Well kick monitoring method in drilling process based on geological engineering big data analysis

Abstract

The invention relates to the technical field of petroleum engineering well kick monitoring methods, in particular to a geological engineering big data analysis-based well kick monitoring method in a drilling process, which comprises the following steps of: establishing a well kick risk standard data set, training the data set to obtain a well kick identification model, and judging well kick working conditions of a real drilled well by using the model. According to the method, the median method is adopted to carry out noise reduction on the logging parameter data, the data quality is high, the method adapts to different noise data and does not depend on high-precision measurement, the relative variation of the logging data is captured, the influence of complex geological structures of wells in different blocks is reduced, and universality is high; the method does not depend on a threshold value of a single parameter, and a calculation result is output in a 0-1 form; abnormal changes of well logging parameters in the well kick development process can be captured, and accurate well kick prediction is achieved; meanwhile, real-time data matching is achieved, and overflow is found as early as possible.

Description

Technical Field

[0001] The invention relates to the technical field of well kick monitoring methods in petroleum engineering, and is a well kick monitoring method in a drilling process based on geological engineering big data analysis. Background Art

[0002] In the process of drilling for oil and gas field development, drilling safety is the primary factor in ensuring economic benefits and personnel safety. The core of ensuring safe drilling is to do a good job of well blowout monitoring. If a well blowout is not discovered in time, it may quickly develop into a blowout. Among many drilling accidents, blowouts are often the most dangerous. Blowouts can cause: (1) casualties of drilling personnel and huge property losses; (2) environmental pollution and affect the production of surrounding agriculture and animal husbandry; (3) toxic gas leakage, stratum collapse and fire caused by blowouts will threaten the lives and safety of residents near the accident well, causing extremely bad social impacts. It has also caused immeasurable losses to the local natural and ecological environment, and the subsequent cleanup work has taken a lot of time, manpower and material resources and lasted for several months. It can be seen that well blowout monitoring is of vital importance.

[0003] (1) At present, scholars at home and abroad have proposed many monitoring methods based on bottom hole pressure measurement and high-precision drilling fluid outlet flow monitoring equipment at the wellhead. However, due to the limitations of formation temperature, pressure and drilling costs, many methods and equipment are not suitable for onshore deep and ultra-deep well drilling sites in the vast western region of my country.

[0004] (2) At present, most of the methods adopted in on-site well blowout monitoring are still based on traditional methods such as mud pool increment and after-effect gas. Such methods have a certain lag, and due to the high noise characteristics of on-site monitoring parameters, traditional monitoring methods are difficult to strike a balance between false alarm rate and accuracy: when the threshold is set high, the false alarm rate is low but missed alarms are prone to occur; when the threshold is set low, the missed alarm rate is low but false alarms are prone to occur. Summary of the invention

[0005] The present invention provides a method for monitoring well kicks during drilling based on geological engineering big data analysis, which overcomes the shortcomings of the above-mentioned prior art and can effectively solve the problem that traditional well kick monitoring methods are mostly based on field experts relying on experience to monitor logging data curves, with high monitoring costs, and the main parameter monitored is the mud pool increment, and this method has a strong hysteresis.

[0006] The technical solution of the present invention is achieved through the following measures: A method for monitoring well kicks during drilling based on geological engineering big data analysis, comprising the following steps:

[0007] Step 1, establish a historical logging parameter set, which includes 17 parameters, namely time, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure, casing pressure, inlet flow, outlet flow, drilling pressure, mechanical drilling speed, pump stroke, inlet density, outlet density, total pool volume, late time, and late well depth;

[0008] Step 2, preprocess the historical logging parameter set to obtain a standard data set, which includes 12 parameters, namely, time interval, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure change, casing pressure, inlet and outlet flow difference, mechanical drilling speed, inlet density, outlet density and total pool volume change; based on the well kick identification of late time and late well depth, the well kick index KD in the actual time and well depth section corresponding to the late time and late well depth when the well kick occurred in the historical logging data is identified to obtain the well kick index KD; the standard data set and the well kick index KD constitute the well kick risk standard data set;

[0009] Step 3, using the well kick risk standard data set as input, using the existing conventional machine learning algorithm for training, and obtaining a well kick identification model that is comprehensively applicable to well kick risks;

[0010] Step 4, collect the actual drilling logging parameter set, the actual drilling logging parameter set includes 17 parameters, namely time, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure, casing pressure, inlet flow, outlet flow, drilling pressure, mechanical drilling speed, pump stroke, inlet density, outlet density, total pool volume, late time, late well depth; perform data preprocessing on the actual drilling logging parameter set to obtain the actual drilling standard data set, the actual drilling standard data set includes 12 parameters, namely time interval, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure change, casing pressure, inlet and outlet flow difference, mechanical drilling speed, inlet density, outlet density and total pool volume change; at the same time, identify the drilling conditions;

[0011] Step 5: After the drilling condition is determined, the condition is input into the kick identification model. After calculation, the kick identification model outputs a classification result in the form of 0 or 1, where 0 represents normal drilling and 1 represents a kick.

[0012] The following are further optimizations and / or improvements to the above technical solutions:

[0013] Furthermore, the kick is marked according to the historical logging data, and the kick index KD in the segment of the late time and the late well depth corresponding to the kick and the actual time and depth is marked as 1, and the other segments are marked as 0.

[0014] Furthermore, the working pump pressure is obtained by multiplying the median of the ratios of all pump pressure values ​​to well depths within 5 minutes after each pump is started until stabilization by the well depth corresponding to the current time. The pump pressure change is equal to the pump pressure at the current time minus the working pump pressure. The working pump pressure is recalibrated each time the pump is started.

[0015] Furthermore, the inlet and outlet flow difference is obtained by subtracting the inlet flow from the outlet flow.

[0016] Furthermore, when the unit of the above-mentioned outlet flow is a percentage, the outlet flow is obtained by dividing the inlet flow at the current moment by the average value of the previous 20 groups of outlet flows and multiplying the outlet flow at the current moment.

[0017] Furthermore, the rated total pool volume is obtained by multiplying the median of the ratio of all total pool volume data in the previous 5 minutes to the corresponding well depth by the median method and the well depth at the current moment. The change in the total pool volume is equal to the total pool volume at the current time minus the rated total pool volume.

[0018] The present invention adopts the median method to reduce the noise of logging parameter data, has high data quality, is adaptable to different noise data, does not rely on high-precision measurement, but captures the relative change of logging data, reduces the influence of complex geological structures of wells in different blocks, and has strong universality; does not rely on the threshold of a single parameter, and the calculation result is output in the form of 0-1; can capture the abnormal changes of logging parameters in the development process of well kick, and realize accurate prediction of well kick; at the same time, real-time data matching is carried out to detect overflow as early as possible. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Attached Figure 1 This is one of the flow charts of the method for monitoring well kick during drilling based on geological engineering big data analysis described in the present invention.

[0020] Attached Figure 2 This is the second flow chart of the method for well kick monitoring during drilling based on geological engineering big data analysis described in the present invention.

[0021] Attached Figure 3 This is one of the comparison curves of different parameters in the stable state and when a kick occurs.

[0022] Attached Figure 4 This is the second comparison curve of different parameters in stable state and when well kick occurs.

[0023] Attached Figure 5 It is the response characteristic curve of different monitoring parameters when well kick occurs.

[0024] Attached Figure 6 It is a schematic diagram of the principle of the well kick identification method based on the late time and the late well depth of the present invention. DETAILED DESCRIPTION

[0025] The present invention is not limited by the following embodiments, and specific implementation methods can be determined based on the technical solution of the present invention and actual conditions.

[0026] The present invention will be further described below in conjunction with embodiments:

[0027] Embodiment 1: As attached Figures 1 to 2 As shown, the well kick monitoring method in the drilling process based on geological engineering big data analysis includes the following steps:

[0028] Step 1, establish a historical logging parameter set, which includes 17 parameters, namely time, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure, casing pressure, inlet flow, outlet flow, drilling pressure, mechanical drilling speed, pump stroke, inlet density, outlet density, total pool volume, late time, and late well depth;

[0029] Step 2, preprocess the historical logging parameter set to obtain a standard data set, which includes 12 parameters, namely, time interval, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure change, casing pressure, inlet and outlet flow difference, mechanical drilling speed, inlet density, outlet density and total pool volume change; based on the well kick identification of late time and late well depth, the well kick index KD in the actual time and well depth section corresponding to the late time and late well depth when the well kick occurred in the historical logging data is identified to obtain the well kick index KD; the standard data set and the well kick index KD constitute the well kick risk standard data set;

[0030] Combined with the literature research and the experience of the field experts of China National Petroleum Corporation Western Drilling and Geology Research Institute (Logging Engineering Branch), refer to Figures 3 to 5 Finally, 17 parameters were selected to form the historical logging parameter set of the present invention. Under normal drilling conditions, the inlet and outlet flow difference, pump pressure, casing pressure, and mechanical drilling speed tend to be stable. As the well depth increases, the total pool volume will increase, and the drilling fluid in the drill pipe and annulus will increase, resulting in a continuous decrease in the total pool volume under normal drilling conditions. After a well kick occurs, due to the invasion of formation fluid, the circulating pressure loss increases, the pump pressure increases, and the mechanical drilling speed is affected to a certain extent. At the same time, the outlet flow increases, which in turn increases the inlet and outlet flow difference and the total pool volume. Due to the complex geological structure of wells in different blocks, it is necessary to calculate the relative changes in the above logging parameters to eliminate the influence of unknown formation information.

[0031] Combined with reference Figure 6 The traditional method of monitoring well kick based on the volume expansion principle is to wait until the well kick develops to a certain extent, which has a time lag, causing most of the drilling fluid in the deepwater wellbore to be contaminated, which is not conducive to subsequent well control operations. The mud pool increment and outlet flow in the logging data are the formation invasion volume Q after the well kick occurs. fThe data circulated to the mud pool for monitoring is actually the corresponding value half a cycle ago, while the pump pressure and mechanical drilling speed will change significantly directly after the formation fluid invades.

[0032] Step 3, using the well kick risk standard data set as input, using the existing conventional machine learning algorithm for training, and obtaining a well kick identification model that is comprehensively applicable to well kick risks;

[0033] Step 4, collect the actual drilling logging parameter set, the actual drilling logging parameter set includes 17 parameters, namely time, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure, casing pressure, inlet flow, outlet flow, drilling pressure, mechanical drilling speed, pump stroke, inlet density, outlet density, total pool volume, late time, late well depth; perform data preprocessing on the actual drilling logging parameter set to obtain the actual drilling standard data set, the actual drilling standard data set includes 12 parameters, namely time interval, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure change, casing pressure, inlet and outlet flow difference, mechanical drilling speed, inlet density, outlet density and total pool volume change; at the same time, identify the drilling conditions;

[0034] Step 5: After the drilling condition is determined, the condition is input into the kick identification model. After calculation, the kick identification model outputs a classification result in the form of 0 or 1, where 0 represents normal drilling and 1 represents a kick.

[0035] Existing conventional machine learning algorithms are not limited to random forests, XGBOOST, and artificial neural networks. During the training process, existing parameter optimization algorithms are used, which are not limited to genetic algorithms, particle swarm optimization algorithms, etc.

[0036] Data preprocessing is mainly to obtain the time interval, pump pressure change, inlet and outlet flow difference and total pool volume change.

[0037] Example 2: As an optimization of the above example, in step 2, the well kick is marked according to the historical logging data, and the well kick index KD within the late time and late well depth corresponding to the well kick and the actual time and depth segment is marked as 1, and other segments are marked as 0.

[0038] Example 3: As an optimization of the above example, the working pump pressure is calculated by multiplying the median of the ratio of all pump pressure values ​​to well depth within 5 minutes after each pump is started until it stabilizes by the well depth corresponding to the current time. The pump pressure change is equal to the pump pressure at the current time minus the working pump pressure. The working pump pressure is recalibrated each time the pump is started.

[0039] Embodiment 4: As an optimization of the above embodiment, the difference between the inlet and outlet flow rates is obtained by subtracting the outlet flow rate from the inlet flow rate.

[0040] Embodiment 5: As an optimization of the above embodiment, when the outlet flow rate unit is a percentage, the outlet flow rate is obtained by dividing the inlet flow rate at the current moment by the average value of the previous 20 groups of outlet flows multiplied by the outlet flow rate at the current moment.

[0041] Example 6: As an optimization of the above example, the rated total pool volume is obtained by multiplying the median of the ratio of all total pool volume data in the previous 5 minutes to the corresponding well depth after noise reduction by the median method by the well depth at the current moment. The change in the total pool volume is equal to the total pool volume at the current time minus the rated total pool volume.

[0042] The data preprocessing method is as follows:

[0043] (1) For the time interval ΔT, due to the inconsistency of real-time data density, the time interval needs to be calculated. The calculation formula is as follows:

[0044] ΔT=timestamp(T 当前数据点时间 )-timestamp(T 前一个数据点时间 )

[0045] (2) For the pump pressure change ΔSPP, we first need to calculate the working pump pressure SPP through the pump pressure 工作 , and then obtain the following calculation formula:

[0046] SPP 工作 =median(SPP 5min / DEPTH 5min )·DEPTH 当前时间

[0047] ΔSPP=SPP 当前时间 -SPP 工作

[0048] Where SPP 工作 represents the working pump pressure, ΔSPP represents the change in pump pressure, SPP 当前时间 Represents the pump pressure at the current time.

[0049] (3) For the inlet and outlet flow difference parameter, since some wells in the western drilling project use baffle flow meters, the outlet flow measured is a percentage, and the outlet flow needs to be converted. The calculation formula is as follows:

[0050]

[0051] Where Q represents the inlet and outlet flow difference, MFI 当前时刻 Represents the inlet flow at the current moment, FLOWOUT 前20组 Represents the first 20 groups of export traffic, FLOWOUT 当前时刻 The current egress traffic.

[0052] (4) For the total pool volume change ΔTPIT, due to the large fluctuation of the total pool volume, the median method medfilt is first used for noise reduction. At the same time, the loss of drilling fluid under normal drilling conditions needs to be eliminated, and the rated total pool volume TPIT needs to be calculated. 额定 , and then calculate the total pool volume change, the calculation formula is as follows:

[0053] TPIT 5min = medfilt(TPIT 5min )

[0054] TPIT 额定 =median(TPIT 5min / DEPTH 5min )·DEPTH 当前时间

[0055] ΔTPIT=TPIT 当前时间 -TPIT 额定

[0056] Where ΔTPIT represents the change in total pool volume, TPIT 当前时间 Represents the total pool volume at the current time, TPIT 额定 Represents the rated total pool volume.

[0057] The following conditions must be met to determine the drilling condition:

[0058] (1)WOB≠0

[0059] (2) SPM>0

[0060] (3)abs(DEPTH 当前时间 -BITDEP 当前时间 )≤0.05

[0061] (4) abs(SPM 当前时间 -mean(SPM 前20组 ))≤5

[0062] WOB stands for weight on bit, SPM stands for pump speed, and BITDEP stands for bit depth.

[0063] In this embodiment, the timestamp function in the data preprocessing method converts the time format in the Python language into a timestamp in seconds.

[0064] In this implementation, the median function in the data preprocessing method is to take the median of an array in the Python language.

[0065] In this embodiment, the mean function in the working condition determination method is to take the average value of the array.

[0066] In this embodiment, the abs function in the working condition determination method takes the absolute value of a value.

[0067] In this embodiment, the medfilt function in the data preprocessing method is a commonly used median method noise reduction processing function in the Python language.

[0068] In this embodiment, the working pump pressure SPP in the data preprocessing 工作 , rated total pool volume TPIT 额定 It needs to be recalculated every time the pump is turned on.

[0069] The above technical features constitute the embodiments of the present invention, which have strong adaptability and implementation effect. Non-essential technical features can be added or reduced according to actual needs to meet the requirements of different situations.

Claims

1. A method for monitoring well kicks during drilling based on geological engineering big data analysis, characterized in that The following steps are involved: Step 1, establish a historical logging parameter set, which includes 17 parameters, namely time, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure, casing pressure, inlet flow, outlet flow, drilling pressure, mechanical drilling speed, pump stroke, inlet density, outlet density, total pool volume, late time, and late well depth; Step 2, preprocess the historical logging parameter set to obtain a standard data set, which includes 12 parameters, namely, time interval, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure change, casing pressure, inlet and outlet flow difference, mechanical drilling speed, inlet density, outlet density and total pool volume change; based on the well kick identification of late time and late well depth, the well kick index KD in the section corresponding to the late time and late well depth when the well kick occurred in the historical logging data is identified according to the actual time and well depth, and the well kick index KD is obtained; The standard data set and the kick index KD constitute the kick risk standard data set; Step 3, using the well kick risk standard data set as input, using the existing conventional machine learning algorithm for training, and obtaining a well kick identification model that is comprehensively applicable to well kick risks; Step 4, collect the actual drilling logging parameter set, the actual drilling logging parameter set includes 17 parameters, namely time, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure, casing pressure, inlet flow, outlet flow, drilling pressure, mechanical drilling speed, pump stroke, inlet density, outlet density, total pool volume, late time, late well depth; perform data preprocessing on the actual drilling logging parameter set to obtain the actual drilling standard data set, the actual drilling standard data set includes 12 parameters, namely time interval, well depth, vertical depth, drill bit depth, drill bit vertical depth, pump pressure change, casing pressure, inlet and outlet flow difference, mechanical drilling speed, inlet density, outlet density and total pool volume change; at the same time, identify the drilling conditions; Step 5: After the drilling condition is determined, the condition is input into the kick identification model. After calculation, the kick identification model outputs a classification result in the form of 0 or 1, where 0 represents normal drilling and 1 represents a kick.

2. The method for monitoring well kicks during drilling based on geological engineering big data analysis according to claim 1, characterized in that The kick is marked according to the historical logging data. The kick index KD in the section of the late time and late well depth corresponding to the kick and the actual time and depth is marked as 1, and the other sections are marked as 0.

3. The method for monitoring well kicks during drilling based on geological engineering big data analysis according to claim 1 or 2, characterized in that The working pump pressure is the median of the ratio of all pump pressure values ​​to well depth within 5 minutes after each pump start until stabilization multiplied by the well depth corresponding to the current time. The pump pressure change is equal to the pump pressure at the current time minus the working pump pressure.

4. The method for monitoring well kicks during drilling based on geological engineering big data analysis according to claim 1 or 2, characterized in that The difference between inlet and outlet flow is the outlet flow minus the inlet flow.

5. The method for monitoring well kicks during drilling based on geological engineering big data analysis according to claim 3, characterized in that The difference between inlet and outlet flow is the outlet flow minus the inlet flow.

6. The method for monitoring well kicks during drilling based on geological engineering big data analysis according to claim 4, characterized in that When the outlet flow rate unit is percentage, the outlet flow rate is the current inlet flow rate divided by the average of the previous 20 groups of outlet flow rates multiplied by the current outlet flow rate.

7. The method for monitoring well kicks during drilling based on geological engineering big data analysis according to claim 5, characterized in that When the outlet flow rate unit is percentage, the outlet flow rate is the current inlet flow rate divided by the average of the previous 20 groups of outlet flow rates multiplied by the current outlet flow rate.

8. The method for monitoring well kicks during drilling based on geological engineering big data analysis according to claim 1, 2, 5, 6 or 7, characterized in that The rated total pool volume is the median of the ratio of all total pool volume data in the previous 5 minutes to the corresponding well depth after noise reduction by the median method multiplied by the well depth at the current moment. The change in total pool volume is equal to the total pool volume at the current time minus the rated total pool volume.

9. The method for monitoring well kicks during drilling based on geological engineering big data analysis according to claim 3, characterized in that The rated total pool volume is the median of the ratio of all total pool volume data in the previous 5 minutes to the corresponding well depth after noise reduction by the median method multiplied by the well depth at the current moment. The change in total pool volume is equal to the total pool volume at the current time minus the rated total pool volume.

10. The method for monitoring well kicks during drilling based on geological engineering big data analysis according to claim 4 or 8, characterized in that The rated total pool volume is the median of the ratio of all total pool volume data in the previous 5 minutes to the corresponding well depth after noise reduction by the median method multiplied by the well depth at the current moment. The change in total pool volume is equal to the total pool volume at the current time minus the rated total pool volume.

Images