CN-122020203-A - Closed loop optimization method and system for effusion detection-recommendation-execution-multi-disc

CN122020203ACN 122020203 ACN122020203 ACN 122020203ACN-122020203-A

Abstract

The invention belongs to the technical field of energy, and discloses a closed-loop optimization method of effusion detection-recommendation-execution-multi-disc, which comprises data acquisition and transmission; data preprocessing and feature engineering, model reasoning and anomaly judgment, advanced prediction and alarm mechanism, result display and manual labeling, and reasoning and delay optimization. The model of the invention can dynamically adapt to the production characteristics and the change trend of different wells, obviously improves the recognition precision of abnormal states such as effusion and the like, reduces the false alarm and missing alarm conditions, and does not rely on post detection any more, but gives out early warning a few hours before the occurrence of the abnormality, so that the operation and maintenance response is more active, and the average production cycle is greatly shortened.

Inventors

ZHOU XIN
Zhou Junxuan
WANG YULONG
LIU CHANGDI
Fu Zidan

Assignees

成都鑫耀天合科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260302

Claims (10)

1. The hydrops detection-recommendation-execution-multi-disc closed-loop optimization method is characterized by comprising the following steps of: S1, an abnormality early warning stage: acquiring production time sequence data such as oil pressure, casing pressure, flow and valve back pressure of a gas well in real time, carrying out characteristic construction on the time sequence data, inputting a hydrops abnormal detection model, and outputting a judging result of whether the gas well is in a hydrops abnormal state or not and a corresponding early warning signal; S2, intelligent recommending stage of the production process measures: Based on the abnormal state of the effusion, the current shaft dynamic parameters and the historical technological measure execution effect data, constructing a technological matching model, and outputting a productive technology scheme aiming at a target gas well; s3, an automatic or semi-automatic execution stage: Generating an execution instruction according to the reproduction process scheme, and sending the execution instruction to a production control system or an operation and maintenance terminal to implement corresponding effusion elimination measures; s4, closed loop evaluation and feedback phase: And collecting production recovery data after the effusion elimination measures are executed, evaluating the effectiveness of the early warning result and the productive technology scheme, and transmitting the evaluation result back to the effusion anomaly detection model and the technology matching model as feedback data so as to realize iterative updating of model parameters.
2. The method of claim 1, wherein the features configured in the anomaly early warning stage include a differential pressure feature between oil pressure and casing pressure, a statistical feature calculated based on a sliding time window, a time differential feature of production parameters, and an anomaly occurrence interval feature.
3. The method of claim 1, wherein the hydrops anomaly detection model is a deep neural network model based on multi-dimensional time sequence feature input, and the output is a probability value corresponding to a normal state and an abnormal state, and the early warning signal is triggered according to whether the probability value of the abnormal state exceeds a preset threshold.
4. An intelligent recommendation method for gas well production process based on historical execution feedback for implementing the hydrops detection-recommendation-execution-duplication closed-loop optimization method according to any one of claims 1-3, characterized in that the method comprises the following steps: a1, constructing a candidate process space, wherein the candidate process space comprises foam drainage, plunger re-production, gas lift re-production and a compressor pressurization process; a2, calculating the similarity between the target gas well and the historical gas well according to the shaft structural parameters, the pressure distribution characteristics and the hydrops degree of the target gas well, and screening to obtain an initial candidate process set; A3, taking the current production state of the target gas well as state information, taking the initial candidate process as optional action, taking the post-recovery yield change and implementation cost as feedback information, and establishing a process selection strategy model; and A4, outputting a reproduction process recommendation result aiming at the target gas well based on the process selection strategy model.
5. The method of claim 4, wherein the similarity is calculated based on multi-dimensional vector similarity of well depth, pipe diameter, pressure distribution characteristics, and historical hydrops extent.
6. The method of claim 4 wherein the process selection strategy model uses the combined evaluation of the yield recovery amplitude and the process implementation cost per unit time after the recovery as feedback information for adjusting the selection weight of each candidate process.
7. A model feedback enhancement method for gas well fluid remediation implementing the fluid accumulation detection-recommendation-execution-multi-disc closed loop optimization method of any one of claims 1-3, the method comprising: b1, collecting production recovery data of a gas well in a preset time window after the execution of the production process is completed; B2, calculating yield recovery amplitude, pressure change trend and reproduction duration index based on the production recovery data; b3, generating a process execution effect evaluation result according to the index; And B4, taking the process execution effect evaluation result as a supervision or reinforcement signal, and respectively correcting an alarm threshold parameter of the effusion anomaly detection model and a strategy parameter of the reproduction process recommendation model.
8. The method of claim 7, wherein the yield recovery amplitude is calculated by comparing average yield changes over a preset period of time before and after performance of the rework procedure.
9. The method of claim 7, wherein the pressure trend is characterized by calculating a rate of oil pressure rise and a magnitude of casing pressure change.
10. The method of claim 7, wherein the model feedback enhancement method retains the manual correction result when correcting the model parameters, and uses the manual correction result and the automatically generated evaluation result together as a model update basis.

Description

Closed loop optimization method and system for effusion detection-recommendation-execution-multi-disc Technical Field The invention belongs to the technical field of energy sources, and particularly relates to a closed-loop optimization method and a closed-loop optimization system for effusion detection, recommendation, execution and multi-disc. Background Wellbore fluid is a core cause of productivity fluctuation and yield loss in shale gas exploitation. The prior art mainly relies on the SCADA system to set a static threshold value for alarming, and has the defects of high false alarm rate, dependence on manual experience analysis, response lag, model capacity loss and the like. Although partial prior art realizes abnormality early warning, the full-chain closed-loop management from 'abnormality discovery' to 'intelligent scheme recommendation' to 'feedback and model evolution' is not formed yet, so that the intelligent level of production operation and maintenance is difficult to continuously improve. Shale gas is an important unconventional natural gas resource, and has been developed and utilized on a large scale in recent years. With the increase of production years, shale gas wells in most of domestic zones gradually enter an old well stage, and the productivity of the shale gas wells shows volatility and instability. According to year-round statistical data of 2024, southwest oil and gas fields average daily effusion wells to 10, single well average production cycle is 2.5 days, single well average daily yield is 1.5 square meters, daily yield loss caused by the average daily yield loss is up to 15 square meters, and year-round yield influence is about 5500 square meters. How to improve the response efficiency of abnormal working conditions such as effusion and the like and the intelligent level of production operation and maintenance become the urgent problem to be solved in the current shale gas production operation and maintenance. In the prior art, the identification and treatment of anomalies such as accumulated liquid in a shaft, shutdown of a compressor, blockage of a pipeline and the like in the shale gas production process still mainly depend on manual experience and a static threshold alarming mechanism set in a SCADA (SupervisoryControlandDataAcquisition) system or a POC (ProductionOperationControl) system. The method has the following remarkable defects: 1. The SCADA/POC system alarms based on manually set rules, cannot adapt to complex working conditions caused by the evolution of a shaft state along with time and environmental change, and is extremely easy to generate the phenomena of missing report or false report. 2. The current abnormality judgment still depends on the experience judgment and manual analysis of field engineering personnel, the efficiency is low, and the accurate and rapid dynamic management is difficult to realize. 3. The data utilization rate is low, and although the underground sensor and the ground monitoring equipment realize a large amount of time sequence data acquisition, the existing system cannot fully mine the hidden dynamic evolution rule. 4. The model capability is lost, an intelligent model with self-learning and automatic feature extraction capability is lacking, and a detection system with high robustness and high adaptability is difficult to build in a multi-well and multi-dimensional data environment. Under the push of digital transformation, the shale gas exploitation area is built into a real-time monitoring system covering a shaft and a ground flow, and standardized acquisition and centralized management of key parameters such as oil pressure, casing pressure, instantaneous flow, valve back pressure and the like are realized. The infrastructure provides a good data environment for the application of the artificial intelligence technology in the shale gas well production dynamic anomaly detection. Under the background, the artificial intelligence, particularly the machine learning technology represented by deep learning, has the advantages of nonlinear modeling, automatic feature extraction, strengthening capability and the like, and can identify fine abnormal trends in a complex dynamic production environment. By introducing the hydrops anomaly detection system based on the deep neural network, the limitation of a traditional rule driving mechanism can be broken through, real-time detection and prediction of various complex anomalies can be realized, the automation and intelligence level of anomaly response is improved, and finally, the fine management and intelligent early warning of the shale gas well production process are realized. Therefore, there is a need for an artificial intelligence-based shale gas and liquid accumulation anomaly detection method and system with high detection accuracy, strong robustness, good real-time performance and self-adaption, which are used for replacing the existing identification mechanism relying on manual an