Gas well operation machine operation optimization management system based on multi-parameter automatic regulation

Abstract

This invention discloses a gas well workover rig operation optimization management system based on multi-parameter automated adjustment, relating to the field of gas well workover rig operation optimization. It primarily targets the operation optimization management of gas well workover rigs, which commonly employ a single-objective control strategy of "maximum production priority." This strategy implicitly assumes a linear relationship between production and energy consumption and a perpetually positive marginal benefit. This invention achieves automated optimization and safety management of the entire gas well workover rig process across multiple dimensions. It abandons the pitfalls of traditional "maximum production priority" single-objective control by accurately identifying the dynamic relationship between production and energy consumption, analyzing hysteresis effects during operation in real time, and integrating environmental and equipment status to achieve dynamic safety early warning. It constructs a full-process management system of "data acquisition - analysis and identification - decision optimization - closed-loop update," effectively solving core problems such as low energy efficiency, execution deviations, and difficulty in predicting safety hazards during gas well workover rig operation.

Description

Technical Field

[0001] This invention relates to the field of gas well workover rig operation optimization technology, specifically a gas well workover rig operation optimization management system based on multi-parameter automated adjustment. Background Technology

[0002] Currently, oil and gas exploration and development is moving towards ultra-deep water, ultra-deep formations, low permeability, and unconventional methods. The operating environment of gas wells is becoming increasingly complex, and the operating efficiency, energy consumption control, and safety management of gas well workover rigs have become core pain points in the industry. As gas field development enters the middle and late stages, a large number of gas wells are entering the decline period or high water cut stage, and problems such as gas slippage loss, increased liquid back pressure, and decreased pump efficiency are becoming increasingly prominent, which puts forward higher requirements for the operation optimization of gas well workover rigs.

[0003] Currently, the optimization management of gas well workover rigs generally adopts a single-objective control strategy of "maximum production priority". By adjusting the stroke frequency through frequency conversion, the real-time production is made as close as possible to the historical maximum value or the geological production limit. This strategy implicitly assumes that production and energy consumption are linearly proportional and that marginal benefits are always positive. It ignores the diminishing marginal benefits of the production-energy consumption curve under complex gas well conditions. As a result, after the stroke frequency exceeds the critical point, energy consumption increases while the production increment approaches zero or even becomes negative. This not only causes energy waste and equipment wear, but may also cause water lock in the gas reservoir due to excessive liquid drainage, affecting the long-term production capacity of the gas well.

[0004] At the operational level, the valve control of gas well workover rigs exhibits a significant hysteresis effect. Traditional technologies lack effective mechanisms for identifying and compensating for this hysteresis effect. When the hysteresis effect intensifies, it can lead to a decrease in valve control accuracy, disruption of the operational process, and even affect operational safety.

[0005] In terms of safety early warning, conventional systems only focus on the equipment's own operating parameters and do not fully integrate external factors such as meteorological environment and nearby interference. They cannot achieve linkage early warning between equipment status and external environment, and it is difficult to predict in advance equipment structural damage or operational failure caused by external interference such as sudden wind speed changes and ground vibration, which poses serious safety hazards.

[0006] In addition, conventional systems suffer from insufficient data acquisition accuracy, inaccurate operating condition identification, and a lack of dynamic adaptability in decision-making. Furthermore, each unit operates independently with poor coordination, making it impossible to form a closed-loop optimization mechanism. This makes it difficult to adapt to the dynamic decline in gas well production capacity, thus hindering the improvement of gas well workover rig operating efficiency and the control of operating costs.

[0007] To address the aforementioned technical shortcomings, a gas well workover rig operation optimization and management system based on multi-parameter automated adjustment is proposed. This system meets the requirements of accurate data acquisition, precise operating condition identification, and dynamic decision-making adaptability, forming a closed-loop optimization mechanism. It efficiently adapts to the dynamic decline in gas well production capacity, thereby improving the operating efficiency of the gas well workover rig and controlling operating costs. Summary of the Invention

[0008] The purpose of this invention is to solve the problems mentioned above by proposing a gas well workover rig operation optimization management system based on multi-parameter automated adjustment.

[0009] The objective of this invention can be achieved through the following technical solution: a gas well workover rig operation optimization and management system based on multi-parameter automated adjustment, comprising an operation optimization management platform, wherein the operation optimization management platform is connected to the following communication links:

[0010] The production and energy consumption relationship identification unit identifies the production and energy consumption relationship of the gas well workover machine and obtains the stable operating points of the gas well workover machine under different strokes. Each operating point includes at least stroke, production and power consumption data.

[0011] A two-dimensional Pareto frontier based on historical operating conditions is dynamically constructed, and the marginal efficiency gradient of the frontier is calculated. A dynamic break-even threshold is generated based on the changing trends of real-time electricity and gas prices and the marginal efficiency gradient of the frontier. The actual marginal efficiency corresponding to the current production increase instruction is calculated, and the actual marginal efficiency is corrected by incorporating production fluctuation entropy. The corrected actual marginal efficiency is compared with the dynamic break-even threshold: if it is greater than the threshold and the number of strokes has not reached the upper limit, incremental production is executed; if it is less than or equal to the threshold, production increase is rejected and exploratory frequency reduction is actively implemented. The Pareto frontier is updated based on the new operating conditions after exploratory frequency reduction, and subsequent dynamic thresholds are recursively adjusted.

[0012] The job execution analysis unit performs execution analysis on the optimization object and optimizes it based on the analysis results; the dynamic safety early warning unit integrates meteorological and environmental vibration data to provide dynamic safety early warnings for the operation of the optimization object.

[0013] Furthermore, the process of generating the dynamic break-even threshold further includes:

[0014] The basic break-even marginal efficiency is calculated based on real-time electricity and gas prices.

[0015] Extract the local marginal efficiency decline acceleration of the Pareto front at the current operating condition projection point and compare it with the historical average decline rate to construct a dynamic penalty factor;

[0016] Multiplying the basic break-even marginal efficiency by the dynamic penalty factor yields the dynamic threshold.

[0017] Furthermore, the calculation and correction process for output fluctuation entropy further includes:

[0018] Collect minute-level production sequences within a set time period in the past, calculate their sample entropy as the production fluctuation entropy, which represents the randomness of production fluctuation; when constructing the cost function for comparison, subtract the dynamic break-even threshold from the actual marginal efficiency, and then subtract the product of the production fluctuation entropy and the preset weight coefficient.

[0019] Furthermore, the exploratory frequency reduction and frontier update process further includes:

[0020] When the production increase order is rejected, the stroke rate is actively reduced by a set exploration step size, and the process waits for at least one stable window duration; the new stable operating point after the frequency reduction is collected, added to the historical operating point set, and the Pareto front is reconstructed.

[0021] Determine if the new operating point is located on the updated Pareto front: if so, mark this frequency reduction as a valid exploration and update the historical average marginal efficiency decline rate; if multiple frequency reductions do not produce a new front point, determine that the system is in the optimal operating zone and stop the exploration.

[0022] Furthermore, before constructing the Pareto front, the extraction and weight correction of local energy efficiency curvature features at the operating point are also included:

[0023] For each historical operating point, the local energy efficiency curvature is calculated based on the second difference between the "stroke-output" plane and the "stroke-power consumption" plane. A positive value of the curvature indicates the region of increasing marginal returns, and a negative value indicates the region of decreasing marginal returns. When constructing the Pareto front, an additional decay factor is applied to operating points with negative curvature and absolute values ​​exceeding a set threshold, in addition to the time-domain decay weight, to reduce their contribution to the front.

[0024] Meanwhile, the strength of the additional decay factor is dynamically adjusted based on the rate of change of the marginal efficiency gradient of adjacent points on the frontier, so that the frontier is not contaminated by inefficient historical operating conditions.

[0025] Furthermore, the execution analysis process of the optimization object in the job execution analysis unit is as follows:

[0026] Extract the valve command sequence corresponding data as valve opening degree, construct the valve command sequence according to time order, and record the feedback time according to the feedback value of the pressure sensor after the valve command is initiated, and construct the pressure sensor feedback sequence according to time order.

[0027] Based on the comparison of adjacent moments in the valve command sequence, the time axis is divided into rising segment, falling segment, and stationary segment; with the valve command sequence as the horizontal axis and the pressure sensor feedback sequence as the vertical axis, the command pressure trajectory is constructed in the non-stationary segment, forming a hysteresis loop.

[0028] Furthermore, data collection and analysis are performed on the hysteresis loop to extract the loop area, which is represented by the area enclosed between the rising curve and the falling curve, and is obtained by integration using the trapezoidal rule.

[0029] If the area of ​​the hysteresis loop continues to increase or exceeds the set red line area threshold, based on the continuous increase in the duration of the operation phase, it is inferred that the optimization object is in the hysteresis effect stage, and the current stage is marked as the severe hysteresis effect stage; conversely, if the area of ​​the hysteresis loop does not continue to increase and does not exceed the set red line area threshold, it is inferred that the optimization object is not in the hysteresis effect stage, and the current stage is marked as the stable hysteresis effect stage.

[0030] Furthermore, during the stable phase of the hysteresis effect, the valve command sequence is executed according to the set rules. However, after entering the severe phase of the hysteresis effect, the period interval of the valve command sequence is increased, and the operation of the optimized object is coordinated simultaneously. The interval of the operation of the optimized object is aligned with the period interval of the valve command sequence, that is, the period interval is lower than the set interval threshold.

[0031] During the severe hysteresis phase, when the optimized objects are running together, the average slope of the rising segment in the middle of the loop and the average slope of the falling segment in the same instruction interval are obtained based on the hysteresis loop. The local slope asymmetry is calculated based on the ratio of the average slopes. If the local slope asymmetry exceeds the set local slope asymmetry threshold, an asymmetry compensation signal is generated and sent to the operation optimization management platform; otherwise, a continuous monitoring signal is generated.

[0032] Furthermore, the dynamic safety early warning process within the dynamic safety early warning unit is as follows:

[0033] Based on the operational phase of the optimization object, determine the environmental parameters and neighboring interference parameters within the operational phase; simultaneously, collect data on the optimization object.

[0034] The current time is used to obtain the completed runtime segment, and the frequency of environmental parameters exceeding the range of parameters of the completed runtime segment during the runtime phase is obtained. At the same time, the duration of the neighboring interference parameter values ​​exceeding the set parameter threshold is collected.

[0035] The current stage is divided into a high-interference stage and a low-interference stage based on threshold comparison.

[0036] Furthermore, the changes in structural response perception data of the optimized object during the operation phase are obtained, as well as the cumulative amount of time when the corresponding type of operation status perception data is outside the set range.

[0037] Based on threshold comparison, the current stage is divided into a state change stage and a state stability stage. When the state change stage coincides with the high interference stage, the current optimization direction is set to optimize the operating environment of the object and control the surrounding interference. When the state change stage coincides with the low interference stage, the current optimization direction is set to optimize the operating state of the object itself, that is, to inspect and maintain the hardware.

[0038] Compared with the prior art, the beneficial effects of the present invention are:

[0039] 1. Through full-process data acquisition and preprocessing, high-precision time alignment of three parameters—stroke rate, output, and power consumption—and accurate division of steady-state operating window are achieved. Transient disturbances and invalid data are eliminated, providing reliable data support for energy efficiency analysis. This solves the problems of insufficient data acquisition accuracy and difficulty in extracting steady-state data in traditional technologies. Furthermore, through the mining of local energy efficiency curvature features, the marginal benefit change law of the output-energy consumption curve is accurately identified, the "efficiency inflection point" is clarified, and the erroneous assumption of "linear proportionality between output and energy consumption" in traditional technologies is broken, avoiding the investment of computing resources in the negative marginal benefit region.

[0040] The dynamic Pareto front construction and feedback correction mechanism, combined with time-domain decay weighting, prioritizes the impact of recent operating conditions and attenuates the weight of historical invalid operating conditions through energy efficiency curvature, avoiding the frontier being "contaminated" by erroneous operating conditions, ensuring the real-time performance and accuracy of frontier construction. Furthermore, the dynamic break-even threshold integrates economic parameters and local frontier curvature, and can dynamically adjust the threshold according to gas well production capacity decline, market price fluctuations, etc., to adapt to different operating condition changes.

[0041] Real-time marginal efficiency calculation and multi-objective decision-making, combined with production fluctuation entropy, take into account energy efficiency, production stability and equipment limits, avoid blindly increasing production or reducing frequency, and achieve the optimal balance between production and energy consumption. It can significantly reduce energy consumption and improve marginal returns during the gas well decline period and high water cut stage. At the same time, through abnormal event identification, it can trigger diagnostic alarms in a timely manner to avoid the impact of misadjustment on gas well production capacity and extend the service life of gas wells.

[0042] 2. Achieve dynamic monitoring and closed-loop optimization of hysteresis effect, adjust execution strategy in real time according to changes in hysteresis effect, ensure coordination between valve control and operation, avoid operational conflicts, improve operation efficiency, and also detect valve operation abnormalities in advance through continuous monitoring and hierarchical early warning, provide a basis for equipment maintenance, reduce downtime due to failure, reduce operation and maintenance costs, adapt to various scenarios such as long-term valve operation and complex operation, and ensure the smoothness and reliability of operation process.

[0043] 3. By linking and analyzing environmental parameters, nearby interference parameters, equipment structural response, and operational status data, we can accurately distinguish between high / low interference stages and status change / stable stages, identify the root causes of abnormal equipment status, avoid blind early warnings and maintenance, and improve the pertinence and accuracy of safety early warnings. When the status change is consistent with the high interference stage, we can accurately control the surrounding interference; when it is consistent with the low interference stage, we can carry out targeted hardware inspection and maintenance. This achieves a precise match between safety early warnings and optimization directions, effectively preventing safety accidents caused by external interference or equipment malfunctions. Attached Figure Description

[0044] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0045] Figure 1 This is a system principle block diagram of the present invention;

[0046] Figure 2 This is a flowchart of the job execution analysis unit in this invention. Detailed Implementation

[0047] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0049] Please see Figures 1-2 As shown, the gas well workover rig operation optimization management system based on multi-parameter automated adjustment includes an operation optimization management platform, which is connected to a production and energy consumption relationship identification unit, an operation execution analysis unit, and a dynamic safety early warning unit.

[0050] Currently, the operation optimization management methods for gas well workover rigs generally adopt a single-objective control strategy of "maximum production priority"—adjusting the stroke rate through frequency conversion to make the real-time production as close as possible to the historical maximum or the upper limit of geological production allocation. The above strategy implies a flawed economic assumption—that production and energy consumption are linearly proportional and that marginal returns are always positive. However, during the decline phase of a gas well or in the high water-cut stage, due to factors such as gas slippage losses, increased liquid back pressure, and decreased pump efficiency, the production-energy consumption curve exhibits a typical law of diminishing marginal returns.

[0051] Once the number of strokes exceeds a certain critical point, the increase in production that can be brought about by each 1% increase in energy consumption approaches zero or even turns into a negative value (excessive liquid discharge leads to water lock in the gas reservoir). Traditional technologies have no mechanism to identify this "efficiency inflection point", and computing resources are continuously invested in the region of negative marginal returns.

[0052] The operation optimization management platform generates a production energy consumption relationship identification signal and sends it to the production energy consumption relationship identification unit. After receiving the production energy consumption relationship identification signal, the production energy consumption relationship identification unit identifies the production energy consumption relationship of the gas well workover machine.

[0053] Phase 1 - Data Acquisition and Preprocessing

[0054] Step 1.1: Raw data acquisition and time alignment;

[0055] Identify the gas well workover machine and mark it as the optimization target. Obtain the execution time of the stroke command during the operation phase of the optimization target, and construct the stroke command sequence S(t) according to the time order (unit: stroke / minute, sampling interval 1 minute). Obtain the instantaneous production and instantaneous power consumption based on the execution time of the stroke command. Construct the instantaneous production sequence Q(t) and instantaneous power consumption sequence E(t) according to the time order of the same sequence, with units of 10,000 cubic meters / day and kWh / day, respectively.

[0056] After data acquisition is completed, time alignment is performed, and three parameter sequences—stroke command S—are read in real time from the machine controller. raw (t) (unit: times / minute, sampling period 1 second), instantaneous output Q raw (t) (unit: 10,000 cubic meters / day, provided by orifice plate flow meter or ultrasonic flow meter, sampling period 1 second), instantaneous power consumption P raw (t) (unit: kW, provided by frequency converter or electricity meter, sampling period 1 second);

[0057] The three sequences are assigned a unified timestamp t. i =i×Ts(T s Linear interpolation alignment is performed for values ​​of 1 second (t = 1 second). Missing values ​​are either padded forward (for missing durations less than 10 seconds) or marked as invalid (for missing durations exceeding 10 seconds). The average value per minute is calculated to generate the downsampled minute-level sequence S(t).m ),Q(t m ),E(t m ), where t m =1, 2, 3, ... minutes, t k For the tth m The kth second within a minute;

[0058] Thrust sequence Unit: times / minute;

[0059] Production sequence Unit: 10,000 cubic meters / day;

[0060] Average power Unit: kW;

[0061] Power consumption sequence Unit: kWh / day, minute-level aligned sequence {S(t m ),Q(t m ),E(t m The time span covers the past 72 hours up to the present moment.

[0062] Step 1.2: Detection of stroke variation points and division of stable operating condition window;

[0063] Minute-level impulse sequence S(t) m );

[0064] A sliding window variance test method is used—with a window length L = 10 minutes—to calculate the standard deviation σS of the number of impulses within the window. If σS < 0.05 impulses / minute and the difference between the impulse rate at the center point of the window and the mean of the adjacent windows is less than 0.1, then the window is marked as a "stable segment." Continuous stable segments are merged to obtain the start and end times of the k-th stable operating condition window. If the interval between adjacent stable segments is less than 5 minutes, they are considered as the same adjustment transient and are merged to obtain K stable operating condition windows. S k Take the median of impulses within the window (to resist transient disturbances). In this specification, the symbol S for all impulse-related events... k S curr S prev S new All use the same unit (times / minute). The subscripts k, curr, prev, and new are only used to distinguish different operating points and have no difference in physical dimensions.

[0065] Step 1.3: Extraction of steady-state yield and power consumption;

[0066] For each steady-state window Wk, the corresponding output sequence Q(t) m ) and power consumption sequence E(t) m (Time range) The length of each steady-state window is set to L. k To avoid the influence of transient adjustments, data from the last 30 minutes of the window (or the entire window if the window length is less than 30 minutes) are used, and the arithmetic mean is calculated to obtain the steady-state output Q for that window. k and steady-state power consumption E k Simultaneously record the window's end time. As the timestamp of that operating condition point, we obtain the set of operating condition points. , k=1...K.

[0067] Step 1.4: Local energy efficiency curvature feature mining;

[0068] Sort the set of operating points P by time;

[0069] For each subset P k (Excluding the first and last points), take its predecessor point P. k−1 and successor point P k+1 Calculate the second-order difference in the "stroke-output" plane: , Calculate the second-order difference in the "stroke-power consumption" plane: , Define the local energy efficiency curvature: ;

[0070] Where λ is the dimensionless normalization coefficient, taken as 1. (Average output over the past 24 hours divided by average power consumption). >0 indicates increasing marginal returns. <0 indicates diminishing marginal returns; the larger the absolute value, the sharper the inflection point, thus obtaining the energy efficiency curvature corresponding to each operating point. , stored in the set of operating points P.

[0071] Phase Two - Building Dynamic Pareto Frontiers

[0072] Step 2.1: Time-domain attenuation weighting;

[0073] Determine the set of operating points P and the current time t. now Calculate the time-domain weights for each point Pk. T decay =72 hours (based on an empirical value for the timescale of gas well productivity decline), weight range (0,1], with closer points having higher weights, resulting in a weighted set of operating point P. w ={(S k Q k E k ,w k )};

[0074] Step 2.2: Two-dimensional Pareto front extraction;

[0075] Based on the weighted set of operating points, with power consumption E as the horizontal axis and output Q as the vertical axis, a variant of the "fast non-dominated sorting" algorithm is used to sort all points in ascending order of power consumption E; the frontier point list F is initialized to empty.

[0076] After traversing the sorted points, if the output Q of the current point is greater than the output of the last point in F (i.e., the output increases with power consumption and it is not dominated by a previous point), then add that point to F. Strictly defined: Point A dominates point B if and only if E A ≤E B And Q A ≥Q B And at least one inequality holds strictly;

[0077] After the traversal is complete, the points in F are arranged in ascending order of power consumption and output, forming the Pareto front, and the sequence of front points is output. , j=1...M, where M≤K.

[0078] Step 2.3: Calculate the gradient of the frontier marginal efficiency;

[0079] Extracting adjacent leading points F from the leading point sequence F. j−1 and F j Calculate the marginal efficiency gradient: (Unit: 10,000 cubic meters / kWh), this gradient represents the increase in output obtained by consuming 1 kWh more electricity from point j−1 to point j, and the rate of change of the gradient (second derivative) is also calculated: ;

[0080] Physical meaning: This indicates that marginal efficiency is decreasing at an accelerating rate, and the region is about to enter the "efficiency cliff," yielding the marginal efficiency gradient sequence corresponding to the frontier point. and gradient rate of change .

[0081] Step 2.4: Feedback correction of the weight of the operating point based on the leading edge curvature;

[0082] Extract historical operating point P and its energy efficiency curvature For the leading edge point F, and for the working condition point P which is not on the leading edge. k If its If the value is negative (in the diminishing marginal returns region) and its absolute value is greater than the threshold of 0.05, then its weight W will be adjusted during the next frontier construction. k Multiply by a decay factor γ=0.5 to reduce its influence on the frontier. This step ensures that the frontier is not "contaminated" by historical erroneous conditions, and the updated weights are used in the next iteration of step 2.1.

[0083] Phase 3 - Generation of Dynamic Break-Even Threshold

[0084] Step 3.1: Input economic parameters and calculate basic thresholds;

[0085] Enter the real-time electricity price C elec (RMB / kWh, obtainable from electricity market API or manually set value), Natural Gas Price C gas (RMB / 10,000 cubic meters, can be read from the gas station's billing system), equipment unit time depreciation cost C dep (RMB / day, adjusted based on equipment original value, lifespan, and operating time), calculate the basic break-even marginal efficiency: ;

[0086] Where ΔQ min =0.1 million cubic meters / day, representing the smallest distinguishable change in output; subtract Cdep / ΔQ from the denominator. min This is to amortize equipment depreciation into incremental costs—increased production requires equipment to run for longer periods, accelerating depreciation and thus obtaining a baseline threshold. (Unit: 10,000 cubic meters / kWh)

[0087] Step 3.2: Calculation of dynamic penalty factor based on frontal local curvature;

[0088] Extracting the frontier marginal efficiency gradient sequence gradient rate of change And the current operating point is located in the leading edge area. Within the projection interval on the front (i.e., finding the point Fj in the front where the power consumption is closest to and slightly greater than the current power consumption, and the point Fj−1 where it is slightly less than the current power consumption), calculate the local marginal efficiency decline rate of the front at the current projection point: ;

[0089] It also calculates the historical average rate of decline in marginal efficiency. Take all frontier intervals from the past 24 hours. Define a dynamic penalty factor for the average of negative values: ;

[0090] Where α = 0.3 (empirical coefficient), ϵ = 10 −6 To prevent division by zero, when the rate of decline of marginal efficiency exceeds the historical average, λ(t) > 1, the threshold is raised, the system becomes more conservative, and a dynamic penalty factor λ(t) is obtained.

[0091] Step 3.3: Dynamic threshold synthesis;

[0092] Extracting the base threshold The dynamic penalty factor λ(t) is used to calculate the final dynamic break-even threshold. Output the marginal efficiency threshold that is updated in real time. The value typically ranges from 0.01 to 0.5 million cubic meters per kWh.

[0093] Phase 4 - Real-time calculation of marginal efficiency and multi-objective decision-making

[0094] Step 4.1: Calculate the actual marginal efficiency;

[0095] Extract current operating point The previous stable operating point (i.e., the point corresponding to the most recently confirmed steady-state window), it is necessary to ensure and There is a clear change in the number of strokes, if the current stroke S curr With S prev If the absolute value of the difference is less than 0.1 times / minute, it is judged as "no effective adjustment has occurred", this stage is skipped, and "maintain current instruction" is output directly; otherwise, the following is calculated: ;

[0096] Also check the sign of the denominator: if E curr -E prev <0 and Q curr -Q prev If the output is greater than 0 (output increases after frequency reduction), it is marked as an "abnormal event," triggering a diagnostic alarm and skipping subsequent decisions, thus obtaining the actual marginal efficiency η. real And effectively adjust the flag position.

[0097] Step 4.2: Calculation of production fluctuation entropy;

[0098] Extract the minute-level output sequence Q(t) from the past 2 hours m ), the time window is [t curr -120,t curr ] Calculate the sample entropy H Q (Normalize to [0,1]), the steps are as follows:

[0099] The sequence is standardized to a sequence x with a mean of 0 and a standard deviation of 1. i ;

[0100] Set the embedding dimension m=2 and the similarity tolerance r=0.2×std(x);

[0101] Calculate the similarity ratio between all subsequences of length m and subsequences of length m+1 to obtain the original sample entropy value SampEn;

[0102] Normalization: Where N = 120 (minutes), H QThe closer the sequence is to 0, the more regular it is (e.g., stable); the closer it is to 1, the more random it is (e.g., violently fluctuating). This yields the production fluctuation entropy H. Q ∈[0,1].

[0103] Step 4.3: Construction and decision-making of the three-dimensional cost function;

[0104] Extract η real η threshold (t), H Q And the current surge S curr and device limit S max (e.g., 8 times / minute), construct the cost function: Where β = 0.05 (weighting coefficient, determined through offline simulation, so that H...) Q When the value is 1, it is equivalent to raising the threshold by 0.05 million cubic meters per kWh. The decision rule is as follows:

[0105] If Φ>0 and S curr <S max If the value is -0.1 (leaving a step size margin), the decision is to execute incremental production: generate instructions ΔS = +0.1 times / minute, mark them as "tentative production increase", and record the current timestamp t. probe ;

[0106] If Φ≤0 and S curr >S min (S) min If the frequency is 3.0 times / minute, the decision is to refuse to increase production and trigger an exploratory frequency reduction.

[0107] In other cases (such as when the stroke limit has been reached or the lower limit has been reached), the output will be "Keep current command";

[0108] The decision type D∈{increase production, explore frequency reduction, maintain} and the corresponding stroke adjustment ΔS are obtained.

[0109] Phase 5 - Exploratory frequency reduction and Pareto frontier update closed loop

[0110] Step 5.1: Trial execution with frequency reduction;

[0111] Extract decision type D = exploration frequency reduction, current impulse S curr Execute a small frequency reduction command ΔS = −0.15 times / minute (the step size is slightly larger than the production increase step size to produce an observable response), send the new stroke command to the frequency converter, and record the trial start time t. explore =t curr Set a frequency reduction instruction and a wait timer.

[0112] Step 5.2: Waiting for a stable window and data acquisition;

[0113] Based on the waiting timer, the real-time data stream after the impulse change is obtained, and the waiting time T is... wait =max(30 minutes, 5 × 60 / s) new (minutes), of which S new For the down-frequency pulses, the second item ensures that at least 5 complete pulse cycles are collected;

[0114] During this period, output and power consumption are continuously monitored, but no new adjustments are made. After the waiting period ends, the new steady-state operating point P is extracted using the same method as in steps 1.2-1.3. new =(S new Q new E new ,t new ), thus obtaining the new operating point P. new .

[0115] Step 5.3: Frontier recalculation and validity assessment;

[0116] Extract the original set of operating points P and add P new And the original frontier F, will P new Add P, re-execute the second stage (steps 2.1-2.3), generate a new Pareto front F′, and determine P. new If the solution is located on F′ (i.e., whether it is the Pareto optimal solution), then mark the frequency reduction as "valid exploration" if it is, otherwise mark it as "invalid exploration". The updated set of working points P′, the updated frontier F′, and the validity flag are obtained.

[0117] Step 5.4: Update the historical average rate of decline in marginal efficiency;

[0118] Extract the valid exploration indicator, the marginal efficiency gradient sequence {∇ηj′} of the new frontier. If valid = true, extract all negative gradient change rates Δ∇ηj′ < 0 from the new frontier and recalculate. If valid=false, then keep the original state. The value remains the same, but an invalid counter is added. count If three consecutive frequency reduction attempts are invalid count If ≥3), the system is determined to be in the "platform zone" of the global Pareto front, and any subsequent frequency reduction attempts are stopped. A message "System has reached the optimal operating zone" is generated, and the updated system is obtained. And a sign indicating whether exploration has ceased.

[0119] Step 5.5: Close the loop and return to the first stage;

[0120] Regardless of whether the decision is to increase production, decrease frequency, or maintain, the system continues to run, repeating the entire process in 1-minute cycles (starting from step 1.1). However, the rebuilding frequency of the Pareto front can be adjusted according to the computational load (e.g., a full rebuild every 10 minutes, with intermediate cycles only updating the current operating point without rebuilding the front).

[0121] The operation optimization management platform generates job execution analysis signals and sends them to the job execution analysis unit;

[0122] After receiving the job execution analysis signal, the job execution analysis unit performs execution analysis on the optimization object and optimizes it based on the analysis results;

[0123] During the operation phase of the optimization object, the initiation of valve commands is recorded, the initiation time of the corresponding valve command is determined, the data corresponding to the valve command sequence is extracted as the valve opening, the valve command sequence is constructed according to the time order, and the feedback time is recorded based on the feedback value of the pressure sensor after the valve command is initiated, and the pressure sensor feedback sequence is constructed according to the time order.

[0124] Based on the comparison of adjacent moments in the valve command sequence, the time axis is divided into rising segment, falling segment and stationary segment. The valve command sequence is used as the horizontal axis and the pressure sensor feedback sequence is used as the vertical axis. In the non-stationary segment, the command pressure trajectory is constructed. Specifically, for the rising segment, the trajectory extends from the low command to the high command.

[0125] For the descent segment, the trajectory extends in the opposite direction, superimposing the ascending segment trajectory and the descending segment trajectory within a complete control cycle (e.g., a round trip from low to high and back to low) to form a hysteresis loop—two non-overlapping curves, with the ascending trajectory located above or below the descending trajectory, depending on the direction of energy dissipation of the system.

[0126] Data collection and analysis were performed on the hysteresis loop, and the loop area of ​​the hysteresis loop was extracted. The loop area is represented by the area enclosed between the rising curve and the falling curve, and is obtained by integration using the trapezoidal rule.

[0127] If the area of ​​the hysteresis loop continues to increase or exceeds the set red line area threshold, based on the continuous increase in the duration of the operation phase, it is inferred that the optimization object is in the hysteresis effect stage, generating a severe hysteresis effect signal and sending it to the operation optimization management platform. After receiving the severe hysteresis effect signal, the operation optimization management platform marks the current stage as the severe hysteresis effect stage.

[0128] Conversely, if the area of ​​the hysteresis loop does not continue to increase and does not exceed the set red line area threshold, it is inferred that the optimization object is not in the hysteresis effect stage, a hysteresis effect normal signal is generated and sent to the operation optimization management platform, and after receiving the hysteresis effect normal signal, the operation optimization management platform marks the current stage as the hysteresis effect stable stage.

[0129] During the stable phase of the hysteresis effect, the valve command sequence is executed according to the set rules. However, after entering the severe phase of the hysteresis effect, the period interval of the valve command sequence is increased, and the operation of the optimized object is coordinated simultaneously. The period interval of the operation of the optimized object is aligned with the period interval of the valve command sequence, that is, the period interval is lower than the set interval threshold.

[0130] During the severe hysteresis phase, when the optimized objects are running together, the average slope of the rising segment in the middle of the loop and the average slope of the falling segment in the same instruction interval are obtained based on the hysteresis loop, and the local slope asymmetry is calculated based on the ratio of the average slopes.

[0131] If the local slope asymmetry exceeds the set local slope asymmetry threshold, an asymmetry compensation signal is generated and sent to the operation optimization management platform. After receiving the signal, the operation optimization management platform performs asymmetry compensation, specifically through asymmetric flutter signal modulation.

[0132] Principle: The amplitude and direction of a conventional chatter signal are independent. Under asymmetric hysteresis, the side with greater frictional resistance requires stronger chatter energy to effectively "shake" the valve core away from the static friction zone.

[0133] Implementation: Based on the local slope asymmetry, calculate the flutter amplitude modulation coefficients in two directions: rising flutter amplitude and falling flutter amplitude (assuming the rising segment has a larger slope and more severe friction). The flutter signal is still a sine wave or square wave, but the amplitudes of the positive and negative half cycles are not equal, forming "bias flutter".

[0134] Effect: Without increasing the total energy, the flutter energy is concentrated in the direction of motion with greater frictional resistance, thus improving the compensation efficiency;

[0135] If the local slope asymmetry does not exceed the set local slope asymmetry threshold, a continuous monitoring signal is generated and sent to the operation optimization management platform. The optimization management platform generates a dynamic safety warning signal and sends it to the dynamic safety warning unit.

[0136] After receiving the dynamic safety warning signal, the dynamic safety warning unit integrates meteorological and environmental vibration data to optimize the dynamic safety warning of the target operation.

[0137] Based on the operational phase of the optimized object, environmental parameters and adjacent interference parameters within the operational phase are determined. The environmental parameters are collected by a three-dimensional ultrasonic anemometer, which is installed near the lightning rod on the top of the derrick to collect wind speed, wind direction, and gust frequency in real time. A high sampling frequency (≥10 Hz) is required to capture instantaneous wind pressure fluctuations that have a significant impact on the structure. The adjacent interference parameters are collected by ground vibration sensors. Triaxial accelerometers are arranged on the four base flanges of the derrick or on the nearby foundation to monitor the ground vibration frequency and amplitude caused by adjacent operations, vehicle traffic, etc.

[0138] Simultaneously, data is collected from the optimized object, including structural response sensing data and operational status sensing data. The structural response sensing data is represented by attaching fiber optic strain gauges (FBGs) to key stress-bearing parts of the derrick (such as the crossbeam of the crane base, the connection between the thigh and the diagonal brace, and the support of the second-level platform) to monitor micro-strain forces in real time. The operational status sensing data is represented by acquiring core process parameters such as hook load, traveling car position, winch torque, brake pressure, and engine speed in real time through the PLC / control system interface.

[0139] The current time is used to obtain the completed runtime segment, and the frequency of environmental parameters exceeding the range of parameters of the completed runtime segment during the runtime phase is obtained. At the same time, the duration of the neighboring interference parameter values ​​exceeding the set parameter threshold is collected.

[0140] If the frequency of environmental parameters exceeding the range of parameters for the completed running phase exceeds the frequency threshold, or the duration of neighboring interference parameters exceeding the set parameter threshold exceeds the duration threshold, then the current phase will be marked as a high interference phase.

[0141] If the frequency of environmental parameters exceeding the range of parameters during the operation phase does not exceed the frequency threshold, and the duration of neighboring interference parameters exceeding the set parameter threshold does not exceed the duration threshold, then the current phase is marked as a low interference phase.

[0142] Acquire the range of changes in structural response sensing data and the cumulative amount of time when the corresponding type of operational status sensing data is outside the set range during the operation phase;

[0143] If the change span of the structural response sensing data of the optimized object exceeds the change span threshold during the operation phase, or the cumulative amount of the corresponding type value of the operation status sensing data exceeds the cumulative amount threshold during operation when it is outside the set range, then the corresponding phase will be marked as a status change phase.

[0144] If the change span of the structural response perception data of the optimized object does not exceed the change span threshold during the operation phase, and the cumulative amount of time when the corresponding type value of the operation status perception data is outside the set range does not exceed the cumulative amount threshold, then the corresponding phase will be marked as a stable phase.

[0145] When the state change phase coincides with the high interference phase, the current optimization direction is set to optimize the object's operating environment and control the surrounding interference. When the state change phase coincides with the low interference phase, the current optimization direction is set to optimize the object's own operating state, i.e., perform hardware inspection and maintenance.

[0146] In summary, the collaborative work of the three core units completely solves the core technical problems in the traditional gas well workover rig operation optimization management, such as low energy efficiency, execution deviation, insufficient safety warning, and poor coordination, thereby achieving intelligent and automated optimization and safety control of the entire gas well workover rig operation process.

[0147] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A gas well operation machine operation optimization management system based on multi-parameter automatic adjustment, characterized in that, This includes a runtime optimization management platform, whose communication connections include: The production and energy consumption relationship identification unit identifies the production and energy consumption relationship of the gas well workover machine and obtains the stable operating points of the gas well workover machine under different strokes. Each operating point includes at least stroke, production and power consumption data. A two-dimensional Pareto frontier of output and power consumption is dynamically constructed based on historical operating conditions, and the marginal efficiency gradient of the frontier is calculated. A dynamic break-even threshold is generated based on the changing trends of real-time electricity prices, gas prices, and the marginal efficiency gradient of the frontier. Calculate the actual marginal efficiency corresponding to the current production increase order, and correct the actual marginal efficiency by combining the output fluctuation entropy; compare the corrected actual marginal efficiency with the dynamic break-even threshold: if it is greater than the threshold and the number of strokes has not reached the upper limit, then execute the incremental production increase; If the value is less than or equal to the threshold, production is rejected and an exploratory frequency reduction is initiated. The Pareto front is updated based on the new operating point after the exploratory frequency reduction, and subsequent dynamic thresholds are recursively adjusted. The job execution analysis unit performs execution analysis on the optimization object and optimizes it based on the analysis results; the dynamic safety early warning unit integrates meteorological and environmental vibration data to provide dynamic safety early warnings for the operation of the optimization object.

2. The multi-parameter automated regulation based gas well operation machine operation optimization management system according to claim 1, characterized in that, The process of generating the dynamic break-even threshold further includes: The basic break-even marginal efficiency is calculated based on real-time electricity and gas prices. Extract the local marginal efficiency decline acceleration of the Pareto front at the current operating condition projection point and compare it with the historical average decline rate to construct a dynamic penalty factor; Multiplying the basic break-even marginal efficiency by the dynamic penalty factor yields the dynamic threshold.

3. The multi-parameter automated regulation based gas well operation machine operation optimization management system of claim 2, wherein, The calculation and correction process for output fluctuation entropy further includes: Collect minute-level production sequences within a set time period in the past, calculate their sample entropy as the production fluctuation entropy, which represents the randomness of production fluctuation; when constructing the cost function for comparison, subtract the dynamic break-even threshold from the actual marginal efficiency, and then subtract the product of the production fluctuation entropy and the preset weight coefficient.

4. The multi-parameter automated regulation based gas well operation machine operation optimization management system of claim 3, wherein, The exploratory downsampling and frontier update process further includes: When the production increase order is rejected, the stroke rate is actively reduced by a set exploration step size, and the process waits for at least one stable window duration; the new stable operating point after the frequency reduction is collected, added to the historical operating point set, and the Pareto front is reconstructed. Determine if the new operating point is located on the updated Pareto front: if so, mark this frequency reduction as a valid exploration and update the historical average marginal efficiency decline rate; if multiple frequency reductions do not produce a new front point, determine that the system is in the optimal operating zone and stop the exploration.

5. The multi-parameter automated regulation based gas well operation machine operation optimization management system of claim 4, wherein, Before constructing the Pareto front, the extraction and weight correction of local energy efficiency curvature features at the operating point are also included: For each historical operating point, the local energy efficiency curvature is calculated based on the second difference between the "stroke-output" plane and the "stroke-power consumption" plane. A positive value of the curvature indicates the region of increasing marginal returns, and a negative value indicates the region of decreasing marginal returns. When constructing the Pareto front, an additional decay factor is applied to operating points with negative curvature and absolute values ​​exceeding a set threshold, in addition to the time-domain decay weight, to reduce their contribution to the front. Meanwhile, the strength of the additional decay factor is dynamically adjusted based on the rate of change of the marginal efficiency gradient of adjacent points on the frontier, so that the frontier is not contaminated by inefficient historical operating conditions.

6. The multi-parameter automated regulation based operating optimization management system for gas well operations machines of claim 1, wherein, The execution analysis process of the optimization object in the job execution analysis unit is as follows: Extract the valve command sequence corresponding data as valve opening degree, construct the valve command sequence according to time order, and record the feedback time according to the feedback value of the pressure sensor after the valve command is initiated, and construct the pressure sensor feedback sequence according to time order. Based on the comparison of adjacent moments in the valve command sequence, the time axis is divided into rising segment, falling segment, and stationary segment; with the valve command sequence as the horizontal axis and the pressure sensor feedback sequence as the vertical axis, the command pressure trajectory is constructed in the non-stationary segment, forming a hysteresis loop.

7. The multi-parameter automated regulation based operating optimization management system for gas well operations machines of claim 6, wherein, Data collection and analysis were performed on the hysteresis loop, and the loop area was extracted. The loop area is represented by the area enclosed between the rising curve and the falling curve, and is obtained by integration using the trapezoidal rule. If the area of ​​the hysteresis loop continues to increase or exceeds the set red line area threshold, based on the continuous increase in the duration of the operation phase, it is inferred that the optimization object is in the hysteresis effect stage, and the current stage is marked as the severe hysteresis effect stage. Conversely, if the area of ​​the hysteresis loop does not continue to increase and does not exceed the set red line area threshold, it is inferred that the optimization object is not in the hysteresis effect stage, and the current stage is marked as the hysteresis effect stable stage.

8. The multi-parameter automated regulation based operating optimization management system for gas well operations machines of claim 7, wherein, During the stable phase of the hysteresis effect, the valve command sequence is executed according to the set rules. However, after entering the severe phase of the hysteresis effect, the period interval of the valve command sequence is increased, and the operation of the optimized object is coordinated simultaneously. The period interval of the operation of the optimized object is aligned with the period interval of the valve command sequence, that is, the period interval is lower than the set interval threshold. During the severe hysteresis phase, when the optimized objects are running together, the average slope of the rising segment in the middle of the loop and the average slope of the falling segment in the same instruction interval are obtained based on the hysteresis loop. The local slope asymmetry is calculated based on the ratio of the average slopes. If the local slope asymmetry exceeds the set local slope asymmetry threshold, an asymmetry compensation signal is generated and sent to the operation optimization management platform; otherwise, a continuous monitoring signal is generated.

9. The gas well workover rig operation optimization management system based on multi-parameter automated adjustment according to claim 1, characterized in that, The dynamic safety early warning process in the dynamic safety early warning unit is as follows: Based on the operational phase of the optimization object, determine the environmental parameters and neighboring interference parameters within the operational phase; simultaneously, collect data on the optimization object. The current time is used to obtain the completed runtime segment, and the frequency of environmental parameters exceeding the range of parameters of the completed runtime segment during the runtime phase is obtained. At the same time, the duration of the neighboring interference parameter values ​​exceeding the set parameter threshold is collected. The current stage is divided into a high-interference stage and a low-interference stage based on threshold comparison.

10. The gas well workover rig operation optimization management system based on multi-parameter automated adjustment according to claim 9, characterized in that, Acquire the range of changes in structural response sensing data and the cumulative amount of time when the corresponding type of operational status sensing data is outside the set range during the operation phase; Based on threshold comparison, the current stage is divided into a state change stage and a state stability stage; When the state change phase coincides with the high interference phase, the current optimization direction is set to optimize the object's operating environment and control surrounding interference; when the state change phase coincides with the low interference phase, the current optimization direction is set to optimize the object's own operating state, i.e., perform hardware inspection and maintenance.

Images