Send Orders for Reprints to Reprints@benthamscience.ae New Reservoir Grading Method for Tight Gas Reservoir – One Case Study in Kekeya Block, Tuha Basin, China

Tight sand reservoir is usually characterized by high heterogeneity and complex pore structure, which makes the permeability calculation a big challenge and leads to difficulties in reservoir classification and productivity evaluation. First, five different Hydraulic Flow Units and respective Porosity-permeability relations were built based on core dataset from Kekeya block, Tuha Basin; and then with BP Neutron Network method, flow unit was classified for un-cored intervals using normalized logging data, and permeability was calculated accordingly. This improved the accuracy of permea-bility calculation and helped a lot on un-cored reservoir evaluation. In addition, based on porosity, permeability and flow unit type, a new reservoir grading chart was set up by incorporating the testing or production data, which provides important guidance for productivity prediction and reservoir development.


INTRODUCTION
Tight sand reservoir usually refers to sandstone with porosity less than 12%, absolute permeability less than 1 mD, water saturation more than 40% and pore throat radius less than 0.5μm [1].Tight sand reservoirs are characterized by complicated pore structure, different kinds of clay minerals, high capillary pressure and high water saturation, which affect reservoir permeability and productivity and bring big difficulties to reservoir grading evaluation using conventional logs.Porosity and permeability are key parameters for reservoir evaluation; porosity can be easily acquired by nuclear magnetic resonance tool or calibration with core dataset.In conventional reservoir, permeability has good correlation with porosity, which can be calculated easier.However, in tight sand reservoirs, beside the intergranular residual pores, pore spaces also include intragranular dissolution pores, as well as micro pores and micro fractures.These different spaces are connected by the fine throats in between [2], which makes the relationship between permeability and porosity no longer follow simple linear relation.In order to get more accurate permeability for reservoir grading, it is necessary to classify the reservoir into different types [3][4][5], and then set up the permeability-porosity relations respectively.
The study area is located in Tuha basin, northwestern of China.The target formation is tight sand reservoir of Jurassic tight sand reservoir, which is mainly debris-feldspar and feldspar-debris sandstone with average porosity of 6% and average permeability of 0.45 mD.In order to better evaluate the tight sand reservoir, the authors proposed an integrated method.Starting with core data analysis, the reservoir was classify d into five different types of Hydraulic Flow Units (HFUs), and for each unit, permeability calculation model can be regressed.For un-cored intervals, HFUs can be classified by using BP Neutron Network, and the permeability can then be calculated accordingly to respective models.Finally, reservoir grading chart was built up by combining porosity, permeability, hydraulic flow unit and production data.

HYDRAULIC FLOW UNIT FUNDAMENTALS
Hydraulic flow unit was first proposed by Amaefule et al. in 1993 [6].It was defined as reservoir unit with similar fluid flowing characteristics.Reservoir Quality Index (RQI) and Flow Zone Indicator (FZI) were also defined by Amaefule et al. according to modified Kozeny -Catmen formular, RQI was defined as:  Since permeability and porosity can be measured from core samples, z , RQI, FZI can be got according to above equations.In the RQIz dual logarithmic diagram, samples of similar FZI values will fall on one line with slope of 1 and intercept of ( ) . Samples on the same lines have similar pore-throat feature and belong to one HFU.

HYDRAULIC FLOW UNIT (HFU) BASED ON CORE DATA
FZI is the integrated response of rock minerals and porethroat features, and it is a unique parameter for each hydraulic unit, which was widely used in HFU classification [7,8].In this case, RQI and FZI were calculated based on coremeasured porosity and permeability of 221 samples from 8 wells.According to dual logarithmic diagram of RQIz , cumulative probability curve of FZI and FZI histogram [9], 5 types of hydraulic units can be identified, namely HU1, HU2, HU3, HU4 and HU5.
Dual logarithmic diagram of RQIz (Fig. 1): As mentioned before, samples of similar FZI values will fall on one line with slope of 1 and intercept of log FZI ( ).While samples of obviously different FZI values will fall on different lines which have the same slope but different intercepts.Cumulative probability curve of FZI (Fig. 3): Due to the presence of random measurement error of core analysis, the FZI values usually distribute around their true mean values.On the cumulative probability graph, samples of normal distribution should fall into a straight line.Each HFU has its own normal distribution, so different HFU should fall into straight lines of different slope.
In addition, characteristics of each HFU were summarized by incorporating thin section and mercury injection data.(1) HU1: FZI<0.33;few intergranular residual pores, some micropores in clay matrix, occasional intragranular dissolution pores, poor connectivity, displacement pressure is more than 0.1MPa.
(5) HU5: FZI>1.6;dissolutioin pores are not developed and displacement pressure is high (about 0.1 MPa), while open fractures are developed, which enhanced the reservoir permeability greatly.

PERMEABILITY CALCULATION BASED ON FLOW UNIT
Since each type HFU has similar fluid-flow characteristics, permeability and porosity show good correlation for each flow unit.Based on the HFU got from core data, the permeability calculation models were built up respectively (Fig. 4).

PERMEABILITY CALCULATION BASED ON LOG DATA
As we know, core data is limited because it's time consuming and expensive.While continuous log data is easier to get and it contains a lot of information.In this paper, we used BP (Back Propagation) neural network technique to classify the logs into different HFUs.
BP algorithm is called Error Back-Propagation algorithm.An interpretation model can be built up by learning and training on a given dataset.The basic idea is by back propagating network output error, it continues to adjust and modify the connection weights and thresholds of the network to minimize the errors.It has self-organizing, self-learning, adaptive, fault tolerance and anti-interference ability, and has been used to predict carbonate log facies and identify volcanic lithologies [10,11].Self checking was done for the training dataset which contains 221 samples, and the coincidence rate was above 85% (Fig. 5).According to BP neutral network model, HFU type of 19 wells were identified using log data, then permeability was calculated based on the permeability-porosity models set up from core dataset.Fig. (6) shows that the calculated permeability agrees well with the core measured permeability.

NEW RESERVOIR GRADING METHOD
On the basis of HFU classification and permeability calculation, we classified the reservoir into different grades by integrating the testing or production data, and built up a useful grading chart which can help to grade the reservoir productivity quickly and effectively.
First, cutoff of effective reservoir were analyzed by using Distribution Function method and Abandon method [12].Fig. (7) is the frequency distribution diagram of porosity, the black line represents samples of dry interval, and the red line represents the samples of producing interval, the cross point of the two curves will be the cutoff according to Distribution Function method, which is about 4.7%.8) is the porosity histogram of testing intervals, and it is believed that small pores which occupy about 20% of the pore spaces contribute very little to the production which can be discarded.According to the two methods, the porosity cutoff of the first type of HFU is about 4.5% and 3.0% for the fifth type of HFU.

Fig. (8). Frequency distribution histogram of porosity within testing intervals.
Draw a vertical line where porosity equal to 4.5% and 3.0% on the cross plots of permeability and porosity for each type of HFUs, you can get the permeability cutoffs corresponding to different types of HFUs.And by connecting the intersecting points together you can get the continuous cutoff line of effective reservoir, which is the black oblique line on the left bottom (Fig. 9).If the porosity and permeability fall into the area below the black line, it will be difficult to produce any gas even after stimulation, which can be classified into Type III.And then, by normalizing the production data, we can find the line corresponding to economic production that is the red oblique line on the right top.If the porosity and permeability falls into the area above the red line, it would be easier to produce economically, which can be classified into Type I.If the points fall between the red and black oblique line, the reservoir can produce some gas, but it still difficult to break even, which is classified into Type II.Fig. (10) shows the grading result of Well A, from the result we can see that, the porosity of the testing interval is higher than 7%, but the permeability is relatively low, and so most intervals are classified into the first and second type of HFUs.However, accordingly to grading chart, the porosity and permeability of this interval fall into to Type I area, which means that this interval can produce economically after stimulation.The testing result shows that, after acidizing, this interval can produce 53682m 3 of gas and 19.76m 3 of oil per day, which matches our grading result very well.Fig. (10).Grading result of one well.

z
is the ratio of pore volume and grain volume,

Fig. ( 1
Fig. (1).Dual logarithmic diagram of RQIz .FZI histogram (Fig. 2): Since FZI distribution is the superposition of multiple lognormal distributions, so FZI histogram can show normal distribution of several flow units and provide them with the corresponding FZI value.However, it is difficult to distinguish single HFU from the overlap region of the histogram.
Using logging data to identify HFU mainly includes four steps: (1) Match the core depth to the log depth, analyze the correlation between FZI and different logging curves and determine the logs sensitive to flow unit type; (2) Normalize the sensitive log curves; (3) Establish the training dataset by incorporating the normalized log data and the flow unit type, get the interpretation model between the log curves and the flow unit type by training the dataset using BP neutral network technique; (4) Apply the interpretation model the logs of un-cored intervals or wells to identify the HFU type.

Fig. (
Fig.(8) is the porosity histogram of testing intervals, and it is believed that small pores which occupy about 20% of the pore spaces contribute very little to the production which can be discarded.According to the two methods, the porosity cutoff of the first type of HFU is about 4.5% and 3.0% for the fifth type of HFU.