Oil and gas reservoir intelligent dynamic monitoring and optimized mining method based on big data

Abstract

The invention relates to the technical field of oil and gas field development, and discloses an oil and gas reservoir intelligent dynamic monitoring and optimized mining method based on big data. According to the method, through multi-source data real-time sensing and dynamic coupling modeling, the limitation of traditional static analysis is broken through, and minute-level accurate description of the state of the oil and gas reservoir is achieved; a machine learning prediction model based on physical constraints effectively fuses a data rule and a geomechanics principle, and the early warning reliability of water-driven dynamics and equipment faults is remarkably improved; a multi-objective optimization strategy of digital twin driving comprehensively plans the recovery ratio, the economy and the safety, and supports intelligent decision making under complex working conditions; the edge control and the block chain technology are combined to form a'sensing-optimizing-executing-auditing 'closed loop, so that the high efficiency, the self-adaptability and the data credibility of the mining process are ensured. According to the overall scheme, oil and gas exploitation is promoted to be transformed to intelligentization, refinement and transparency, and technical support is provided for cost reduction, efficiency improvement and sustainable development of an oil field.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas field development, and specifically to an intelligent dynamic monitoring and optimized exploitation method for oil and gas reservoirs based on big data. Background Art

[0002] Current dynamic monitoring of oil and gas reservoirs mainly relies on periodic static tests (such as pressure build-up well testing, production logging) and manual experience analysis. Conventional methods obtain key parameters by deploying discrete sensors (such as downhole pressure gauges, flow meters), establish a static geological model by combining numerical simulation software (such as Eclipse, CMG), and formulate an exploitation plan based on historical data. Some existing technologies have attempted to introduce machine learning algorithms (such as random forest, support vector machine) to predict production capacity indicators, but their model training mostly relies on offline historical data, and local optimization is achieved by manually adjusting injection and production parameters. In addition, existing systems usually adopt centralized data storage, and ensure data reliability through regular manual auditing.

[0003] There are still deficiencies in the existing technologies:

[0004] The coverage density of traditional sensor networks is low and the sampling frequency is insufficient, resulting in monitoring blind spots for key parameters (such as fracture propagation, fluid front), and the regulation decision-making depends on periodic manual analysis, unable to respond to dynamic changes in real time; geological modeling, seepage simulation and equipment condition analysis are independent of each other, lacking a multi-physical field coupling mechanism, and model updates rely on manual intervention, making it difficult to adapt to complex scenarios such as dynamic degradation of reservoir physical properties; machine learning models lack physical law constraints, the interpretability of prediction results is poor, and the optimization process ignores the balance between economy and safety. At the same time, centralized data storage is vulnerable to tampering and difficult to meet the needs of multi-party collaborative auditing. Therefore, an intelligent dynamic monitoring and optimized exploitation method for oil and gas reservoirs based on big data is proposed. Summary of the Invention

[0005] Aiming at the deficiencies of the existing technologies, the present invention provides an intelligent dynamic monitoring and optimized exploitation method for oil and gas reservoirs based on big data to solve the problems in the background art.

[0006] To achieve the above object, the present invention provides the following technical solution: An intelligent dynamic monitoring and optimized exploitation method for oil and gas reservoirs based on big data, comprising the following steps:

[0007] Step 1, multi-source data acquisition and preprocessing: Deploy heterogeneous sensor networks in the wellbore, surface gathering pipelines and reservoir to collect multi-dimensional production data including formation pressure, temperature, fluid flow rate, water cut, downhole motor vibration spectrum, casing deformation data in real time, and perform noise filtering, spatio-temporal alignment and normalization processing on the original data;

[0008] Step 2, Dynamic Coupling Modeling: Integrate the preprocessed real-time data with the 3D geological model and historical development data to construct a multi-physical field coupling analysis model that includes reservoir physical property parameters, fluid migration laws, and equipment operating status;

[0009] Step 3, Intelligent Prediction and Diagnosis: Based on the coupling model, establish an injection-production response prediction module through machine learning algorithms to real-time output prediction results of reservoir pressure distribution, water drive front position, fracture closure risk index, and equipment failure probability;

[0010] Step 4, Optimization Decision Generation: According to the prediction results, simulate the production effects under different combinations of production parameters in the digital twin system, and use multi-objective optimization algorithms to generate injection-production volume adjustment plans, optimized pump pumping frequencies, and chemical flooding agent injection strategies;

[0011] Step 5, Closed-loop Execution and Feedback: Send the optimization plan to the wellsite edge controller for execution, synchronously collect the production data after execution, and update the prediction model parameters through an online learning mechanism to form a closed-loop control link of "perception - decision - execution - verification".

[0012] Preferably, in the above Step 1:

[0013] The heterogeneous sensor network includes distributed fiber optic acoustic sensors (DAS), downhole permanent pressure gauges, multiphase flow meters, and high-precision wellhead temperature sensors;

[0014] Data preprocessing uses a sliding window dynamic calibration method to compensate for sensor drift, and the compensation formula is:

[0015] x corrected =x raw ·(1 + α·Δt) + β·sin(2πf noise t)

[0016] where x raw is the original sensor reading; α is the sensor time drift coefficient, calibrated through daily zero-point self-check data; Δt: the time interval since the last calibration; β is the environmental electromagnetic interference amplitude; f noise is the main interference frequency, identified through fast Fourier transform.

[0017] Preferably, the multi-physical field coupling analysis model in Step 2 includes three-way coupling of a geomechanics module, a seepage module, and an equipment operating condition module;

[0018] Use a loose coupling iterative algorithm to achieve data interaction between modules, and complete within each time step:

[0019] Update the reservoir permeability tensor according to the change in in-situ stress;

[0020] Calculating the fluid pressure distribution based on the new permeability;

[0021] Back-calculating the casing deformation amount through the fluid pressure and correcting the vibration constraint conditions of the equipment.

[0022] Preferably, in step three:

[0023] The machine learning algorithm uses a physics-informed neural network (PINN), and its loss function includes:

[0024]

[0025] where u pred is the predicted value of pressure / saturation output by the network; u obs is the measured pressure / saturation data; K is the permeability tensor; p is the fluid pressure field; λ1, λ2 are the weight coefficients of the data term and the physical constraint term.

[0026] Preferably, in step four:

[0027] The digital twin system includes a full-chain simulation model of formation - wellbore - surface equipment;

[0028] Multi-objective optimization uses an improved NSGA-II algorithm, and its fitness function includes:

[0029] F = [ER, NPV, 1 / WC] T

[0030] where ER is the predicted recovery factor, calculated through the material balance equation; NPV is the net present value, considering dynamic parameters such as oil price and operating cost; WC is the comprehensive water cut, with different weight coefficients assigned to different well groups.

[0031] Preferably, in step five:

[0032] The edge controller deploys an adaptive PID algorithm, and the control variables include:

[0033]

[0034] where u(t) is the control output (mA); e(t) is the deviation between the set value and the actual value; K p is the proportionality coefficient, dynamically adjusted according to the reservoir pressure gradient; K i is the integral coefficient, satisfying K i = 1 / (2τ) with the injection-production response delay time τ; K d is the differential coefficient, which is forced to zero when the sand production warning is activated.

[0035] Preferably, a risk hedging mechanism is introduced in the optimization decision of step four, and a conservative exploitation plan is enabled when one of the following situations occurs:

[0036] The single-day decline of international oil prices exceeds 7%;

[0037] The monitored value of CO2 concentration underground breaks through the critical value of H2S synergistic corrosion;

[0038] The load rate of the regional power grid has been higher than 90% for 2 consecutive hours.

[0039] Preferably, the integration of the blockchain trusted deposit mechanism in the fifth step specifically includes:

[0040] Writing the hash value of the optimized decision parameter into the smart contract:

[0041] H decision = SHA3-256(Q inj ||Q prod ||Timestamp)

[0042] where Q inj is the water injection volume adjustment instruction, Q prod is the oil production setting value, and Timestamp is the UNIX timestamp;

[0043] When the deviation between the actual production data and the predicted value exceeds 15%, the blockchain deposit audit process is automatically triggered to trace the original sensor data and the intermediate results of model calculation.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] Through real-time perception of multi-source data and dynamic coupling modeling, the present invention breaks through the limitations of traditional static analysis and realizes minute-level accurate characterization of the oil and gas reservoir state; the machine learning prediction model based on physical constraints effectively integrates data laws and geomechanics principles, significantly improving the early warning reliability of water drive dynamics and equipment failures; the multi-objective optimization strategy driven by digital twins coordinates recovery efficiency, economy and safety, and supports intelligent decision-making under complex working conditions; the combination of edge control and blockchain technology forms a "perception-optimization-execution-audit" closed loop, ensuring the efficiency, self-adaptability and data credibility of the mining process. The overall solution promotes the transformation of oil and gas exploitation towards intelligence, refinement and transparency, providing technical support for cost reduction, efficiency improvement and sustainable development of oilfields.

[0046] Other features and advantages of the present invention will be described in the subsequent specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained by the structures pointed out in the specification, claims and drawings. Description of the Drawings

[0047] Figure 1This is the flow chart of the intelligent dynamic monitoring and optimized exploitation method for oil and gas reservoirs based on big data in the present invention. Detailed implementation manners

[0048] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art in the technical field of the present invention without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0049] Please refer to Figure 1 , an intelligent dynamic monitoring and optimized exploitation method for oil and gas reservoirs based on big data in the present invention, includes the following steps:

[0050] Step 1: Multi-source data collection and preprocessing

[0051] 1. Sensor deployment:

[0052] Downhole monitoring:

[0053] Install a distributed fiber optic temperature / sonic wave sensor (DTS / DAS) with the model of Silixa XT-DTS every 15 meters in the horizontal well section. The temperature measurement range is -40°C to 300°C, the accuracy is ±0.1°C, and the sonic wave sampling rate is 2 kHz, which is used to monitor the wellbore temperature distribution and microseismic events.

[0054] Deploy a fiber Bragg grating pressure sensor (FBG) 5 meters above the perforation section, with a measurement range of 0 - 100 MPa and a resolution of 0.01 MPa, to collect formation pressure data in real time.

[0055] Surface monitoring:

[0056] Install a multiphase flowmeter (Roxar MFM 2600) at the wellhead to measure the flow rates of oil, gas, and water phases, with an accuracy of ±1.5% and a data update frequency of 1 second.

[0057] Arrange vibration sensors (Emerson CSI 9420) on the gathering and transportation pipelines, with a monitoring frequency range of 10 Hz - 10 kHz, to detect abnormal vibrations of pumps and valves.

[0058] 2. Data preprocessing:

[0059] For the problem of sensor time drift, the sliding window dynamic calibration method is adopted, and the compensation formula is:

[0060] x corrected = x raw ·(1 + α·Δt) + β·sin(2πf noise t)

[0061] Among them, x raw is the original reading of the sensor; α is the sensor time drift coefficient, calibrated through the zero-point self-check data every day; Δt: the time interval (in hours) since the last calibration; β is the environmental electromagnetic interference amplitude (mV); f noise is the main interference frequency (Hz), identified through fast Fourier transform;

[0062] Example: The original reading x raw of a certain pressure gauge = 12.5 mV, the time drift coefficient α = 0.002 / h (calibrated through the daily zero-point self-check), the time interval since the last calibration Δt = 10 h, the interference amplitude β = 0.2 mV, and the noise frequency f noise = 50 Hz. Then the corrected data is:

[0063] Xcorrected = 12.5×(1 + 0.002×10) + 0.1×sin(100πt) = 12.75 + 0.1sin(100πt) mV

[0064] Noise filtering: Perform wavelet threshold denoising on the DAS acoustic wave signal, select the sym8 wavelet basis, decompose it into 6 layers, perform soft threshold processing, and retain the formation fracture characteristic signal (<500 Hz).

[0065] Eliminate sensor drift and electromagnetic interference, and improve data reliability; the error of the compensated pressure data is reduced from ±0.5 MPa to ±0.1 MPa, the signal-to-noise ratio (SNR) of the acoustic wave signal is increased to 40 dB, effectively identify fracture propagation events, and meet the requirements of high-precision modeling.

[0066] Step 2: Dynamic coupling modeling

[0067] 1. Construction of multi-physical field coupling model:

[0068] 1.1. Geomechanics module: Solve the control equation based on the in-situ stress data (inverted through fiber optic strain):

[0069]

[0070] Among them, σ is the in-situ stress tensor, ρ is the rock density, and g is the acceleration due to gravity;

[0071] 1.11. Microseismic event correction: According to the source location (x i , y i , z i ), and the energy E i , update the shear modulus μ:

[0072]

[0073] Among them, A i is the fracture area, and ν is the Poisson's ratio.

[0074] 1.12. Seepage module: The non-Darcy seepage equation is used to describe the flow in the fracturing cracks:

[0075]

[0076] Among them, p is the fluid pressure; K is the permeability, and the initial value is determined by core experiments; μ is the fluid viscosity; β is the non-Darcy coefficient.

[0077] 1.13. Equipment condition module: Calculate the allowable vibration speed according to the casing strain ∈:

[0078]

[0079] Among them, E is the elastic modulus of the steel; ρ steel is the density of the steel; ∈ is the measured value of the strain.

[0080] 2. Loose coupling iterative algorithm:

[0081] In each time step (Δt = 1 hour), the following operations are performed in sequence:

[0082] 2.1 Permeability update: If the in-situ stress changes by Δσ = 5 MPa, according to K new = K old ·e 0.02Δσ make a correction to obtain K new = 8 × e 0.01 = 8.84 mD;

[0083] Pressure field calculation: Solve the seepage equation to obtain the pressure p = 35.2 MPa at a certain grid point;

[0084] Equipment constraint update: Calculate v max = 0.65 m / s, and if this value is exceeded, the frequency reduction protection is triggered.

[0085] Step three: Intelligent prediction and diagnosis

[0086] 1. Training of the physics-informed neural network (PINN):

[0087] Network structure: Input layer (14 features such as pressure, temperature, and flow rate), 4 hidden layers (256 neurons, with the activation function being Swish), output layer (water saturation S w , fracture risk index R f );

[0088] Loss function:

[0089]

[0090] Among them, u predis the predicted pressure / saturation value for network output (dimensionless); u obs is the measured pressure / saturation data (dimensionless); K is the permeability tensor; p is the fluid pressure field; λ1, λ2 are the weight coefficients of the data item and the physical constraint item.

[0091] Training data: 10-year historical production data of an oilfield (500,000 sets of samples), Adam optimizer (learning rate 0.001), batch size 512, trained until the loss converges (about 100,000 steps).

[0092] 2. Abnormal diagnosis mechanism:

[0093] Improve the accuracy of water cut prediction and fault diagnosis;

[0094] Water cut mutation warning: When the predicted S w rises by more than 5% within 2 hours, trigger the injection-production ratio adjustment algorithm:

[0095] ΔQ inj = k p (S w,target - S w,pred ) + k i ∫(S w,target - S w,pred )dt

[0096] where k p = 0.8 is the proportionality coefficient; k i = 0.2 is the integral coefficient; S w,target = 70% is the target water saturation.

[0097] Equipment fault diagnosis: Analyze the vibration spectrum of the electric pump. If the amplitude of the 63Hz (bearing defect characteristic frequency) component > 0.4g, determine that the fault probability > 85%.

[0098] Example:

[0099] Predict the S of a well group w rising from 65% to 72%, the actual monitoring is 71.3%, with an error of 0.7%. Automatically adjust the water injection volume from 250m 3 / d to 280m 3 / d;

[0100] Detect the 63Hz component of 0.45g of an electric pump, give an early warning 36 hours in advance, and replace the bearing to avoid shutdown losses.

[0101] Step Four: Optimization Decision Generation

[0102] Increase the recovery rate and reduce the carbon emission intensity. Risk hedging reduces the losses caused by oil price fluctuations;

[0103] 1. Digital Twin System:

[0104] Build a full 3D geological model on the Petrel platform with 1 million grid cells, integrating real-time production data (refresh frequency: 30 seconds);

[0105] The virtual wellbore model includes detailed components such as casing, packers, pump valves, etc., and simulates the temperature-stress coupling effect.

[0106] 2. Multi-objective optimization algorithm:

[0107] Improve the parameters of the NSGA-Ⅲ algorithm: population size 200, crossover probability 0.9, mutation probability 0.05, number of iterations 500;

[0108] Fitness function:

[0109] F = [ER, NPV, 1 / WC, -E carbon T

[0110] Among them, ER is the predicted recovery rate, calculated through the material balance equation; NPV is the net present value, considering dynamic parameters such as oil price and operating cost; WC is the comprehensive water cut, with different weight coefficients assigned to different well groups; E carbon is the carbon emission index, and the weight is dynamically adjusted according to the carbon tax policy.

[0111] Constraint conditions:

[0112] Bottom-hole flowing pressure P wf ≥15 MPa (higher than the bubble point pressure of 14.2 MPa);

[0113] Injection-production ratio 0.9 ≤ ∑Q inj / ∑Q prod ≤1.2.

[0114] 3. Risk hedging mechanism:

[0115] When the single-day decline rate of international oil price > 8%, switch to the low-production and stable-production mode, reducing power consumption by 15% - 20%;

[0116] When the downhole H2S concentration > 25 ppm, automatically inject corrosion inhibitor (concentration 30 ppm) and increase the pH to > 6.5.

[0117] Example:

[0118] After optimization, the daily production of a certain well group increased from 800 tons to 850 tons, the water cut decreased from 68% to 65%, and the annual NPV increased by $15 million; during the period of sharp decline in oil price, the automatic adjustment plan saved $25,000 in electricity costs per day.

[0119] Step Five: Closed-loop Execution and Blockchain Evidence Storage

[0120] ​Short control response time; blockchain evidence storage shortens data audit time and reduces costs.

[0121] 1. Edge control execution:

[0122] Adaptive PID control rule:

[0123]

[0124] Among them, u(t) is the control output (mA); e(t) is the deviation between the set value and the actual value; K p is the proportionality coefficient, dynamically adjusted according to the reservoir pressure gradient; K i is the integral coefficient, and satisfies K i = 1 / (2τ) with the injection-production response delay time τ; K d is the differential coefficient, which is forced to zero when the sand production warning is activated.

[0125] Example: For a well with a pressure deviation e(t) = 1.5 MPa e(t) = 1.5 MPa, calculate the control output:

[0126]

[0127] Adjust the frequency of the injection pump from 45 Hz to 48 Hz, and the water injection volume increases by 12 m 3 / h.

[0128] 2. Blockchain evidence storage:

[0129] Hash calculation: Write the hash value of the optimized decision parameter into the smart contract:

[0130] H decision = SHA3-256(Q inj ||Q prod ||Timestamp)

[0131] Among them, Q inj is the water injection volume adjustment instruction, Q prod is the set value of the oil production volume, and Timestamp is the UNIX timestamp.

[0132] Example:

[0133] Hdecision = SHA3-256(300||1800||1659376800000)

[0134] Among them, Q inj = 300 m 3 / d, Q prod = 1800 m 3 / d, Timestamp = 2022-08-02 12:00:00 (UNIX timestamp);

[0135] Smart contract trigger: When the deviation of the actual oil production volume > 12%, automatically retrieve the original DAS data and intermediate model parameters to locate the abnormal source (such as the drift of a certain FBG pressure gauge exceeding the limit).

[0136] This embodiment provides a method for intelligent dynamic monitoring and optimized exploitation of oil and gas reservoirs based on big data. By deploying heterogeneous sensor networks such as distributed fiber optic sensors and multiphase flow meters, data such as formation pressure, temperature, and fluid flow rate are collected in real time. A multi-physical field coupling model of geomechanics-seepage-equipment working conditions is constructed, and a physics-informed neural network (PINN) is used to accurately predict water cut and fracture risk. Based on digital twin and improved NSGA-Ⅲ algorithm, a multi-objective optimization scheme is generated, and injection-production parameters are adjusted in real time in combination with adaptive PID edge control to form a closed-loop control link. At the same time, blockchain hash deposit is used to ensure the immutability of data and full-process traceability. Practical applications show that this method improves the recovery rate, reduces the ineffective water injection volume, increases the accuracy of fault warning, reduces the carbon emission intensity, and has the core advantages of efficient exploitation, intelligent operation and maintenance, and green and low-carbon.

Claims

1. A method for intelligent dynamic monitoring and optimized exploitation of oil and gas reservoirs based on big data, characterized in that: The following steps are involved: Step 1: Multi-source data collection and preprocessing: Deploy heterogeneous sensor networks in the wellbore, surface gathering and transportation pipelines, and reservoirs to collect multi-dimensional production data including formation pressure, temperature, fluid flow rate, water content, downhole motor vibration spectrum, and casing deformation data in real time, and perform noise filtering, spatiotemporal alignment, and normalization on the raw data; Step 2: Dynamic coupling modeling: Integrate the pre-processed real-time data with the three-dimensional geological model and historical development data to build a multi-physics coupling analysis model that includes reservoir physical parameters, fluid migration laws, and equipment operating status; Step 3: Intelligent prediction and diagnosis: Based on the coupling model, an injection-production response prediction module is established through machine learning algorithms to output the prediction results of reservoir pressure distribution, water flooding front position, fracture closure risk index and equipment failure probability in real time; Step 4: Optimization decision generation: Based on the prediction results, the production effects under different mining parameter combinations are simulated in the digital twin system, and a multi-objective optimization algorithm is used to generate an injection and production adjustment plan, an optimized value for pumping frequency, and a chemical displacement agent injection strategy; Step 5: Closed-loop execution and feedback: The optimization plan is sent to the well site edge controller for execution, and the production data after execution is collected synchronously. The prediction model parameters are updated through the online learning mechanism to form a closed-loop control link of "perception-decision-execution-verification".

2. The method for intelligent dynamic monitoring and optimized exploitation of oil and gas reservoirs based on big data according to claim 1 is characterized in that: In the step 1: The heterogeneous sensor network includes distributed fiber optic acoustic sensors (DAS), downhole permanent pressure gauges, multiphase flow meters, and wellhead high-precision temperature sensors; The data preprocessing adopts the sliding window dynamic calibration method to compensate for the sensor drift. The compensation formula is: x corrected =x raw ·(1+α·Δt)+β·sin(2πf noise t) Among them, x raw is the original reading of the sensor; α is the sensor drift coefficient, which is calibrated by the daily zero self-test data; Δt: the time interval from the last calibration; β is the amplitude of the environmental electromagnetic interference; f noise The main interference frequency is identified by fast Fourier transform.

3. The method for intelligent dynamic monitoring and optimized exploitation of oil and gas reservoirs based on big data according to claim 1 is characterized in that: The multi-physics field coupling analysis model in step 2 includes a three-way coupling of a geomechanics module, a seepage module, and an equipment operating condition module; A loosely coupled iterative algorithm is used to implement data interaction between modules, which is completed within each time step: Update the reservoir permeability tensor according to the change of ground stress; Calculate fluid pressure distribution based on the new permeability; The casing deformation is inferred through fluid pressure to correct the equipment vibration constraint conditions.

4. The method for intelligent dynamic monitoring and optimized exploitation of oil and gas reservoirs based on big data according to claim 1 is characterized in that: In the step three: The machine learning algorithm adopts physical information neural network (PINN), and its loss function includes: Among them, u pred is the pressure / saturation prediction value output by the network; u obs is the measured pressure / saturation data; K is the permeability tensor; p is the fluid pressure field; λ1, λ2 are the weight coefficients of the data item and the physical constraint item.

5. The method for intelligent dynamic monitoring and optimized exploitation of oil and gas reservoirs based on big data according to claim 1 is characterized in that: In the step 4: The digital twin system includes a full-chain simulation model of formation-wellbore-surface equipment; The multi-objective optimization adopts the improved NSGA-II algorithm, and its fitness function includes: F=[ER,NPV,1 / WC] T Among them, ER is the predicted recovery factor, which is calculated by the material balance equation; NPV is the net present value, which takes into account the dynamic parameters of oil price and operating cost; WC is the comprehensive water cut, which assigns weight coefficients to different well groups.

6. The method for intelligent dynamic monitoring and optimized exploitation of oil and gas reservoirs based on big data according to claim 1 is characterized in that: In the step 5: The edge controller deploys an adaptive PID algorithm, and the control variables include: Where, u(t) is the control output (mA); e(t) is the deviation between the set value and the actual value; K p is the proportionality coefficient, which is dynamically adjusted according to the reservoir pressure gradient; K i is the integral coefficient, and satisfies K with the injection-production response delay time τ i =1 / (2τ); K d It is the differential coefficient, which is forced to zero when the sand warning is activated.

7. The method for intelligent dynamic monitoring and optimized exploitation of oil and gas reservoirs based on big data according to claim 1 is characterized in that: A risk hedging mechanism is introduced in the optimization decision of step 4, and a conservative mining plan is activated when one of the following situations occurs: International oil prices fell by more than 7% in a single day; The CO2 concentration monitoring value underground exceeded the critical value of H2S synergistic corrosion; The regional power grid load rate remained above 90% for 2 hours.

8. The method for intelligent dynamic monitoring and optimized exploitation of oil and gas reservoirs based on big data according to claim 1 is characterized in that: The blockchain trusted evidence storage mechanism is integrated in step 5, specifically including: Write the optimization decision parameter hash value into the smart contract: <h2 style=";text-align:left;direction:ltr">H<h2 style=";text-align:left;direction:ltr"> decision <h2 style=";text-align:left;direction:ltr"> =SHA3-256(Q<h2 style=";text-align:left;direction:ltr"> inj <h2 style=";text-align:left;direction:ltr"> ∣∣Q<h2 style=";text-align:left;direction:ltr"> prod <h2 style=";text-align:left;direction:ltr"> ∣∣Timestamp) Among them, Q inj Q is the water injection adjustment instruction. prod is the oil production setting value, Timestamp is the UNIX timestamp; When the actual production data deviates from the predicted value by more than 15%, the blockchain evidence audit process is automatically triggered to trace the original sensor data and the intermediate results of the model calculation.

Images