Frontiers in Computing and Intelligent Systems ISSN: 2832-6024 | Vol. 11, No. 3, 2025 162 Analysis of the Main Control Factors Affecting the Production Capacity of Surig Gas Field Su 53 Area Xiaobo Li a, Haibiao Wang b Production Optimisation, Tanggu Tianjin, 300459, China a lixb13@cosl.com.cn, b wanghb75@cosl.com.cn Abstract: Accurately identifying the main controlling factors of the fracturing effect of horizontal wells in tight gas reservoirs, and then effectively guiding the optimisation of the fracturing scheme, is the key to enhance the fracturing capacity of horizontal wells in tight gas reservoirs. Relying on the geological and engineering big data of Su53 block in Surig gas field, four algorithms, namely, Spearman coefficient, maximum mutual information coefficient, Copula entropy and grey correlation, were used to identify the main controlling factors affecting the fracturing effect of horizontal wells under different completion methods, and the weights of the four algorithms were weighted and combined in order to reduce the contingency of the evaluation results of a single weight, and the geological and engineering big data of the 126 barehole horizontal wells and 21 cased horizontal wells of this block were summed up. The study summarises the relationship between the geological and fracturing parameter characteristics of 126 openhole horizontal wells and 21 cased horizontal wells in the block and the test production of a single well, and puts forward suggestions for efficient development of horizontal wells at a later stage. The study shows that the formation pressure coefficient is the main controlling factor for determining the single-well productivity of horizontal wells in this zone, and the productivity of barehole completed horizontal wells is controlled by the fracturing fluid return rate, effective reservoir encounter rate and mud content, while the productivity of cased horizontal wells is controlled by the number of fractured sections, effective reservoir encounter rate and average section length. Keywords: Sulig Gas Field; Horizontal Wells; Production Capacity; Weights; Main Control Factors. 1. Introduction The analysis of factors affecting gas well production capacity points out the direction of how to develop the gas reservoir efficiently, clarifies the relationship between various factors and production capacity, and has guiding significance for the deployment of new wells and development adjustment in the later stage of the study area, which is conducive to proposing the optimisation technical countermeasures for horizontal well development [1]. Experts and scholars at home and abroad have also done a lot of research on the main control factors of horizontal well production capacity in different oil and gas reservoir types. Zhao Hongtao et al [2] used grey correlation method to analyze the correlation between each parameter and the specific oil recovery index (SORI), to determine the main control factors and to establish a prediction model of SORI, which provides a certain guiding basis for the preferential selection of the test layer of the exploratory wells in the regional thick oil reservoirs and the production capacity estimation before the test. Xue Ting et al [3] used grey correlation method and random forest algorithm to systematically analyze the degree of influence of geological, fracturing construction and other parameters on the production capacity, clarified the main controlling factors of single-well production capacity, and optimised the deployment of geological wells and fracturing construction parameters. However, at present, the analysis and evaluation of gas well productivity are all for the same completion method, and the change of completion method in the same block may lead to inconsistency in the main controlling factors of horizontal well productivity, resulting in the mismatch between the fracturing construction plan in the gas field site and the geological conditions of the reservoir, which restricts the gas field from realising stable production and efficient development. Therefore, there is an urgent need to clarify the indicators of gas well productivity under different completion methods, and to find out the influence of geological and engineering parameters with good correlation on productivity. In this paper, Spearman's correlation coefficient, maximum mutual information coefficient, Copula entropy and grey correlation method are used to analyze the weights of the factors influencing the production capacity of horizontal wells under different completion methods in the target block, and the weighted combination of the weights of the four algorithms is used to reduce the contingency of a single weight on the evaluation results, to sum up the laws of the influence of the geological and construction parameters of the block on the production capacity, and to choose the optimal values of the parameters. The results of the four algorithms are combined to reduce the chance of single weighting on the evaluation results, summarize the influence of geological and construction parameters on gas production, and select relatively better parameter values. 2. Block Overview Sulige gas field Su53 zone is a low porosity, low permeability large distribution of lithological gas reservoirs, with the characteristics of rapid phase change of reservoir sands, small effective thickness of single reservoir layer, and strong non-homogeneity of reservoir sands. In order to be more conducive to fracturing and transformation, the zone has carried out technological innovation and tests under the transformation of completion methods since 2018, horizontal wells are transformed from bare eye completion to casing completion, the effectiveness of inter-segmental sealing is improved, casing completion bridge-plug linkage fracturing 163 has become the mainstream process mode, the fracturing parameters are continuously optimised, and the production rate of single wells is significantly improved [4]. Fig 1. Evolution of horizontal well completion and fracturing process in Sourig gas field Its main reservoirs are the sand body of Box 8 section of the Lower Stone Box Formation and the sand body of Shan 1 section in the upper part of the Shanxi Formation. Box 8 section is mainly a braided diversion river deposition, with coarse sandstone, sandstone conglomerate and medium- coarse sandstone dominating in the section structure. The Shan 1 section generally exhibits the characteristics of curvilinear fluvial deposition, and the sand body is relatively narrow and small in thickness, with obvious binary structure. The porosity of the reservoir Shan 1 section is generally 5.0% to 12.0%, with an average of 8.0%, and the permeability ranges from 0.1×10-3 to 1.0×10-3µm2, with an average of 0.503×10-3µm2. The porosity of the Box 8 section is generally 5.0% to 14.0%, with an average of 8.9%, and the permeability ranges from 0.1×10-3 to 1.0×10-3µm2, with an average of 0.782×10-3µm2, which is a low porosity and low permeability reservoir. 3. Methodology for Analysing the Main Control Factors Affecting Production Capacity There are many factors affecting post-fracturing capacity, and the degree of influence often varies greatly. Clarifying the degree of influence of each factor on post-fracturing capacity is a prerequisite for the optimisation of fracturing process. In this paper, we propose to use multifactor analysis to analyse the degree of contribution of reservoir geology and construction parameters to gas production in a block. There are various methods being used to measure the correlation between sample characteristics, and common correlation metrics are as follows. 3.1. Spearman's Coefficient Method Spearman correlation coefficient has the advantages of being independent of the data magnitude and insensitive to abnormally large numbers, etc. Spearman correlation coefficient is calculated as follows:          1 2 2 1 1 N n n n B N N n n n n X X Y Y W i X X Y Y            (1) The value of Spearman correlation coefficient is between - 1 and +1; -1 means that the two variables are completely negatively correlated, +1 means that the two variables are completely positively correlated, and 0 means that the two variables are completely unrelated; the closer to 1 or -1 means that the stronger the linear correlation of the two variables. Using this coefficient to select the best comprehensive evaluation method, it is necessary to select a group of samples first, and at the same time, establish the criteria for ranking the samples in a reasonable level of comprehensive evaluation. Then according to the different evaluation methods on the sample of different rank ordering and reasonable rank ordering of the degree of correlation between the rank of the Spearman coefficient size to select the best [5]. 3.2. Maximum Mutual Information Factor Maximal Information Coefficient (MIC) is a non- parametric method proposed by Reshef as a tool for exploratory data analysis. Compared to traditional linear correlations or exponential relationships, MIC is capable of discovering a wide enough range of correlations on a sufficiently large sample set and is not limited by specific relationship types. The core idea is to encapsulate the dataset from block to block through scatterplot chunking. If there is a correlation between variables, a grid can be plotted on the scatterplot, exhausting all the possible methods of grid delineation, calculating mutual information under each delineation and using this as the information coefficient, and taking the maximum value as the maximum information coefficient. Due to its advantages of adaptivity, non-linearity and freedom from distributional assumptions, it has been widely used in data analysis in bioinformatics, medicine, finance and other fields. For a bounded set 2D R and *,x yn n N , definition:     * , , * |x y GI D n n max I D (2) In the formula, *I denotes the maximum value of mutual information of grid G for column xn and row yn , and  |GI D denotes the value of mutual information of D under grid G partitioning. Define the identity matrix  M D of the bivariate data set:       * 2 , , log * , x y x y I D n n M D min n n  (3) In the formula, I is the maximum value of M. For the bivariate data set D and the number of samples n, the 164 maximum information coefficient can be expressed as formula (3), which is normalised to give the weights of the influencing factors.       * x xn n B nMIC D max M D  (4) 3.3. Copula Entropy Copula function is a function used to describe the dependence between multidimensional random variables and is independent of the marginal distribution function, which is defined as follows: 1 1 1( 1, 2,..., ) ( 1 1 ( 1), 2 2 ( 2),..., ( ))C u u un P X F u X F u Xn Fn un      (5) In the formula, where X1, X2, ......Xn are n random variables, F1, F2, ......Fn are their marginal distribution functions, and u1, u2, ......un are coordinates on the unit hypercube. Copula entropy is based on the concept of Copula function. It can be used to measure the complexity of dependencies between multidimensional random variables. Copula entropy is defined as follows: H(C) ... (u1,u2,..., un) log (u1,u2,..., un) du1du2 dunC C    (6) In the formula, C is the Copula function and H(C) is the Copula entropy. 3.4. Grey Correlation Method Grey correlation is a method used to study the correlation between factors. The basic idea is to convert the relationship between factors and targets into similarities between factors, and then to derive the grey correlation between each factor and the target factor by comparing the similarities between the factors. Grey correlation analysis can be used to deal with various types of data, and its main advantage is that it can be analysed in the case of lack of data or incomplete data, and it does not need to carry out complex statistical processing such as hypothesis testing or parameter estimation on the data, and the specific application steps are as follows. (1) Take the unit reservoir average daily pressure drop production as a reference series, and the influence factors as a comparison series, and carry out dimensionless processing. (2) Calculate the grey correlation coefficient between the comparison series and the reference series using equation (7).     min max 0 0 max i i k k          (7) In the formula, max and min represent the absolute difference between the largest and smallest sample data in the comparative series respectively, and represent the absolute interpolation of the corresponding sample values in the comparative series and the reference series.  is the resolution coefficient, ranging from 0 to 1, which is an important parameter to control the difference between the correlation coefficients, and is generally taken as 0.5. The following principles are followed in taking the value: firstly, the resolution coefficient is dynamically taken according to the actual situation of the sequence; secondly, when there is a singular value in the sequence, the resolution coefficient is taken as a small value in order to reduce the influence of the singular value; thirdly, when the sequence is relatively smooth, the resolution coefficient is taken as a large value to reflect the overall correlation of the correlation coefficient. overall nature of the correlation. According to this principle, while considering the outliers dominating the system correlation value, the mean value of the difference of all absolute values can be expressed by y .    0 1 1 1 m n y i i k Y k Y k n m        (8) In the formula, n and m are the number of samples and the number of influences, respectively, noting that * y max      , when the relationship max 3 y   is satisfied, 1.5     ; when max 3 y   , 1.5 2     . (3) Calculating the grey correlation 0i :  0 0 1 1 n i i k k n      (9) (4) Calculating the weights and normalising the correlations gives the correlation weights for each comparison series  GW i :   0 0 1 i G m i i W i      (10) 4. Analysis of Factors Affecting the Capacity of Horizontal Wells 4.1. Data Preprocessing The data collected in this paper are from the actual production of the Su53 block. Due to the differences in the data records of the same block and the presence of missing values, or outliers in the actual production data, it is not possible to train directly. Therefore, before analysing, it is necessary to carry out data cleaning and other operations to obtain higher precision data collection: (1) Missing value processing Currently, there are two main types of missing value processing methods: 1) Directly delete sample groups or features containing missing values. If a group of samples or a feature has too much missing data, the group of samples or the feature value is deleted. 2) Fill in the sample groups or features containing missing values. Specific filling methods include: 1 plurality, median and mean filling method. 2 neighbouring values to fill, generally using the data before and after the missing value to fill. 3 Predictive modelling methods for filling, building corresponding machine learning models for missing value prediction filling. 4 Lagrangian and other interpolation methods for filling. 5 KNN algorithm for filling, by 165 comparing the corresponding features in the complete dataset and the missing data, and calculating the distance between the missing data and each sample in the complete dataset, then the missing data value is obtained by averaging multiple samples with the smallest distance. In this case, the sample distance is calculated as follows  2 1 ( , ) m i i i d p q p q    (11) In the formula, ( , )d p q —— the distance between two samples; ip , iq —— the corresponding point data of different samples. According to the above two processing methods of missing values, combined with the characteristics of gas field data and the specifics of the collected field data of fractured horizontal wells, the steps of processing missing values in this paper are as follows: 1) Since the two production indicators, gas production and water saturation, are affected by multiple factors, there is a certain degree of randomness in processing the missing values of these two production indicators purely from the perspective of data. Therefore, considering the accuracy of the results, if a group of samples is missing the two production indicators of gas production and water saturation, the group of samples will be deleted directly. 2) Features with more than half of the missing values in the original data are deleted. 3) Considering that there is not much difference in data such as layer conditions and production system between horizontal wells in the same block. Therefore, in this paper, the vacant values of features such as mud content, return rate, total porosity and total fluid volume in the collated data samples are filled by KNN. The figure reflects the principle of the KNN algorithm, whose basic idea is to find sample states in the data that are similar to the current state, and apply the sample states that match the current state to the prediction. Although the KNN method is simple and effective, the value of K also affects the final result at the same time. Therefore, the data is fitted by using different values of K while utilising the Random Forest algorithm, which has a strong ability to perform in small samples. Fig 2. Principle of the KNN algorithm (2) Outlier Handling Before outlier processing, outlier detection should be performed first. The commonly used detection methods are as follows: 1) Lajda ( 3 ) criterion: assuming equal precision measurements of variables, if the residual error b of a given measurement bx satisfies 3b bv x x    , bx is considered a bad value with a large error value and it is removed. In the formula, x is the arithmetic mean of the measured values. 2) Box plot analysis: In a box plot, an outlier is usually considered to be greater than the upper quartile + 1.5 times the interquartile spacing, or less than the lower quartile + 1.5 times the interquartile spacing. In this case, the upper quartile means that 1/4 of all values are greater than it; the lower quartile means that 1/4 of all values are less than it; and the interquartile spacing refers to the difference between the upper quartile and the lower quartile. Fig 3. Principle of box-and-line diagram construction 3) According to the operating experience of field engineers, the range of all variable data is firstly sorted out, and then the outliers are identified according to the range of each variable. Considering the specificity of data samples in the gas field 166 field, this paper adopts the 2nd method for outlier identification. For a small number of data with excessive errors in recording, manual intervention is also required to ensure the screening accuracy of outliers. A total of 94 outlier samples are identified through this comprehensive method, and considering the small number of samples obtained from collation, the method of treating outliers as missing values and filling them in using KNN is adopted. After processing the anomalous values and missing values, the available samples of 147 wells were obtained from 241 horizontal wells in the block, of which 126 were barehole wells and 21 were cased wells, so as to establish a database for analysing the main controlling factors of the production capacity of Su-53 block. Fig 4. Identification of outliers in the dataset 4.2. Portfolio Weighting The above four models can all solve the degree of contribution of each factor to gas production after fracturing of horizontal wells, in order to avoid the problem of chance caused by a single evaluation method, firstly, the impact weights of each parameter obtained by different evaluation methods are linearly normalised, and then each production impact indicator is combined, and the parameter weights are summed up according to formula (12) to obtain the final combined weights  W i . Fig 5. Spearman's coefficient method to obtain weights WB(i) Fig 6. Maximum mutual information method to obtain weights WD(i) Fig 7. Copula entropy to obtain weights WC(i) Fig 8. Grey correlation method to get weights WG(i) 167                   1 B D C G m D C G i B W i W i W i W i W i W i W i W i W i           (12) 4.3. Analysis of Main Control Factors Based on the actual data of Su53 block in Ordos Basin, 126 barehole horizontal wells and 21 cased horizontal wells with relatively complete data are taken as the research objects to analyse the main influencing factors of horizontal well production capacity under different completion methods. 4.3.1. Barehole Completion Horizontal Wells According to the gas test data and model calculation weighting results, 10 parameters with relatively large combined weighting values (formation pressure coefficient, return rate, effective reservoir encounter rate, mud content, etc.) were selected for comparative analysis, as shown in Tables 1 and 2. The average weighting value of geological factors is larger than the average weighting value of engineering factors, which indicates that the geological factors have a higher degree of influence on the production than the engineering factors on the production. Table 1. Combined weights of different geological factors for horizontal wells with barehole completion Factor Spearman's coefficient Maximum mutual information coefficient copula entropy Grey correlation Combination weights Stratigraphic pressure coefficient 0.160 0.078 0.067 0.047 0.352 Silt content 0.077 0.054 0.069 0.041 0.242 Total porosity 0.073 0.063 0.055 0.044 0.235 Permeability 0.065 0.046 0.064 0.043 0.218 Table 2. Combined weights of different engineering factors for horizontal wells with barehole completions Factor Spearman's coefficient Maximum mutual information coefficient copula entropy Grey correlation Combination weights Fracturing fluid return rate 0.126 0.053 0.034 0.037 0.249 Effective reservoir encounter rate 0.105 0.052 0.040 0.046 0.243 Fluid consumption of single section 0.053 0.056 0.038 0.046 0.193 Total sand volume 0.031 0.040 0.068 0.046 0.185 Proportion of precursor fluid 0.016 0.050 0.069 0.049 0.184 Total liquid nitrogen 0.064 0.044 0.029 0.043 0.180 The influential factors with larger combination weights are plotted against the cumulative one-year gas production at the beginning of the gas well in a scatter plot, and it can be seen from Figs. 9-12 that a high pressure coefficient corresponds to a high production rate, and with the increase of the block exploitation years, the reservoir pressure appears to be attenuated, which results in the decrease of the stable production capacity of the gas wells, and the gas production capacity of the reservoirs in high-pressure zones is much larger than that of the low-pressure zones, and the gas reservoir development should prioritise the reservoirs with a relatively high pressure. Due to the high return rate, the fracture will not be able to be effectively supported and closed, which is not conducive to fluid seepage, and the production capacity decreases with the increase of fracturing fluid return rate. There is an obvious positive correlation between the production capacity and the drilling encounter rate, and the size of the drilling encounter rate determines the scale of effective fracturing construction, and the drilling encounter rate should be improved by optimising the trajectory of the borehole. In conclusion, the post-fracturing production capacity of barehole completed horizontal wells has a strong correlation with geological conditions. Geological factors are the basis of reservoir production capacity, so the geological sweet spot should be preferred for the development of horizontal wells, and the scale of fracturing is the guarantee of production capacity, so the scale of fracturing should be increased appropriately with the combination of economic cost and actual site conditions. Fig 9. Relationship between Stratigraphic pressure coefficient and annual cumulative gas production Fig 10. Relationship between fracturing fluid return rate and annual cumulative gas production 168 Fig 11. Relationship between effective reservoir drilling rate and annual cumulative gas production Fig 12. Relationship between mud content and annual cumulative gas production 4.3.2. Casing Completion Horizontal Wells As the development of Surig gas field shifts from rich area to sub-rich area, the reservoir quality gradually decreases, and the production rate of horizontal wells put into production decreases year by year. In order to increase the production of single well, casing completion is gradually adopted in horizontal well completion in Sulige 53 block, and it has been proved to be more conducive to fracturing and reforming, and it is of great significance to clarify the main controlling factors of production capacity corresponding to casing completion for the adjustment of the development plan of this gas reservoir in the next step. Ten parameters (formation pressure coefficient, total porosity, number of fractured sections, effective reservoir encounter rate, etc.) with relatively large weight values are selected for comparison and analysed in Tables 3 and 4, and the factors that have a greater influence on the production capacity of casing-completed horizontal wells are the formation pressure coefficient, the number of fractured sections, the effective reservoir encounter rate and the average section length. Table 3. Combined weights of different geological factors for casing completion horizontal wells Factor Spearman's coefficient Maximum mutual information coefficient copula entropy Grey correlation Combination weights Stratigraphic pressure coefficient 0.115 0.064 0.034 0.048 0.261 Total porosity 0.051 0.042 0.055 0.044 0.192 Table 4. Combined weights of different engineering factors for casing completion horizontal wells Factor Spearman's coefficient Maximum mutual information coefficient copula entropy Grey correlation Combination weights Number of fractured sections 0.097 0.045 0.040 0.049 0.231 Effective reservoir encounter rate 0.050 0.073 0.040 0.047 0.209 Average section length 0.070 0.045 0.050 0.044 0.209 Fracturing fluid volume per stage 0.070 0.082 0.008 0.045 0.206 Total sand volume 0.035 0.051 0.068 0.044 0.197 Number of fractures 0.046 0.046 0.057 0.045 0.194 Horizontal section length 0.096 0.036 0.011 0.045 0.188 Percentage of pre- fracturing fluid 0.030 0.039 0.069 0.046 0.184 Due to the serious lack of data on annual cumulative gas production from cased completions in this block, in order to reduce the error and improve the accuracy of the results of the analysis of the main control factors of production capacity, the influencing factors with larger combination weights are plotted on a scatter plot with the cumulative gas production of the gas wells in the initial 90 days (Figs. 13-16). Similarly, low formation pressure seriously affects the gas production capacity of cased wells in this reservoir, which is a technical challenge that needs to be solved urgently. Compared with openhole wells, the number of fracturing sections has a greater impact on the production capacity of cased wells, because fracturing in sections of cased wells is more controllable and precise, and increasing the number of fracturing sections (decreasing the average section length) is conducive to increasing the scale of fracturing construction and improving the production of a single well, provided that the difference in the length of horizontal sections is not too large. Increasing the amount of fluid injected into each fracture stage can make the fracture extend effectively, while increasing the amount of proppant increases the effective fracture volume and creates a good seepage channel for gas well production. 169 Fig 13. Relationship between Stratigraphic pressure coefficient and 90-day cumulative gas production Fig 14. Relationship between fracturing fluid volume per stage and 90-day cumulative gas production Fig 15. Relationship between total sand content and 90-day cumulative gas production Fig 16. Relationship between average segment length and 90-day cumulative gas production 5. Conclusion (1) The reliability of the weights of the factors influencing the production of tight gas wells after fracturing in the study block obtained by using the combined weighting method is better than that of the single weighting calculation method, which can effectively make up for the solution bias caused by the single method due to the limited data samples or differences in the method principles. (2) The pattern of horizontal well production capacity under different completion methods and single fracturing influencing factors is not obvious from the analysis of field data, which indicates that the horizontal well production capacity is affected by a combination of many factors such as geological and engineering conditions. (3) In the Su-53 area, in order to obtain high production capacity, gas wells need to be geologically located in the zone of good physical properties and gas content, and at the same time, horizontal well development should be implemented according to local conditions, adopting a relatively reasonable method of reservoir modification and maximising the use of geological reserves. (4) Formation pressure attenuation is the main factor restricting the efficient development of gas reservoirs in this block, so we should optimise the fracturing construction plan for low-pressure reservoirs, and put forward efficient development techniques and technical countermeasures for horizontal wells suitable for low-pressure zones. References [1] Zhao Hongbing. Research on the technology of judging the main control factors of horizontal well production capacity in volumetric fracturing of tight reservoirs[D]. Xi'an Petroleum University,2021. [2] ZHAO Hongtao, YU Xi, YU Weiqiang, FAN Xinlei. Analysis and application of main controlling factors of thick oil reservoir capacity in Bohai Sea[J]. Complex Oil and Gas Reservoirs, 2022, 15(01):44-47. [3] XUE Ting, HUANG Tianjing, CHENG Liangbian, MA Shuwei, SHI Jianchao. Optimisation of production control factors and development countermeasures for horizontal shale oil wells in Qingcheng Oilfield, Ordos Basin[J]. Natural Gas Geoscience, 2021,32(12):1880-1888. [4] LI Xianwen, WANG Lili, WANG Wenxiong, XIAO Yuanxiang, CHEN Baochun, ZHANG Yanming, ZHOU Changjing, MA Zhanguo, SHI Huaqiang. Key technological innovation and efficient development practice of fracturing based on small borehole completion--A case study of tight gas reservoir in Surig gas field[J]. Natural Gas Industry,2022, 42 (09):76-83. [5] Shang Weiping. Optimisation of comprehensive evaluation methods using Spearman's coefficient[J]. Jiangsu Statistics, 1996 (08): 19-21.