the evaluation of coalbed methane recoverability. However, coal is an unconventional reservoir, which serves as both the source as well as the reservoir for methane, displaying intensive heterogeneity (Vishal et al2013b, Zhang et al2010). The complex relationship among logging responses, coal composition, gas content, fracture etc makes coalbed methane logging evaluation more challenging. Kowaiski (1975) evaluated coal parameters with density and acoustic logs. Mullen (1989) evaluated the industrial coalbed composition and the gas content of the coalbed methane reservoir in the northeast of the San Juan Basin with density logs. Ahmed et al (1991) calculated the content of coalbed composition and gas content with ash content, cleat porosity and density logging. Sun (1991) evaluated the content of primary coalbed composition with resistivity and density logs. Li and Chen (1996) presented a method to calculate coalbed gas content from a density log on the basis of coal density and gas content measurement in the laboratory. Gao et al (2003) interpreted and analysed ash content with core calibration logging based on a coal composition model. Bhanja and Srivastava (2008) presented the limitations of gas content evaluation with only density logging, and developed a comprehensive evaluation method using acoustic, volume density, litho-density and Gamma logs. Based on the modern nonlinear method, without consideration of a specific mathematical model between coal parameters, gas content and logging response, a complex relationship between coalbed parameters and logging responses is expressed implicitly by Pan and Huang (1998). They evaluated coalbed methane content by using the BP neural network method and achieved high accuracy. Fu et al (1999) adopted the stepwise regression method to establish a mathematical model of coalbed methane content with density, gamma ray, sonic and resistivity logging responses in the Huaibei Sunan syncline. Lian et al (2007) established the relationship between coalbed methane content and gamma, density and acoustic logging responses with the support vector machine regression method. Yang et al (2010), in the Huainan mining area, quantitatively evaluated the physical parameters of a 16 unit coalbed with grey relational analysis. The outputs of coalbed methane follow the process of ‘desorption–diffusion–seepage’. Fracture development is significant to whether a coalbed could become an effective coalbed methane reservoir, the productivity, as well as the subsequent exploitation. Sibbit (1985) proposed a method to calculate fracture aperture with a dual laterolog. Hoyer (1991) quantitatively calculated fracture porosity of different angles with deep and shallow resistivity. Chen et al (1997) introduced the methods proposed by Sibbit and Hoyer to China. Chen and Wang (2003) evaluated a cleat in the coalbed field of the Tuha basin with deep and shallow resistivity. Zhang and Cai (2009) carried out fine evaluation of the coalbed structure and fracture identification with micro-resistivity scanning image logging. Li et al (2011) determined the fracture aperture and the fracture porosity of a coalbed with a dual laterolog and a density log. 2. General geology of the target area The target area lies in Hancheng city, Shanxi Province, China. The target coalbed in the Hancheng area is the upper Palaeozoic in the northern Weibei uplift of the Ordos basin. The main coalbeds are the Carboniferous Taiyuan Formation and the Permian Shanxi Formation, and consist mostly of lean coal and meagre coal. The Taiyuan Formation consists of interactive marine and terrestrial deposits with a formation thickness of about 50 m. There are three to nine layers that contain coal. The 11 coalbed and 5 coalbed are the main target layers. The Permian Shanxi Formation is a terrestrial deposit with a formation thickness of about 60 m, in which there are one to four coal layers. The main target of the Shanxi Formation is layer 3. During the period of coalbed deposition, the crust was mainly in subsidence, therefore, the lithology is stable with large thickness, and the coalbed has good gas generating and storage capacity. Coal is an unconventional reservoir which serves as both the source as well as a reservoir for methane, displaying intensive heterogeneity. The complex relationship among logging responses, coal composition, gas content, fracture etc, makes coalbed methane logging evaluation more challenging. In view of the target zone, based on coal core calibration logging theory, a multi-information nonlinear log processing method is adopted to evaluate coal composition and gas content with the SADE neural network. Based on numerical simulation of the dual laterolog response of a coalbed fracture, a fast calculation method is adopted to predict coalbed fracture porosity. 3. Analysis of coal core measurements 3.1. Gas content analysis Coalbed composition and gas content of 152 coal cores of layer 3, layer 5, and layer 11 in the study zone were analysed. The isothermal absorption method was used for the measurement of coal core gas content with the IS-100 Adsorption isotherm instrument made by Teera Tek company. The experimental temperature was set to be the same as the reservoir temperature (25–35 °C). The grain grade of coal samples is 60. After being processed by equilibrium water, the absorbate is methane with purity of 99.99%. The average gas content of the three main target layers is 10.86 m3 t-1, as shown in table 1. Table 1. Coal core test results. Coalbed . Gas content Gcad (%) . Moisture content Mad (%) . Ash content Aad (%) . Volatile content Vad (%) . Fixed content FCad (%) . Vitrinite reflectance R0 (%) . 3 Data range 1–19 0.7–1.7 5–35 11–13 45–75 1.85–2.55 Average 9.38 0.93 21.43 11.14 65.71 2.05 5 Data range 1–17 0.1–1.5 5–75 9–27 5–85 1.85–2.55 Average 10.76 0.85 21.1 11.72 65.25 2.16 11 Data range 3–19 0.3–1.5 5–75 7–19 15–85 1.85–2.55 Average 11.52 0.73 26.27 10.78 62.78 2.03 Coalbed . Gas content Gcad (%) . Moisture content Mad (%) . Ash content Aad (%) . Volatile content Vad (%) . Fixed content FCad (%) . Vitrinite reflectance R0 (%) . 3 Data range 1–19 0.7–1.7 5–35 11–13 45–75 1.85–2.55 Average 9.38 0.93 21.43 11.14 65.71 2.05 5 Data range 1–17 0.1–1.5 5–75 9–27 5–85 1.85–2.55 Average 10.76 0.85 21.1 11.72 65.25 2.16 11 Data range 3–19 0.3–1.5 5–75 7–19 15–85 1.85–2.55 Average 11.52 0.73 26.27 10.78 62.78 2.03 Open in new tab Table 1. Coal core test results. Coalbed . Gas content Gcad (%) . Moisture content Mad (%) . Ash content Aad (%) . Volatile content Vad (%) . Fixed content FCad (%) . Vitrinite reflectance R0 (%) . 3 Data range 1–19 0.7–1.7 5–35 11–13 45–75 1.85–2.55 Average 9.38 0.93 21.43 11.14 65.71 2.05 5 Data range 1–17 0.1–1.5 5–75 9–27 5–85 1.85–2.55 Average 10.76 0.85 21.1 11.72 65.25 2.16 11 Data range 3–19 0.3–1.5 5–75 7–19 15–85 1.85–2.55 Average 11.52 0.73 26.27 10.78 62.78 2.03 Coalbed . Gas content Gcad (%) . Moisture content Mad (%) . Ash content Aad (%) . Volatile content Vad (%) . Fixed content FCad (%) . Vitrinite reflectance R0 (%) . 3 Data range 1–19 0.7–1.7 5–35 11–13 45–75 1.85–2.55 Average 9.38 0.93 21.43 11.14 65.71 2.05 5 Data range 1–17 0.1–1.5 5–75 9–27 5–85 1.85–2.55 Average 10.76 0.85 21.1 11.72 65.25 2.16 11 Data range 3–19 0.3–1.5 5–75 7–19 15–85 1.85–2.55 Average 11.52 0.73 26.27 10.78 62.78 2.03 Open in new tab 3.2. Coal component test analysis 3.2.1. Experimental conditions Moisture content is measured by using the nitrogen drying method. Under the temperature of 105–110 °C, dehydrate the coal samples for 48 h, after air drying. The moisture is determined according to the change in mass of the coal sample before and after drying. The air dried coal samples are fully combusted under the high temperature of 815 ± 10 °C. Ash content is determined by the mass of the residue. Put coal samples in a container with air isolation, and heat for 7 min under the high temperature of 900 ± 10 °C. The volatile constituent is determined using the mass change from before and after heating. The fixed carbon content is the rest of the coal sample after the removal of moisture, ash and volatile content. 3.2.2. Coal composition analysis Moisture content and ash content have a certain influence on the absorption ability (Amorino et al2005, Bachu 2007, Clarkson and Bustin 2000). The existence of moisture reduces the absorption surface area of coal to coalbed methane and blocks the channel for coalbed methane seeping into micro pores. Ash in pore spaces affects absorption and reservoir properties. As the moisture and ash content grow, the coalbed methane content reduces. The coal composition analyses show that (table 1) moisture content is mainly distributed from 0.5% to 1.2%, which belongs to ultra-low moisture coal. The ultra-low moisture is beneficial for coalbed methane gathering. Ash content has the main range 0%∼40%, covering all the ash content level. Volatiles and fixed carbon are key indicators of coal metamorphism. The volatile content of the study area is mainly distributed from 8% to 14%. Fixed carbon mainly ranges from 60% to 90%, displaying a high coal metamorphism degree, which is beneficial for the absorption of coalbed methane. 3.3. Coal rank characteristics Vitrinite reflectance and volatile content are the main indicators of coal ranking. As the degree of coalification intensifies, the vitrinite reflectance increases and the volatile yield decreases. A vitrinite reflectance measurement system, made by the LEICA company, is used to measure the reflectance of coal samples. The instrument is mainly composed of an MSP200 micro photometer, a microscope, and an oil immersion lens. The volatile of the dry ash-free base is mainly distributed from 10% to 30%, and the vitrinite reflectance distributed from 1.7 to 2.6, peaking at about 2.0. The core samples are mainly high metamorphic lean and meagre coal, indicating high gas generation ability and ample gas content. 3.4. Correlation analysis of coal parameters Figures 1–3 show good correlations between ash content and fixed carbon, volatile of dry ash-free base (Vdaf) and fixed carbon, and true density (TRD) of coal samples and ash content respectively. Based on the good correlation between true density and ash content, if a calibration relationship between density log and experimental true density is established, the ash content of a coalbed can be predicted from its density log. Figure 1. Open in new tabDownload slide Relationship between ash content and fixed carbon. Figure 1. Open in new tabDownload slide Relationship between ash content and fixed carbon. Figure 2. Open in new tabDownload slide Relationship between FCad and Vdaf. Figure 2. Open in new tabDownload slide Relationship between FCad and Vdaf. Figure 3. Open in new tabDownload slide Relationship between ash content and true density. Figure 3. Open in new tabDownload slide Relationship between ash content and true density. Since the study area mainly contains lean and meagre coal, the distribution ranges of the volatile content and the vitrinite reflectance are narrow. Figure 4 shows that the ash content and volatile content have poor correlation. Vitrinite reflectance correlates poorly to ash content and volatile content, as shown in figures 5 and 6. Coal composition is an important factor affecting coalbed gas content. Especially, ash content and fixed carbon content have good correlation with the adsorption capacity of a coalbed, as shown in figures 7 and 8. When ash content increases, gas content decreases, and when fixed carbon content increases gas content increases. Figure 4. Open in new tabDownload slide Relationship between ash content and volatile content. Figure 4. Open in new tabDownload slide Relationship between ash content and volatile content. Figure 5. Open in new tabDownload slide Relationship between vitrinite reflectance and ash content. Figure 5. Open in new tabDownload slide Relationship between vitrinite reflectance and ash content. Figure 6. Open in new tabDownload slide Relationship between vitrinite reflectance and volatile content. Figure 6. Open in new tabDownload slide Relationship between vitrinite reflectance and volatile content. Figure 7. Open in new tabDownload slide Relationship between gas content and ash content. Figure 7. Open in new tabDownload slide Relationship between gas content and ash content. Figure 8. Open in new tabDownload slide Relationship between gas content and fixed carbon. Figure 8. Open in new tabDownload slide Relationship between gas content and fixed carbon. 4. Correlation between coal core physical parameters and logging information 4.1. Coalbed logging response In general, a coalbed has high resistivity, high acoustic transit time, high neutron, low natural gamma, low photoelectric absorption cross section and low density response characteristics, and a coalbed is classified on the basis of these characteristics. In the study area, the natural gamma value is among 20–100 API, the density is distributed over 1.2–1.6 g cm-3, the acoustic transit time is 350–500 µs m–1, the neutron (CNL) is 30% to 60%, and the resistivity is 100–20 000 Ω m, as shown in figure 9. It shows that part of the coalbed layers have high natural gamma, low resistivity and a high degree of overlap with sandstone and limestone. Thus, the low gamma and high resistivity cannot meet the demand of classifying a coalbed, and can only be references. Characteristics of low density, high neutron and high acoustic transit time can be used to classify a coalbed. From the good logging response characteristics, no obvious difference is shown among the three main coalbed layers. Figure 9. Open in new tabDownload slide Logging response characteristics of the coalbed in the study area. Figure 9. Open in new tabDownload slide Logging response characteristics of the coalbed in the study area. 4.2. Correlation between coalbed physical properties and logging responses According to the correlation analysis of coal composition, ash component and fixed carbon show good correlation. Because of the low moisture content of the target area (<1%), the volatile content can be determined by Aad + Vad + FCad = 1. Therefore, once the relationship between ash content and logging response is established, other coal parameters can be estimated by the ash content. As shown in figure 10, ash content has a certain correlation with natural gamma, acoustic time, density and neutrons. However, the correlation is not an obvious single linear relationship, indicating that a more sophisticated model is needed. Figure 10. Open in new tabDownload slide Correlation between logging response and ash content. Figure 10. Open in new tabDownload slide Correlation between logging response and ash content. 5. Coalbed methane evaluation based on the SADE algorithm neural network 5.1. Neural network training with the SADE algorithm Every log by and large responds to coal parameters such as ash content, moisture content, fixed carbon content and gas content. This can be observed from figure 10. An individual log is not enough to make any authentic estimate of these parameters. It is fairly good to use the integrated log information to evaluate the coalbed (Bhanja and Srivastava 2008). The self-adaptive differential evolutionary (SADE) algorithm is used to solve nonlinear optimum problems with many local optima (Huang et al2006, Mohseni M 2011, Singh et al2012). SADE is proposed by combining simulated annealing with a difference evolution algorithm, which has good diversity at an early stage and good convergence at a late stage. It does not have the prematurity shortcoming of the classical artificial neural network algorithm and has favourable general search ability and robustness (Li et al 2012). Based on the SADE algorithm, the neural network is used to establish the prediction model of coal parameters and gas content. With the neural network of a single hidden layer, the output can be expressed as (Li et al 2012): 1 where outputk are coal parameters and gas content; k is the output number; inputi are the input parameters; m and n are the node number and the node number of the hidden layer, respectively; f stands for excitation function; w and v are weights. Take the weights of absolute error and relative error as the new error function: 2 where O1, O2 stand for weight coefficient; EP and Er stand for absolute error and relative error function respectively. 3 4 where is the expected output. Taking the weights w and v as independent variables, the training process of the neural network can be converted to the minimization of an unconstrained problem. To solve this problem, the temperature is used to improve the control of the solving process. When the temperature is higher than the critical temperature, it is in the initial optimization period, adopting Scheme DE1 (equation (5)) to increase the population diversity. When the temperature reduces to below the critical temperature, as in the later optimization period, then use Scheme DE2 (equation (6)), where 5 6 where NP is the population number, 5–10 times the unknown parameters, xbest is the best of the current population, F is a scaling factor which is generally 0.5–1, λ is the control parameter. With the temperature controlling strategy, a poor solution is accepted at a certain probability at the initial period; at a later period, the greedy algorithm is adopted to select the next generation. In the calculation process, the global optimal solution is saved during the entire optimization process with an elitist strategy. 5.2. Sample calculation and result analysis Ash content, vitrinite reflectance, fixed carbon content and gas content have obvious responses in gamma, acoustic transit time, neutron and density logging. Therefore, we select these parameters as the training array of the model; meanwhile, we select core samples on which to carry out an actual forecast. The resistivity log is not considered because it has more sensitivity to the fracture. The training and prediction results of ash component, vitrinite reflectance, gas content (shown in figures 11–13 respectively) confirm that the predicting model established by the neural network model has high accuracy. Figure 11. Open in new tabDownload slide Comparison between predicted ash content and coal core analysis result. Figure 11. Open in new tabDownload slide Comparison between predicted ash content and coal core analysis result. Figure 12. Open in new tabDownload slide Comparison between predicted vitrinite content and coal core analysis reflectance and coal core analysis results. Figure 12. Open in new tabDownload slide Comparison between predicted vitrinite content and coal core analysis reflectance and coal core analysis results. Figure 13. Open in new tabDownload slide Comparison between predicted gas content and coal core analysis results. Figure 13. Open in new tabDownload slide Comparison between predicted gas content and coal core analysis results. 6. Fracture evaluation model based on a dual laterolog numerical simulation 6.1. Conductivity model of a coalbed fracture Generally, a coalbed develops two sets of orthogonal or oblique vertical fracture systems, both of which are perpendicular to the bedding plane. The dominant set of roughly parallel, throughgoing, extensive fractures are face cleats, the secondary but less developed are butt cleats. Under in situ conditions, the coal seams can contain additional fracture sets, parallel to the bed plane (Laubach et al 1998). The fracture systems are the main seepage and output channels of coalbed methane and are the key factors controlling coalbed permeability (Pashin et al2004). The coalbed model is simplified to be composed of equally spaced coal-based and vertically crossed fractures, as shown in figure 14. σb and σf stand for the conductivity of the coal matrix and pore fluid, respectively. h1 and d1 stand for the fracture aperture and the distance between two horizontal fractures, respectively; h2 and d2 are the aperture of a vertical cleat and the distance between two face cleats, respectively; h3 and d3 stand for the aperture of a butt cleat and the distance between two cleats, respectively. For the coal seams, h1, h2, h3 are actually far less than d1, d2, d3; σf  is usually greater than σb. The conductivity of a coalbed on a certain direction is equivalent to the parallel and series connections of fracture fluid and coal matrix, therefore displaying obvious anisotropy, and the conductivity tensor is expressed as: 7 where 8 9 10 According to the experimental observation and analysis of coal core cleat density in the study area, the average distance between fractures is about 0.003–0.01 m, so the dual laterolog response shows that a coalbed has macro electrical anisotropy (Deng et al 2011), therefore, the fracture aperture and interval can be replaced by fracture porosity 11 where φ1 stands for porosity of horizontal fracture, φ2 is porosity of the vertical face cleat, φ3 is porosity of the vertical butt cleat and the total fracture porosity is φf, 12 So the elements of the formation conductivity tensor of fracture-induced macroscopic electrical anisotropy can be simply expressed: 13 14 15 6.2. The calculated method for the laterolog response of a fracture For anisotropic media, the differential form of Ohm's law can be written in the form of matrix and vector components (Zhang 1984), 16 where j is the current density vector. E is the electric field intensity vector, σ is the medium conductivity tensor, as shown in equation (7). The dual laterolog response is mainly affected by the product of fracture porosity and fluid conductivity in the fracture, the fissure occurrence, and the coal matrix conductivity. To determine the dual laterolog response requires calculation of a continuous and smooth function of ϕ, satisfying the equation, 17 where ϕ is the potential distribution function. On the constant voltage electrode, ϕ is a known constant. On the constant current electrode, ϕ is an unknown constant. On the surface of a constant current electrode, the supply current of the electrodes is constant. On the insulating boundary surface, the normal current is 0. This problem is generalized to a functional extremum problem (Zhang 1984), the functional is 18 where σij is the (i, j) element of the conductivity tensor, ξ1 = x,ξ2 = y, ξ3 = z, IE and UE are the current and the potential of the electrode E respectively. The sum is of all the electrodes. Ω is the solving area, which is the three-dimensional space removal of the electrode system. The solving area is divided into 120 380 tetrahedral elements. The improved front solving (Zhang 1984) is adopted to solve this functional. 6.3. The logging response characteristics of fractured formation Figure 15 shows the deep and shallow resistivity responses under different combinations of fractures and cleats, where the horizontal axis is the proportion of face cleat porosity accounting for the total fracture porosity (φh) and cleat porosity (φv). The total porosity of fractures and cleats is 0.1%. The vertical axis shows the dual laterolog responses. RLLD and RLLS are the deep and shallow apparent resistivity, respectively. The borehole diameter (D) is 20 cm, mud resistivity Rm is 1Ω m, the liquid resistivity of the fracture Rf is 0.1 Ω m. The matrix resistivity of the coalbed (Rb) is 1000 Ω m, and the total porosity of fractures and cleats is 0.1%. The responses show that the deep and shallow apparent resistivities decrease as the difference between the face cleat porosity and the butt cleat porosity decreases. The dual laterolog resistivities decrease as φh increases. When φh is small, the difference between deep and shallow resistivity is positive. When φh is large, the deep and shallow resistivities show an obvious negative difference. Figure 14. Open in new tabDownload slide Coalbed fracture model. Figure 14. Open in new tabDownload slide Coalbed fracture model. Figure 15. Open in new tabDownload slide Dual laterolog response to the fracture and cleats. Figure 15. Open in new tabDownload slide Dual laterolog response to the fracture and cleats. 6.4. Relationship between the dual laterolog response and fracture porosity The dual laterolog apparent resistivity is transformed into apparent conductivity, as shown in figures 16 and 17, the relationship between the dual laterolog apparent conductivity and fracture porosity is linear, so the laterolog is often used to evaluate the coal fracture or cleats (Deng et al 2010, Chen and Wang 2003, Li et al2011). A coalbed with different coal ranks shows different conductivity. As coal matrix resistivity increases, the dual laterolog apparent conductivity decreases accordingly. When fracture porosity is small, the dual laterolog responses mainly reflect the coal matrix conductivity. Figure 16. Open in new tabDownload slide Relationship between deep laterolog conductivity and fracture porosity. Figure 16. Open in new tabDownload slide Relationship between deep laterolog conductivity and fracture porosity. Figure 17. Open in new tabDownload slide Relationship between shallow laterolog conductivity and fracture porosity. Figure 17. Open in new tabDownload slide Relationship between shallow laterolog conductivity and fracture porosity. 6.5. The dual laterolog computational fracture method 6.5.1. Fracture porosity computational model with coalbed matrix resistivity Rb ≤ 100 Ω m Horizontal fracture is well-developed (0.7 < a ≤ 1) Horizontal fracture and vertical fracture are both well-developed (0.3 ≤ a ≤ 0.7) Vertical fracture is well-developed (0 ≤ a < 0.3) where Rf stands for the fluid resistivity in the fracture, Clld and Clls are respectively the apparent conductivity of a dual laterolog. 6.5.2. Fracture porosity computational model with coalbed resistivity 100 Ω m < Rb ≤ 1000 Ω m Horizontal fracture is well-developed (0.7 < a ≤ 1) Horizontal fracture and vertical fracture are both well-developed (0.3 ≤ a ≤ 0.7) Vertical fracture is well-developed (0 ≤ a < 0.3) 6.5.3. Fracture porosity calculation model with coalbed resistivity Rb ≥ 1000 Ω m Horizontal fracture is well-developed (0.7 < a ≤ 1) Horizontal fracture and vertical fracture are both well-developed (0.3 ≤ a ≤ 0.7) Vertical fracture is well-developed (0 ≤ a < 0.3) 6.6. Validity analysis of the fracture porosity computation Establish coalbed fracture model. With coal bed resistivity ranging from 50 to 5000 Ω m and the fracture porosity ranging from 0.01% to 4%, according to the different fracture states, the three-dimensional finite element method is adopted to calculate the dual laterolog response. According to formulas (9)–(11), the fracture porosity is inverted. The inversion results indicate that when the horizontal or vertical fracture is well-developed, the calculation result is ideal, as shown in figure 18, while when the horizontal fracture and vertical fracture are developed similarly, the computation error is large. Figure 18. Open in new tabDownload slide Validation of fracture porosity calculation. Figure 18. Open in new tabDownload slide Validation of fracture porosity calculation. 7. Case study The coal composition, vitrinite reflectance and gas content are predicted with the improved SADE method for the Carboniferous and Permian system coalbed in the HC region, as shown in figure 19. The stick plot is the coal core test result, and the 7th, 8th and 9th tracks show ash content, fixed carbon and volatile content, respectively. The 10th and 11th tracks show the vitrinite reflectance and predicted gas content, respectively. The predicted results and the test results show good consistency. Based on the fracture porosity with the dual lateral logging in the study region, as shown in figure 20, the stick diagram is the fracture porosity of the coal core. The 7th and 8th tracks show that the calculated fracture porosity and the coal core fracture density have good correlation. Figure 19. Open in new tabDownload slide The predicated results of coal composition and gas content of the coalbed in Well X1. Figure 19. Open in new tabDownload slide The predicated results of coal composition and gas content of the coalbed in Well X1. Figure 20. Open in new tabDownload slide The predicated results of the fracture porosity of the coalbed in Well X2. Figure 20. Open in new tabDownload slide The predicated results of the fracture porosity of the coalbed in Well X2. 8. Conclusion The coalbed gas reservoir in the target area mainly develops meagre coal and lean coal, with ultra-low moisture content, varied ash content, low volatile content of coal base, rich gas content, good correlation between ash content and fixed carbon, dry ash-free base volatile and fixed carbon, fixed carbon and true density. The log responses of a coalbed methane reservoir in the study area show obvious characteristics. Low density, high neutron and high acoustic travel time can be key indicators to divide coalbeds, whereas the correlation between coal parameters and log response, and between gas content and log response are complex in the study area. The evaluation of coal property, coal rank and gas content with the improved the SADE algorithm neural network model has high accuracy, and this method is applicable to coalbed methane reservoir parameter evaluation. A fast computation model for fracture porosity is established on the basis of the dual laterolog numerical simulation. When either the horizontal or the vertical fracture is relatively developed, the calculation of fracture porosity has high accuracy, which can basically meet the requirements of coalbed fracture evaluation. Acknowledgments This study is funded by the National Natural Science Foundation of China (41174009) and the National Major Science and Technology Projects of China (2011ZX05020, 2011ZX05035, 2011ZX05003, 2011ZX05007). We would like to express our sincere appreciation to Professor Mo Xuanxue, geologist and member of the Chinese Academy of Sciences, for his instructive guidance. We would also like to express our heartfelt thanks to the referees’ valuable suggestions which helped us to improve this paper significantly, and give our great thanks for Dr Ren Haitao's precious suggestions and hard revision work on this paper. References Ahmed U , et al. , 1991 An advanced and integrated approach to coal formation evaluation SPE Annu. Technical Conf. and Exhibition (pg. 755 - 770 ) Al-Jubori A , et al. , 2009 Coalbed methane: clean energy for the world , Oilfield Rev. , vol. 21 (pg. 4 - 13 ) OpenURL Placeholder Text WorldCat Amorino C , et al. , 2005 CO2 geological storage by ECBM techniques in the Sulcis area (SW Sardinia Region, Italy) 2nd Int. Conf. Clean Coal Technologies for Our Future 10–12 May Cagliary, Italy Bachu S . , 2007 Carbon dioxide storage capacity in uneconomic coal beds in Alberta, Canada: methodology, potential and site identification , Int. J. Greenhouse Gas Control , vol. 1 (pg. 374 - 385 ) 10.1016/S1750-5836(07)00070-9 Google Scholar Crossref Search ADS WorldCat Crossref Bhanja A K , Srivastava O P . , 2008 A new approach to estimate CBM gas content from well logs SPE Asia Pacific Oil and Gas Conf. and Exhibition 20–22 Oct Perth, Australia Chen H L , Wang R Y . , 2003 Coalbed methane logging interpretation and application , Tuha Oil Gas , vol. 8 (pg. 164 - 169 ) OpenURL Placeholder Text WorldCat Chen J F , Liu B M , Bie Y P . , 1997 The application of logging in the coalbed methane reservoir evaluation , Coal Technol. , vol. 4 (pg. 36 - 40 ) OpenURL Placeholder Text WorldCat Clarkson C R , Bustin R M . , 1997 The effect of methane gas concentration, coal composition and pore structure upon gas transport in Canadian coals: implications for reservoir characterization Proc. Int. Coalbed Methane Symp. University of Alabama, Tuscaloosa, AL 12–16 May (pg. 1 - 15 ) Clarkson C R , Bustin R M . , 2000 Binary gas adsorption/desorption isotherms: effect of moisture and coal composition upon component selectivity , Int. J. Coal Geol. , vol. 42 (pg. 241 - 272 ) 10.1016/S0166-5162(99)00032-4 Google Scholar Crossref Search ADS WorldCat Crossref Deng S G , Li Z Q , Chen H . , 2010 The simulation and analysis of array lateral log response of fracture in coalbed methane reservoir , Coal Geol. Explor. , vol. 38 (pg. 55 - 60 ) OpenURL Placeholder Text WorldCat Deng S G , et al. , 2011 Response characteristics of array laterolog and their primary inversion in reservoirs with fracture-induced anisotropy , Pet. Sci. , vol. 8 (pg. 11 - 16 ) 10.1007/s12182-011-0109-1 Google Scholar Crossref Search ADS WorldCat Crossref Faiz M M , et al. , 2007 Evaluating geological sequestration of CO2 in bituminous coals in the southern Sydney Basin, Australia: natural analogue , Int. J. Greenhouse Gas Control , vol. 1 (pg. 223 - 235 ) 10.1016/S1750-5836(07)00026-6 Google Scholar Crossref Search ADS WorldCat Crossref Fu X H , Lu G Z , Qin J , Jia H Y . , 1999 Determination of coalbed gas content and classification of coalbody structure with log response , Well Logging Technol. , vol. 23 (pg. 112 - 115 ) OpenURL Placeholder Text WorldCat Gao X C , Zhang C C , Duan T L . , 2003 Probe into interpretation on logging informations of coalbed gas , Coal Geol. China , vol. 15 (pg. 54 - 57 ) OpenURL Placeholder Text WorldCat Hoyer D . , 1991 Evaluation of coal fracture porosity from dual laterolog SPWLA 32nd Annu. Logging Symp. 16–19 June (pg. 1 - 15 ) Huang V L , Qin A K , Suganthan P N . , 2006 Self-adaptive differential evolution algorithm for constrained real-parameter optimization IEEE Congress on Evolutionary Computation Canada 16–21 July (pg. 17 - 24 ) Kowalski J J . , 1975 Formation strength parameters from well logs SPWLA 16th Annu. Logging Symp. 4–7 June Paper N Laubach S E , Marrett R A , Olson J E , Scott A R . , 1998 Characteristics and origins of coal cleat: a review , Int. J. Coal Geology , vol. 35 (pg. 175 - 207 ) 10.1016/S0166-5162(97)00012-8 Google Scholar Crossref Search ADS WorldCat Crossref Li H , et al. , 2012 A new method predicting reservoir properties with neural network based on SADE algorithm , Well Logging Technol. , vol. 36 (pg. 585 - 589 ) OpenURL Placeholder Text WorldCat Li J Q , et al. , 2011 Evaluation of the reservoir permeability of anthracite coals by geophysical logging data , Int. J. Coal Geol. , vol. 87 (pg. 121 - 127 ) 10.1016/j.coal.2011.06.001 Google Scholar Crossref Search ADS WorldCat Crossref Li J S , Chen Y . , 1996 Calculation of coalbed gas content with density logging value , Logging Sect. , vol. 5 (pg. 70 - 73 ) OpenURL Placeholder Text WorldCat Lian C B , et al. , 2007 Coalbed gas content predication method based on logging information , Acta Mineral. Sin. Suppl. , vol. 22 (pg. 446 - 447 ) OpenURL Placeholder Text WorldCat Mohseni M , Afroomand A , Mohsenipour F . , 2011 Optimum coordination of overcurrent relays using SADE algorithm , 16th Electrical Power Distribution Conf. (Bandar Abbas, Iran, April 19–20) OpenURL Placeholder Text WorldCat Mullen M J . , 1989 Coalbed methane resource evaluation from wireline logs in the northeastern Sanjuan Basin: a case study Low Permeability Reservoirs Symp. Denver, CO 6–8 March Pan H P , Huang Z H . , 1998 Log interpretation model of determining coalbed coal quality parameters , Geoscience , vol. 12 (pg. 447 - 451 ) OpenURL Placeholder Text WorldCat Pashin J C , Jin G H , Payton J W . , 2004 Three-dimensional computer models of natural and induced fractures in coalbed methane reservoirs of the Black Warrior Basin , Ala. Geol. Surv. Bull. , vol. 174 pg. 62 OpenURL Placeholder Text WorldCat Pashin J C , et al. , 2009 Carbon sequestration and enhanced recovery potential of mature coalbed methane reservoirs in the Black Warrior basin , Carbon Dioxide Sequestration in Geological Media: State of the Science  , vol. vol 59 Grobe M , Pashin J C , Dodge R L . Tulsa, OK American Association of Petroleum Geologists (pg. 125 - 147 ) Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Qiu H J . , 2008 Coalbed methane exploration in China AAPG Annu. Convention 20–23 Apr. San Antonio, TX Sibbit A M . , 1985 The dual laterolog response in fractured rocks SPWLA 26th Annu. Logging Symp. 17–20 June Dallas, TX Singh R , et al. , 2012 A comparative study of generalized regression neural network approach and adaptive neuro-fuzzy inference systems for prediction of unconfined compressive strength of rocks , Neural Comput. Appl. , vol. 21 (pg. 1 - 8 ) OpenURL Placeholder Text WorldCat Sun X H . , 1991 Logging data coalbed analysis method research , Coal Geol. China , vol. 3 (pg. 71 - 77 ) OpenURL Placeholder Text WorldCat Vishal V , Ranjith P G , Singh T N . , 2013a CO2 permeability of Indian bituminous coals: implications for carbon sequestration , Int. J. Coal Geol. , vol. 105 (pg. 36 - 47 ) 10.1016/j.coal.2012.11.003 Google Scholar Crossref Search ADS WorldCat Crossref Vishal V , et al. , 2013b Numerical modeling of Gondwana coal seams in India as coalbed methane reservoirs substituted for carbon dioxide sequestration , Energy , vol. 49 (pg. 384 - 394 ) 10.1016/j.energy.2012.09.045 Google Scholar Crossref Search ADS WorldCat Crossref Yang X C , et al. , 2010 The application of grey correlation analysis in the coalbed methane reservoir evaluation , J. Xian Technol. Univ. , vol. 30 (pg. 202 - 206 ) OpenURL Placeholder Text WorldCat Zhang G J . , 1984 , Electrolog Beijing Petroleum Industry Press (pg. 31 - 56 ) Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Zhang L L , Cai W Y . , 2009 The evaluation method of imaging well logging data in coalbed methane reservoir , Well Logging Technol. , vol. 33 (pg. 571 - 574 ) OpenURL Placeholder Text WorldCat Zhang S H , Tang S H , Tang D Z , Pan Z J , Yang F . , 2010 The characteristics of coal reservoir pores and coal facies in Liulin district, Hedong coal field of China , Int. J. Coal Geol. , vol. 81 (pg. 117 - 127 ) 10.1016/j.coal.2009.11.007 Google Scholar Crossref Search ADS WorldCat Crossref Zhang S R . , 2011 Proposed effective developing method for the deep coalbed methane , China Coalbed Methane , vol. 8 (pg. 18 - 20 ) OpenURL Placeholder Text WorldCat Zou C N . , 2011 , Unconventional Petroleum Geology Beijing Geology Publishing House (pg. 111 - 147 ) 10.1016/B978-0-12-397162-3.00004-9 Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Crossref © 2013 Sinopec Geophysical Research Institute
TI - Integrated petrophysical log evaluation for coalbed methane in the Hancheng area, China
JO - Journal of Geophysics and Engineering
DO - 10.1088/1742-2132/10/3/035009
DA - 2013-06-01
UR - https://www.deepdyve.com/lp/oxford-university-press/integrated-petrophysical-log-evaluation-for-coalbed-methane-in-the-6aHIKmML0l
VL - 10
IS - 3
DP - DeepDyve
ER -