Research on the Influencing Rules of Gas Hydrate Emission Dissipation Coefficient Based on Subspace Spectrum Clustering

Featured by high energy density, low combustion pollution and large quantity, natural gas hydrate has become one of the research hotspots in Sanlutian Field of Muri Coalfield since 2008, when China first drilled natural gas hydrate samples in the permafrost area of Qilian Mountains, Qinghai-Tibet Plateau. However, the study on the controlling factors of gas hydrate accumulation is still shallow, which hinders the exploration and development of natural gas hydrate resources. The controlling factors of gas hydrate accumulation mainly include temperature and pressure conditions, gas source conditions, sedimentary conditions and structural conditions, among which structural conditions are the important one. In this paper, the Sanlutian Field of Muri Coalfield in the southern margin of Qilian Mountains is selected as the main research area and natural gas hydrate is taken as the research subject. According to the sample test results, the geological structure pattern is inferred, and the occurrence characteristics, gas source and hydrate composition of Sanlutian gas hydrate samples are further obtained. Finally, the gas hydrate accumulation model at the end of structure is summarized and illustrated.


Introduction
The gradual decrease of conventional oil and gas resources worldwide, accompanied by the constantly increasing demands on energy and intensified supply-demand pressure, keep urging the scientists to continuously seek for new alternative energies. Since China explored the first physical sample in the perpetually frozen region with medium and low latitude on Qilian Mountain in 2008, natural gas hydrate (NGH) has aroused great concerns and attentions among scientists [1]. As a kind of new natural energy and an important unconventional oil gas resource, NGH is not only an ideal substitute for conventional oil and gas resources, but also a kind of new green and clean energy with low pollution and high efficiency, thus shedding new light on the future direction in this field [2].
As indicated by a new round of evaluation works on coal resource potential that has just been completed, it can be seen that the total quantity of coal resources of Qilian Mountain region reaches 14.082 billion tons. Coal measures serve as an important carrier of comprehensive mineral resources. Besides coal seams, various kinds of mineral resources are reserved or associated in the coal measures in different forms. Gaseous energies mainly contain coalbed methane (CBM), sandstone gas and shale gas, etc., [3] and solid-state energies mainly contain bauxite, sedimentary uranium mine and oil shale etc., as well as solid-state NGH formed under specific conditions, etc. Referring to a saying of Novak, the Russian Minister of Energy when he was having an interview on Apr. 16, 2013, "exploitation of NGH instead of shale gas can raise a revolution in the world energy market." As the standing stock of NGH is twice the total amount of shale gas and natural gas, technologies that are more suitable for exploitation of NGH will be invented in the long run though there has been no effective exploration and exploitation yet. In view of this, the exploitation cost will consequently be reduced [4,5]. Aimed at the broadly-distributed permafrost distribution areas in China, especially the permafrost regions on Qinghai-Tibet Plateau, academic circles have attached great importance to them and believe that these regions have the geological conditions for the formation of NGH. As early as 1999, when the national special project of "a new round of big survey on territorial resources" was started in China, a series of researches have been carried out [6]. The finding of NGH in Muri coalfield, Qinghai Province has set off another round of climax in exploration and research. Ministry of Land and Resources, China Geological Survey, Science and Technology Department of Qinghai Province and China National Administration of Coal Geology, etc., all set up projects and carried out deeper researches successively [7,8].
After years of researches on permafrost regions of Qinghai-Tibet Plateau, key breakthroughs were finally made by Chinese scientific researchers in Muri region, Qilian Mountain, Qinghai Province. As the first country in the world to collect the aquo-complex samples from perpetually frozen regions, China's discovery of aquo-complex in Muri region and the accompanied researches on all aspects have a great significance on theoretical researches [9]. Compared with NGH from sea areas and polar region tundra, NGH from medium and low latitude present different characteristics in terms of existing conditions, rules and main controlling factors, etc. Therefore, development of fundamental researches on characteristics of NGH is helpful to the development and utilization in Qinghai-Tibet Plateau regions.

Dissection of Typical Oil and Gas Reservoir in Active Regions of Mud Diapiric Fold
The Muri coal and gas field at the southern edge of Qilian Mountain belongs to the gas field groups at the eastern part of Yinggehai Basin, which is the largest commercial natural gas field found in this basin through exploration up to now, and its geographical position is at the northwestern part of central mud diapir structure belt in Yinggehai Basin ( Fig. 1) [10]. With a water depth of 75 m, the main pay zone of this gas field is Yinggehai of Pliocene series, now over 10 exploratory wells have been drilled in Muri coalfield 2/3 at the southern edge of Qilian Mountain, through which it was explored that when the gas bearing area was 287.7 k, the geological reserves of natural gas was explored to be 1296.38 × l0 8 m 3 , of which pure hydrocarbon gas took 612 × 10 8 m 3 , this gas field was found through exploration in the 1990s and was put into exploitation in 2003, up to the end of 2011, the accumulated output of natural gas was 171.77 × 10 8 m 3 . Shallow gas reservoir of Muri coalfield at the southern edge of Qilian Mountain is located at the northwestern part of central mud diapir structure belt in Yinggehai Basin [11], whose structure is a brachyaxis anticline with inherited development that is formed through the mud diapiric fold function. Its long axis is 21 km and the minor axis is 12 km; the entrapment is of a large area, of which the area of the largest entrapment can reach 287.7 km 2 , and it is of a high closure (closure range is 219-254.8 m) and a shallow burial, as shown in Fig. 2. The anticline structure is steep in the east and gentle in the west; a northsouth breakage is developed on the tectonic axis, distributed in the shape of arborization, and the anticline structure is divided into eastern and western blocks, causing the effect that different blocks composing the wings have different pressure systems and natural gas components. Muri coal gas field at the southern edge of Qilian Mountain is a structural gas pool mainly consisting of mud diapiric fold anticlines and fault control, parts of which are lithologic gas reservoirs. The gas reservoirs contain multiple gas groups, and the distribution of gas and water is controlled by the structure and faults, which is a structural gas pool with the driving type of elastic water drive. Pressure coefficient of the gas reservoir is 1031.14, so it is a gas pool with atmospheric pressure.
The shallow gas reservoir of Muri coalfield at the southern edge of Qilian Mountain is the Yinggehai stratum of Pliocene epoch, shoals and sand bank sedimentations similar to shore-shallow sea phase belt are formed by underwater highlands deposited in the shallow sea environment, which show a high-mesoporous and low-hypotonic feature on the whole. The cap rock is the whole set marine mudstone formed by Yinggehai series and Ledong series, which is of a large thickness and good sealing feature. There are 3 types in the reservoirs of Muri coalfield at the southern edge of Qilian Mountain: the lithology of the first type is thick-bedded sandstone porous-type reservoir, which mainly consists of grey thick-layer fine sandstones, sometimes mixed with siltstone stripes, whose general thickness of single layer is 5 m. The second type is fractured argillaceous siltstone reservoir, which mainly consists of argillaceous siltstones, whose general thickness of single layer is 15 m. The lithology of the third type is argillaceous siltstone reservoir, with the reservoir space consisting of matrix micropores.
It is indicated by the geochemical characteristics of shallow gas reservoir of Muri coalfield at the southern edge of Qilian Mountain that, its source rocks are mainly from Sanya-Huangliu marine mudstone of Miocene series, the lithology of source rocks is mainly silty mudstone and mud rock, and its organic matter type is humic mixed type-humic type. The TOC content is 0.39%-0.70% with an average proportion of 0.50%, the contents of chloroform bitumen A and total hydrocarbon are respectively 0.0561% and 353 × 10 -6 , which can reach the level of good source rocks on the whole. There are great changes in the chemical constitution of natural gas in the Muri coal gas field at the southern edge of Qilian Mountain, in different air layers and even the same air layer, features of natural gas from different fault blocks and especially the contents of non-hydrocarbon gas and isotope composition of methane are widely different, and an obvious anisotropism is showed up. It can be divided into 3 types based on the content of hydrocarbon and non-hydrocarbon: the first type is the hydrocarbon gas reservoir whose natural gas component is mainly methane (over 75%), the constituent content of heavy hydrocarbon is low (3.37%), the value of aridity coefficient is 0.99, most of which is dry gas; the second type is CO2 gas reservoir (content is as high as 5588%); and the third type is N2-rich gas reservoir, whose Na content is as high as 1531%, and the CO2 content is less than 5%.

Spectral Clustering Based on K-harmonic Mean
If dissipation coefficient analysis data is seen as a document, and the labels of dissipation coefficient analysis data are seen as the key words of the document, so if we conduct clustering with the feature of labels of dissipation coefficient analysis data, clustering of dissipation coefficient analysis data can be conducted through the method of this paper. Currently, popular clustering methods are mainly methods based on division with K-Means as representative and spectral clustering method, etc. Spectral clustering is to cluster based on graph theory and similarity among data, as it has no relationship with dimensionality of data points and simply related to the number of them, it is suitable for non-measure spaces. It has been broadly concerned, but traditional clustering algorithm is sensitive to selection of initial center, which makes its computational result unstable, and it is easy to be caught in partial minimal points. Meanwhile, there is particularity on the clustering problems of this paper, vectors of this paper are generally sparse vectors containing multiple 0, which has brought about difficulties to the selection of clustering center. Thus, the performance of spectral clustering algorithm is improved through introduction of K-harmonic means. K-harmonic means (KHM) algorithm [11] is a kind of clustering algorithm based on the center, through calculating the harmonic mean of distance from data points to the clustering center, its performance function is constructed through this algorithm. Expression of the algorithm is as follows: Algorithm 1: K-harmonic mean spectral clustering algorithm Input: n data points ) , ( Step 2: construct a Laplacian matrix, Step 4: convert the row vector of matrix Z to unit vector and acquire matrix Step 5: cluster each row of matrixY into k categories through KHM algorithm. As the harmonic average of distance from data points to the center of all clustering replaces the minimum distance from data points to the center of clustering in KHM algorithm, it overcomes the sensitiveness on original value.

Dissipation Coefficient Analysis Data Clustering Based on Subspace
Set that } ,...

Low-Level Features and Semantic Mapping Relation of Dissipation Coefficient Analysis Data
Dissipation coefficient analysis data in each semantic category are divided into different regions, and the extracted low-level feature of the divided regions is expressed with vector f . f is expressed with 24dimension vector. Based on the similarity relation of low-level regional features, KHM algorithm is used to cluster the similar regions of the same semantic category into the same blob once again and form K ) blobs. Through the number of optimized blobs K, internal spur of each blob is concentrated as far as possible and the distance among blobs is kept away to the largest extent. To confirm the optimal value of K, several Based on Davies-Bouldin index [11], Formula (2) is as follows: (2) Acquire that the K at its minimum value should be the optimal K value, that is: Thus k blobs are acquired in each semantic category. Those blobs inherit the semantics i L that they are in.
Through twice clustering, training dataset of dissipation coefficient analysis T is divided into several semantic categories , each semantic category is expressed with the central features of the usable blobs and the key words In any semantic category, distribution of joint probability of blob i b and key word i i j L w ∈ is acquired through the following formula: In the formula, is the priori probability of category i C , ) , | ( can be acquired through calculation of the following formula:

Empirical Analysis
It can be known from the measured data of research area (Tab. 1) that the HI scope if coal sample is 55-187 mg/g and the average value is 132.73 mg/g, and it is divided into III2-type (three-category fivegroup method) kerogens based on standard of coal organic matter types, which are converted to III-type kerogens through three-category inquartation rules [12,13]. The scope of (S1 + S2) is 72.09-185.31 mg/g with the average value of 129.21 mg/g, which is divided into III1-type (three-category five-group method) kerogens, and is converted to II-type kerogens through three-category inquartation rules. HI scope of mudstone sample is 29-476 mg/g and the average value is 213 mg/g, it is divided into II-type kerogens based on standard of continental facies source rock organic matter type; scope of D is 3.74%-41.77% with the average value of 19.56%, which is divided into II:-type kerogens; scope of (S 1 + 52) is 0.23-15.45 mg/g with the average value of 6.58 mg/g, which is divided into II1-type kerogens. IH scope of oil shale samples is 112-311 mg/g, its average value is 163 mg/g, which is divided into II:-type kerogens based on standard of continental facies source rock organic matter type; scope of D is 10.67%-26.68% with the average value of 14.79%, which is divided into II:-type kerogens; scope of ( S1 + S2) is 1.98-11.41 mg/g with the average value of 4.79 mg/g, which is divided into II1-type kerogens. The division results can be seen in Figs. 3 and 4.   Based on collection and settlement of research and experimental data of predecessors and actual measurement of samples collected from well fields and NGH drill holes of the adjacent region (measured by Emmi of national key laboratory of coal resources and safe mining), the results show that the minimum Ro value of vitrinite reflectance of coal sample is 0.71%, the maximum value is 2.26% and the average value is 1.18%. There is a little measured data of oil shale and mudstone, the Ro value of oil shale is 0.94%; the minimum Ro value of mudstone is 0.72%, the maximum value is 0.95% and the average value is 0.86%.
In addition, in the practical investigation of drill cores, oil patches can be seen many times (see Fig.  5). Thus it can be known from analysis in above text that oil shale and mudstone of the research area enter the mature stage, in which petroleum and wet gas can be generated; while coals mainly enter the maturehigh-mature stage, in which petroleum, wet gas and condensate gas can be generated, individual samples show that the coals have entered the postmature stage, in which dry gas can be generated.

Conclusion
It is thought through comprehensive analysis that control actions of construction on reservoir forming of NGH can be divided into 4 aspects: (1) Structural framework controls the distribution of NGH: NGH mainly gathers in blob M1 and S1, etc., and it is more concentrated in areas with deeper burial depth; in addition, F2 and F30 breakage has a significant control action on its reserves.
(2) Tectonic movements control the formation of hydrocarbon gas and warm-pressing stabilized zones, in addition, it also transforms the reservoirs in 2 ways, namely deformation and deflection: main period of hydrocarbon generation of source rocks in well fields was the end of early Baiscicus, and hydrocarbon generation was stopped in later periods; formation time of perpetually frozen area of Qilian Mountain was no early than around 3.6 Ma till now, then the formation of warm-pressing stabilized zone was also no early than around 3.6 Ma till now; tectonic movements have caused permanent deformation and deflection on the reservoirs, enlarging or reducing the burial depth of gases, then it enters the warm-pressing stabilized zone and thus forming NGH reservoirs.
(3) Different structural configurations have different preserving functions on NGH: breakages and fractures provide passages for the movement of hydrocarbon gases, breakage crushed zones and fractures provide reservoir spaces beneficial for the gases, and construction formation circles store and accumulate the gases.
(4) Similar to tectonic movements, destructions of structures on NGH can be divided into transformation and deflection damages: transformation of reservoirs let off the gases and thus no NGH can be formed; deflection of reservoirs make the gases crop out and disperse, thus no NGH can be formed, or the gases are deeply buried and cannot enter the warm-pressing stabilized zones to form NGH or to be explored.

Conflicts of Interest:
The authors declare that they have no conflicts of interest to report regarding the present study.