An intelligent prediction and optimization control method for the operation state of an oilfield sewage settling tank

Abstract

The present application relates to a kind of intelligent prediction and optimization control method for oilfield sewage settling tank operating state, it includes using the modular design reconfigurable settling system of fusion multimodal perception, collection time series data;In combination with physical information neural network PINN algorithm, the mapping relationship between the working characteristics of actual settling tank and experimental settling tank is constructed: through experimental settling tank data, the performance of actual settling tank is predicted;Experimental settling tank operating state identification and prediction, establish multi-parameter fuzzy comprehensive evaluation model;The intelligent decision-making and automatic control of the settling system operation;Markov decision process MDP is used as the optimization framework of control decision, and the PPO algorithm of migration enhancement is combined to optimize control decision, to realize the automatic adjustment of system performance;Realize the continuous self-optimization for Markov decision process MDP decision process.The present application solves the problem that settling tank operating state is difficult to real-time monitoring and dynamic regulation and control in the traditional oilfield sewage treatment process.

Description

TECHNICAL FIELD

[0001] The present application relates to oilfield sewage treatment technology, in particular to an intelligent prediction and optimization control method for the operation state of an oilfield sewage settling tank. BACKGROUND

[0002] With most of the oilfields in China entering the late development stage, the oil production technology gradually shifts to tertiary oil recovery (EOR) technology (such as polymer flooding, chemical flooding), resulting in significant changes in the composition of oilfield produced water, and the complexity of the composition intensifies the presence of chemical agents such as polyacrylamide (HPAM) and residual surfactants in the produced water, forming stable emulsified oil and colloidal suspended solids, and the efficiency of traditional gravity settling separation decreases by 30%~50%. At the same time, the water quality indicators are more stringent, and the reinjection water needs to meet the requirements of "SY / T 5329-2022 Clastic Rock Reservoir Water Quality Index" for suspended solids (≤10 mg / L), oil content (≤15 mg / L), and particle size (≤2 μm), while the existing process is difficult to stabilize and meet the standards. High dosage of agents (such as demulsifiers and flocculants) leads to a 20%~40% increase in treatment costs, and non-compliant reinjection water causes formation plugging (permeability decreases by 50%~70%) and injection pressure multiplication (from 5 MPa to more than 12 MPa), significantly increasing equipment maintenance and oil production energy costs, and economic and environmental pressures coexist.

[0003] The current oilfield sewage settling separation process mainly relies on static gravity settling tanks, and its efficiency is very limited. There are dead zones and short flow phenomena in the traditional settling tank, heat transfer is uneven, and the effective separation volume utilization rate is less than 60%. When the water quality fluctuates (such as polyacrylamide concentration and temperature changes), there is a lack of real-time control means, and there is a dynamic response lag, which prolongs the settling time and relies excessively on manual experience to adjust the process parameters, making it difficult to match the optimal working condition under complex water quality. SUMMARY

[0004] The purpose of the present application is to provide an intelligent prediction and optimization control method for the operation state of an oilfield sewage settling tank, which solves the technical problem of real-time monitoring and dynamic control of the operation state of the settling tank in the traditional oilfield sewage treatment process.

[0005] The technical solution adopted by the present application to solve its technical problems is that the intelligent prediction and optimization control method for the operation state of an oilfield sewage settling tank comprises the following steps:

[0006] Step one, using the modular design of the reconfigurable settling system with fusion multi-modal perception, collect time series data, including flow, particle concentration, flow field vorticity; the settling system integrates PIV particle image velocimetry, distributed optical fiber sensor, conductivity-turbidity multi-parameter sensor array and temperature dyeing technology, builds a multi-dimensional physical field dynamic monitoring network, realizes three-dimensional visualization reconstruction of experimental settling tank internal flow pattern characteristics and quantitative analysis of particle migration trajectory;

[0007] Step two, combined with the physical information neural network PINN algorithm, the mapping relationship between the working characteristics of the actual settling tank and the experimental settling tank is constructed: through the experimental settling tank data, the performance of the actual settling tank is predicted;

[0008] Step three, experimental settling tank running state recognition and prediction; establish a multi-parameter fuzzy comprehensive evaluation model, use the multi-parameter fuzzy comprehensive evaluation model, according to the current experimental settling tank running data, real-time judge the experimental settling tank running state, and establish evaluation index; combined with machine learning algorithm to predict the possible situation of continuous running of the current state of the settling system, and the change of future running state with time;

[0009] Step four, the intelligent decision and automatic control of the settling system; through accurate monitoring of the dynamic variables directly related to the settling process, using Markov decision process MDP as the optimization framework of control decision, combining with the optimization of control decision of migration enhanced PPO algorithm, realizing the automatic adjustment of system performance, the dynamic variables include oil layer thickness, turbidity, temperature, pH value and suspended solids concentration;

[0010] Step five, decision verification and Markov decision process MDP dynamic correction closed-loop optimization control, through the deep coupling of GBDT and reinforcement learning, realize the continuous self-optimization of Markov decision process MDP decision process.

[0011] The multi-parameter fuzzy comprehensive evaluation model in step three of the above scheme is as follows:

[0012] ;

[0013] In the formula: p w is the density of water; p o is the density of oil; g is the acceleration of gravity; α is the dynamic coefficient; Q is the liquid distribution flow; k DAF is the efficiency coefficient of air flotation unit; A b is the total bubble surface area flux; vb is the bubble rising velocity; d 32 is the Sauter mean bubble diameter; C oil is the oil content; C oil,crit is the critical emulsification concentration of oil phase; E a is the apparent activation energy; R is the ideal gas constant; T is the heating coil temperature; T ref is the reference temperature; f coil x c , y c z c is the heating coil spatial position influence function at spatial position ( x c , y c z c ); P i is the water distribution port i =1, heating coil i =2, flotation unit i =3 spatial coordinate; P opt is the optimal layout position of each component; L char is the characteristic length, taking 20% of the tank diameter; w i is the weight coefficient; m T is the fluid viscosity; A settlle is the effective settling area; Q max is the critical flow rate; C TSS is the sewage turbidity; β is the oil phase correction coefficient.

[0014] The modular design of the reconfigurable settling system integrating multi-modal perception in the above scheme includes an experimental module, a data acquisition module, and a data set analysis control module, and each module cooperates to form a complete experimental feedback loop;

[0015] ​​The experimental module comprises an experimental settling tank, a power unit, a heating unit and a flotation unit. The experimental settling tank is provided with a central column tube and a central reaction cylinder. The flotation unit is a dissolved air pump. The dissolved air pump sucks air and sewage together. The impeller of the dissolved air pump cuts the sewage and air into fine foam, so that they are fully mixed. At the same time, the high pressure generated by the high-speed rotation of the impeller fully dissolves the sucked gas into the water. The dissolved air water is released under normal pressure through a pressure reducing valve, and micro bubbles with a diameter of 20-50 μm are generated.

[0016] The data acquisition module is integrated with a PIV particle image velocimetry device, a CCD camera array, a multi-spectral turbidity meter and a distributed sensor group. The PIV particle image velocimetry device and the CCD camera are triggered synchronously by laser pulses to realize sub-micron dynamic analysis of the particle size distribution of oil droplets. The multi-spectral turbidity meter adopts a multi-wavelength scattered light intensity detection technology and an embedded temperature and pressure compensation algorithm to realize online calibration of the turbidity change in the settling process. The distributed sensor group includes a micro pressure difference transmitter and an array type temperature probe to form a multi-point physical parameter monitoring network of the internal flow field of the settling tank.

[0017] The data analysis control module is configured with a motion feature extraction unit, a multi-physical field coupling analysis unit and a settling efficiency evaluation unit. The motion feature extraction unit adopts an improved PTV particle tracking algorithm to perform Gaussian fitting denoising processing on continuous frame images, and calculates the oil droplet velocity field, acceleration field and motion trajectory fractal dimension. The multi-physical field coupling analysis unit analyzes the oil droplet coalescence and breakup dynamics by establishing a correlation matrix model of turbidity-particle size-flow rate. The settling efficiency evaluation unit constructs a particle group settling prediction model based on the Fokker-Planck equation, and outputs key indicators such as settling rate and interface clarity.

[0018] The step two in the above scheme is specifically:

[0019] Step 1, constructing a physical information neural network architecture, input layer: parameters obtained from the operation of the experimental settling tank; output layer: actual settling tank performance parameters, including particle settling efficiency and turbidity removal rate;

[0020] Define the loss function:

[0021] ;

[0022] In the formula: data is a data matching item, and a small amount of actual settling tank calibration data is taken during calculation Y real ;

[0023] ;

[0024] phys is a physical constraint term;

[0025] scale is a similarity criterion constraint term;

[0026]

[0027] Adjusting the allocation of each item dynamically through the validation set;

[0028] Step 2, time series data preprocessing and physical information neural network (PINN) model training;

[0029] Enhance the time series data, and perform dimensionless processing on the time series data. At the same time, the time series data is interpolated to match the actual tank time axis. The experimental tank settlement time series data is used to train the basic PINN, and a high-precision mapping relationship between the experimental settlement tank and the actual settlement tank is established.

[0030] Step 3, dynamic update, continuously fine-tune the model with new data, and optimize the physical information neural network (PINN) model.

[0031] The specific method of step three in the above scheme:

[0032] Step 1, obtain data, including water distribution flow, air flotation intensity, heating temperature, sewage turbidity, and oil content;

[0033] Step 2, establish a multi-parameter fuzzy comprehensive evaluation model;

[0034] Based on the mass conservation and kinetic equation, the settlement efficiency is constructed h Model:

[0035] ;

[0036] Subfunction expansion:

[0037] (1) Air floatation intensification factor Γ:

[0038] ;

[0039] In the formula: k DAF Non-air floatation unit efficiency coefficient, related to the spatial position of the floatation unit; A b Total bubble surface area flux, m² / s; v b Bubble rising velocity; d 32 Sauter average bubble diameter; C oil,crit Oil phase critical emulsification concentration;

[0040] (2) Thermal effect correction term T eff :

[0041]

[0042] where: E a Ea is the apparent activation energy; R R is the ideal gas constant; T ref T0 is the reference temperature;

[0043] f coil ( x c , y c z c ) is the heating coil spatial position influence function at spatial position ( x c , y c z c );

[0044] (3) Spatial synergy factor ξ space :

[0045]

[0046] where: P i x, y, z are the spatial coordinates of the water distribution port ( i =1), heating coil ( i =2), and flotation unit ( i =3); P opt x0, y0, z0 are the optimal layout positions of each component; L char L is the characteristic length, taken as 20% of the diameter of the tank; w i w is the weight coefficient;

[0047] (4) Foundation settlement rate η0

[0048] ;

[0049] where: m(T) μ is the fluid viscosity; A settle Aeff is the effective settling area; Q max Qc is the critical flow rate;

[0050] Thus:

[0051] ;

[0052] The multi-parameter fuzzy comprehensive evaluation model is trained by the data obtained in step 1, the multi-parameter fuzzy comprehensive evaluation model is utilized, the above water distribution flow, air floatation intensity, heating temperature, sewage turbidity, oil content and dynamic gradient characteristics are input, and the target variable is the oil content of the separated water, the water content of the separated oil and the settling time, and after continuous iteration, the optimal experimental parameters are comprehensively evaluated;

[0053] Step 3, data dimension reduction, determining the correlation between each time series data and the settling tank efficiency, and judging the influence weight of each time series data;

[0054] Step 4, state identification and prediction.

[0055] Beneficial effects:

[0056] (1) The present application innovatively combines similar theoretical criteria and physical information neural network technology to construct a wastewater settling tank scale-down experimental system. Firstly, the similarity theory is used to guide the experimental model design, which greatly reduces the experimental cost and implementation difficulty while ensuring the similarity of the flow field. Secondly, the physical information neural network is used to fuse the measured data and the fluid mechanics equation, breaking through the limitations of traditional empirical formulas and realizing high-precision mapping of the multi-physical field characteristics of the experimental tank and the prototype tank. Thirdly, the hybrid modeling method of data-driven and mechanism model collaborative optimization can effectively overcome the poor generalization of pure data models and significantly improve the adaptability to complex working conditions. Through small-scale experiments combined with intelligent algorithm extrapolation, the high cost of full-scale testing can be avoided, and key parameters such as flow pattern distribution and settling efficiency can be obtained simultaneously, providing reliable basis for optimizing the structure parameters and operation control strategy of the actual tank.

[0057] (2) The present application innovatively combines mechanism model and machine learning algorithm to construct an optimization control system driven by physical law and intelligent decision-making. The mechanism model based on multiphase flow dynamics equation provides interpretable physical law constraints for the system, ensuring that the core settling process conforms to the principles of fluid mechanics. The multi-parameter fuzzy comprehensive evaluation model processes the uncertainty and non-linear correlation between process parameters and settling conditions through membership functions and fuzzy rule bases, identifies the state of the experimental equipment and judges the optimal working parameters. The Markov decision process model outputs dynamic decisions to optimize equipment working parameters, maximizing global benefits under time-varying conditions. The transfer enhanced deep learning algorithm optimizes the output decisions, and finally the gradient boosting decision tree algorithm is used to correct the decision model based on the deviation between the theoretical efficiency and the actual situation. The cooperation between algorithms can improve the settling efficiency in oilfield wastewater settling tanks and promote unmanned control, effectively overcome the technical bottlenecks of model mismatch and decision lag in traditional methods, and have both physical interpretability and dynamic optimization capability.

[0058] (Three) The application is aimed at the collaborative optimization design of the structure parameters and process parameters of the settling tank. Through the construction of a modular experimental device integrating intelligent sensor networks and automatic actuators, a working state prediction model is established by combining machine learning algorithms, and a self-adaptive control strategy is developed to realize the dynamic optimization of the settling process, including computational fluid dynamics simulation, intelligent sensor data fusion, prediction control model construction, and self-adaptive regulation methods based on real-time monitoring.

[0059] (Four) The application focuses on the two core directions of poly-containing sewage settling separation and intelligent control. Through multi-physical field coupling experiments, intelligent control technology combined with machine learning algorithms is developed to innovate traditional process modes. The self-adaptive regulation algorithm based on machine learning is developed to optimize the operating parameters of the settling tank in real time, which will significantly reduce the comprehensive cost of oilfield sewage treatment, ensure the stability of the reinjection water quality, and extend the development cycle of the formation water injection.

[0060] (Five) The application realizes real-time intelligent identification of the operating state of the settling tank (normal / abnormal, efficient / inefficient), multi-parameter diagnostic analysis of abnormal and inefficient conditions, and automatic generation of optimization decision schemes, solving the technical problems of real-time monitoring and dynamic regulation of the operating state of the settling tank in traditional oilfield sewage treatment processes, and providing an innovative solution for the intelligent and unmanned operation and management of oilfield water treatment equipment. Through the deep integration of deep learning algorithms and Internet of Things technology, the system realizes accurate judgment of the operating state, early warning of faults, and autonomous optimization of operating parameters, significantly improving the operating efficiency and management level of the oilfield sewage treatment system.

[0061] (Six) The application builds an experimental device that can simulate the oil-water separation process of the oilfield sewage settling tank, realize the visualization of the flow field and temperature field, and provide research conditions for in-depth study of the heat transfer and flow characteristics of the settling separation process, the formation of the flotation process control method, and the optimization of the heater structure, the water collection and distribution port structure, and the process parameters.

[0062] (Seven) The application can realize the visualization of the flow and temperature in the uninterrupted liquid distribution and intermittent liquid distribution process of the settling tank, obtain detailed three-dimensional temperature field and velocity field data, and test data is rich and accurate. The experimental conditions can be accurately controlled, which is beneficial to the in-depth study of the temperature and flow coupling characteristics of the oilfield sewage settling process, and can further calculate parameters such as the synergy angle and uniformity, providing experimental conditions for the study of the flow and heat transfer coupling characteristics of the oilfield sewage settling tank under different processes.

[0063] (Eight) The present application has a particle image velocimetry (PIV) device, which can obtain the flow and temperature distribution data of the medium in the experimental settling tank by shooting tracer particles and temperature dyes mixed in the experimental medium in advance, and the collected data has the characteristics of completeness, detail and synchronization of temperature field and velocity field; by combining the temperature data of the temperature sensor with the temperature data of the temperature dye, more accurate three-dimensional temperature field data can be obtained.

[0064] (Nine) The inlet and outlet of the experimental settling tank and the corresponding pipeline are provided with a plurality of valves, which can freely switch the process, so as to realize the simulation of various working processes, such as closing the liquid distribution port and opening the liquid collection port to lower the liquid level, closing the liquid collection port and opening the liquid distribution port to raise the liquid level, and adjusting the opening degree of the liquid collection port and the liquid distribution port. And the wall surface of the pipeline is covered with heat preservation material, and the fluid temperature entering the experimental settling tank can be more accurately and effectively controlled by adjusting the heating device. The rotation speed control of the centrifugal pump in the pipeline system cooperates with the flow sensor to accurately control the flow.

[0065] (Ten) The present application simultaneously measures three-dimensional temperature field and velocity field, and the measurement data can be used to analyze the coupling characteristics of three-dimensional temperature and velocity, such as but not limited to calculating the synergy angle of temperature field and velocity field, the uniformity of temperature field, the energy utilization rate of the device, the average flow velocity and the average temperature.

[0066] (Eleven) The present application realizes the precise recombination of multiple modules in three-dimensional space through high-precision quick-connect interface and intelligent positioning system, restores the actual equipment structure, maps the actual settling tank working characteristics by analyzing the experimental device operating parameters, forms a closed-loop research and development system that can support orthogonal experiment design and real-time feedback optimization. At the same time, other emerging technologies can be introduced to judge the efficiency improvement ability of the settling tank and optimize the structure parameters and working parameters of the emerging technologies by using machine learning algorithm. With a very high module reuse rate, an economical scientific research platform is created, and multiple discipline expansion interfaces are reserved, providing an innovative solution with precision, flexibility and expandability for emerging experimental technologies of sewage settling tanks. BRIEF DESCRIPTION OF DRAWINGS

[0067] Fig. 1 is a schematic diagram of an experimental settling tank;

[0068] Fig. 2 is a schematic diagram of a modular design reconfigurable settling system integrating multi-modal perception;

[0069] Fig. 3 is a schematic diagram of the implementation process of a physical information neural network algorithm for mapping from an experimental settling tank to an actual settling tank;

[0070] Fig. 4 is a schematic diagram of the implementation process of a principal component analysis algorithm for dimensionality reduction of experimental data of an experimental settling tank;

[0071] Fig. 5 A flowchart for implementing a multi-parameter fuzzy comprehensive evaluation method model for equipment operation state recognition and prediction is shown in Figure 1.

[0072] Fig. 6 A flowchart for implementing a Markov decision process algorithm is shown in Figure 2.

[0073] Fig. 7 A flowchart for implementing a migration enhanced PPO algorithm for decision scheme optimization is shown in Figure 3.

[0074] Fig. 8 A flowchart for implementing a gradient boosting decision tree algorithm for decision scheme backtracking verification is shown in Figure 4.

[0075] Fig. 9 A technical roadmap of the present application is shown in Figure 5.

[0076] In the figure: 1 - liquid preparation pipe; 2 - oil collection tank; 3 - liquid collection pipe; 4 - liquid outlet pipe; 5 - liquid inlet pipe; 6 - heating coil; 7 - air distribution pipe; 8 - tank wall; 9 - tank top; 10 - tank bottom; 11 - central reaction cylinder; 12 - central column pipe; 13 - experimental settling tank; 14 - data acquisition device; 15 - water bath heating box; 16 - centrifugal pump; 17 - power supply; 18 - cylindrical lens; 19 - pulsed laser; 20 - synchronizer; 21 - high-speed CCD camera; 22 - square cavity; 23 - sensor. DETAILED DESCRIPTION

[0077] The present application will be further described below with reference to the accompanying drawings: Example 1

[0078] Referring to Figs. 1-9 This intelligent prediction and optimization control method for the operation state of the oilfield wastewater settling tank includes the following steps:

[0079] Step one, use the modular design of the reconfigurable settling system that fuses multi-modal perception to collect time series data, including flow rate, particle concentration, and flow field vorticity.

[0080] Step two, combine the physical information neural network (PINN) algorithm to construct the mapping relationship between the actual settling tank and the experimental settling tank in terms of working characteristics: predict the performance of the actual settling tank based on the data from the experimental settling tank.

[0081] Step three, recognize and predict the operation state of the experimental settling tank; establish a multi-parameter fuzzy comprehensive evaluation model, use the multi-parameter fuzzy comprehensive evaluation model to judge the operation state of the experimental settling tank in real time based on the current operation data of the experimental settling tank, and establish evaluation indexes; then combine the machine learning algorithm to predict the possible situations that may occur under the current state of the settling system and the changes in future operation state over time.

[0082] Step four, the settlement system runs intelligent decision and automatic control; through accurate monitoring of dynamic variables directly related to the settlement process, adopting Markov decision process (MDP) as the optimization framework of control decision, combining the PPO algorithm of migration enhancement to optimize the control decision, the automation adjustment of system performance is realized, and the dynamic variables include oil layer thickness, turbidity, temperature, pH value and suspended solids concentration.

[0083] Step five, decision verification and dynamic correction of decision model closed-loop optimization control, through the deep coupling of GBDT and reinforcement learning, the continuous self-optimization of MDP model decision process is realized.

[0084] The modular design of the reconfigurable settlement system in the application fuses multi-modal perception, including an experimental module, a data acquisition module and a data set analysis control module, and the modules work together to form a complete experimental feedback closed loop.

[0085] The experimental module includes a basic settlement tank body, a power unit, a heating unit and a flotation unit; wherein the basic settlement tank (i.e. the experimental settlement tank) body includes a tank wall 8, a tank top 9, a tank bottom 10, a liquid inlet pipe 5, a liquid outlet pipe 4, an oil outlet pipe, a gas distribution pipe, a central column pipe 12 and a central reaction cylinder 11, the central column pipe 12 is connected to the liquid collecting pipe 3, the central reaction cylinder 11 is connected to the liquid distribution pipe 1, the oil-water mixture enters the central reaction cylinder 11 from the liquid inlet pipe 5, as the liquid level in the central reaction cylinder 11 rises, reaches the liquid distribution pipe 1, and the oil-water mixture enters the experimental settlement tank from the liquid distribution pipe 1; the liquid in the experimental settlement tank flows out from the liquid collecting pipe 3 and enters the central column pipe 12, the liquid level in the central column pipe 12 rises until the height of the liquid outlet pipe 4, and the liquid is discharged from the liquid outlet pipe 4. The main equipment of the power unit is a centrifugal pump, which is used to provide power for liquid distribution, liquid collection and oil collection, etc. The heating unit is a water bath device connected to the heating coil in the experimental settlement tank 13, making it a constant temperature heat source. The mechanical equipment of the flotation unit is a dissolved air pump, which uses a multiphase dissolved air pump to suck air and sewage together, the impeller of the dissolved air pump cuts the sewage and air into fine foam, making them fully mixed, at the same time, the high pressure generated by the high-speed rotation of the impeller can also fully dissolve the sucked gas into the water, and the combined action makes it form stable dissolved air water, when these dissolved air water is released at normal pressure through the pressure reducing valve, it will produce tiny gas bubbles with a diameter of 20-50 μm, the dissolved air pump is located outside the experimental settlement tank 13, and the gas distribution pipe 7 is arranged in the experimental settlement tank 13.

[0086] (2) The data acquisition module is integrated with a PIV particle image velocimetry device, a high-speed CCD camera array, a multi-spectral turbidimeter and a distributed sensor group. Among them, the PIV particle image velocimetry device and the high-speed CCD camera 21 are triggered by laser pulse synchronization, realizing sub-micron dynamic analysis of the particle size distribution (0.1-500 μm range) of oil droplets; the turbidimeter adopts multi-wavelength scattered light intensity detection technology (working wavelength range 400-900 nm), cooperates with the embedded temperature and pressure compensation algorithm, realizes online calibration of the turbidity change in the settling process; the distributed sensor group includes a micro pressure difference transmitter (range 0-10 kPa, accuracy ±0.1% FS) and an array temperature probe, forming a multi-point physical parameter monitoring network of the internal flow field of the settling tank. The PIV particle image velocimetry device includes a high-speed CCD camera 21, a square cavity 22, a cylindrical lens 18, a pulsed laser 19 and a synchronizer 20.

[0087] (3) The data analysis control module is based on the construction of a high-performance computing workstation, and is configured with: 1) a motion feature extraction unit, which adopts an improved PTV particle tracking algorithm, performs Gaussian fitting denoising processing on continuous frame images, calculates oil droplet velocity field, acceleration field and motion trajectory fractal dimension; 2) a multi-physical field coupling analysis unit, which establishes a correlation matrix model of turbidity-particle size-flow rate, analyzes the oil droplet coalescence and breakup dynamics characteristics; 3) a settling efficiency evaluation unit, which establishes a particle group settling prediction model based on the Fokker-Planck equation, outputs key indicators such as settling rate and interface definition.

[0088] The present application is based on the similarity theory and the physical information neural network to design a scaled experiment model according to the actual sewage settling tank:

[0089] (1) The experimental device is designed strictly in accordance with the structural parameters and operating conditions of the actual settling tank, and by establishing accurate geometric similarity and dynamic similarity criteria, it is ensured that the experimental model and the actual settling tank are highly consistent in fluid dynamics characteristics. According to the geometric characteristics of the actual settling tank, an experimental model is established by scaling down by a factor of 1:12, and the geometric similarity error is less than 1%, ensuring that the experimental tank can accurately map the working characteristics of the actual equipment;

[0090] (2) By analyzing the physical process of the actual working condition, the Froude number (Fr), Reynolds number (Re) or Stokes number (Stk) is selected as the main similarity criterion, and the fluid viscosity, flow rate or particle size of the experimental model is adjusted to meet the similarity condition. In the experimental model, sensor arrays, turbidimeters and PIV particle image velocimetry devices are arranged to collect time series data including flow rate, particle concentration, temperature and flow field vorticity;

[0091] (3) By combining the physical information neural network (PINN) algorithm, the mapping relationship between the actual and experimental settling tank working characteristics is constructed. The physical law and machine learning are combined, and the basic physical laws such as fluid mechanics and particle settling are integrated into the training process, so that the network can not only fit the experimental data, but also ensure that the model output conforms to the known physical constraints. Reduce the demand for a large number of labeled data, improve the prediction accuracy and generalization ability of the model.

[0092] The present application is based on a modular detachable settling tank experimental device to explore the method of optimal equipment structure:

[0093] Firstly, a basic settling tank body integrating multi-modal monitoring functions is built, and an optical sensor observation window is provided at the top, and a detachable sludge collection bin is configured at the bottom; secondly, a structure parameter adjustable module group is installed at the key parts of the tank body, covering adjustable liquid collecting / distributing system (interface number, height distribution, geometric shape) to control inlet flow field distribution and fluid dynamics characteristics; adjustable heating unit (number of coil rings, pipe diameter specification, spatial layout) to optimize the internal temperature field gradient of the equipment, adjustable air flotation device (gas release head density, gas distribution pipe diameter parameter, installation height) to optimize the distribution of micro-bubbles and the capture efficiency of oil droplets, and adjustable inlet / outlet liquid pipeline system; During the experiment, a data capture sensor network is deployed synchronously, different structure parameter combinations are set up for comparison experiments, standard sewage samples (controllable oil content, temperature, flow rate) are used for dynamic testing, CFD numerical simulation is used as offline verification, high-speed camera technology is used to capture oil droplet coalescence behavior, and the system records core performance indicators such as oil phase floating rate, water outlet oil phase concentration, and sludge deposition morphology, and finally determines the optimal collaborative configuration scheme of each module through multi-dimensional parameter correlation analysis.

[0094] The modular design of the multi-modal perception reconfigurable settling system integrates PIV particle image velocimetry device, distributed optical fiber sensor, conductivity-turbidity multi-parameter sensor array and temperature dyeing technology, and builds a multi-dimensional physical field dynamic monitoring network, which can realize three-dimensional visualization reconstruction of internal flow characteristics of the settling tank and quantitative analysis of particle migration trajectory.

[0095] Under the action of the power unit, the oily sewage is pumped into the experimental settling tank by the centrifugal pump. The tank body of the experimental settling tank is made of high-transparency organic glass material, and the size ratio of the experimental settling tank is reserved. The experimental settling tank has a vault, and a plurality of test holes are arranged on the vault. The test tube extends into the experimental settling tank 13 through the test hole, and the part extending out of the vault is connected with the data acquisition and control system through a wire. A plurality of temperature sensors 23 are installed on the test tube. The upper and lower parts of the tank body are respectively provided with a settling tank liquid preparation port and a liquid collecting port. The settling tank liquid inlet, the settling tank liquid outlet and the heating coil 6 inlet are connected with the corresponding branch pipelines. Valves, pressure sensors and flow sensors are arranged on the branch pipelines, and an experimental pipeline is formed. The other end of the experimental pipeline is connected with a group of water bath heating boxes 15. A group of centrifugal pumps 16 are installed between the water bath heating boxes 15 and the experimental settling tank 13. By switching the valves, the different branch pipelines are connected with each other, the experimental process is switched, and different experimental conditions are constructed. The experimental settling tank chamber aims to simulate the actual flow environment of the oily sewage, and provides a place for studying the behavior of oil droplets under different conditions.

[0096] The present application focuses on the key similarity parameters such as Reynolds number and Froude number, and realizes accurate simulation of flow field characteristics. The settling system is equipped with advanced measuring devices and data acquisition systems, which can accurately measure key parameters such as velocity field and concentration distribution.

[0097] (1) Multi-dimensional parameter system construction and correlation modeling, taking the process parameter set (inlet and outlet flow, air flotation intensity, heating temperature) and the environmental parameter set (oil content, temperature, PH value, suspended solids concentration) as input variables, and taking the quality of the removed medium (water content of the removed oil / oil content of the removed water) and the settling efficiency (processing time) as decision variables, a mathematical model of parameter interaction is established;

[0098] (2) Data feature space reconstruction, principal component analysis is used to decouple the multi-dimensional monitoring data, the principal component load spectrum is extracted by covariance matrix decomposition, the key sensitive parameters are selected based on the variance contribution rate and the weight coefficients are quantified, and the nonlinear mapping relationship between process-environment parameters and settling efficiency is revealed;

[0099] (3) Intelligent identification of running state, a fuzzy comprehensive evaluation model is constructed, real-time monitoring data flow and historical database are integrated, parameter influence intensity is quantified by using membership function, dynamic diagnosis and risk prediction of running condition are realized, multi-step trend deduction is carried out by using recurrent neural network, and a comprehensive evaluation index system including stability index and energy efficiency index is established;

[0100] (4) Self-adaptive optimization and control, a parameter optimization mechanism is constructed based on reinforcement learning algorithm, process control parameters and environmental compensation coefficients are iteratively updated, multi-objective collaborative optimization is realized, a "monitoring-evaluation-optimization" closed-loop control architecture is formed, and finally an optimal parameter combination scheme meeting the decision constraints is output.

[0101] (5) Mapping the actual production equipment working characteristics, through analyzing the experimental parameters of the experimental settling tank, and combining the aforementioned physical information neural network (PINN) algorithm, the working characteristics of the actual equipment of the oil field are inversed. By combining the experimental data with the constraints of the physical model, the deep relationship between the experimental parameters and the working characteristics of the actual equipment is excavated. By inverting the working characteristics of the actual equipment of the oil field, theoretical support can be provided for equipment optimization design, operation strategy adjustment and performance improvement,

[0102] The present application realizes accurate regulation and autonomous optimization of the settling process by constructing a "physical device-sensing system-intelligent algorithm" three-in-one intelligent control architecture:

[0103] (1) Structural parameter optimization unit: Through the multi-parameter comprehensive fuzzy evaluation algorithm combined with orthogonal experimental design, the topological optimization of the component layout of the settling tank is carried out, and at the same time the three-dimensional flow field simulation results are generated by CFD for comparison.

[0104] (2) Process parameter adjustment unit: Install control components at each component of the settling tank, execute the MDP model and the PPO algorithm enhanced by transfer learning, dynamically calculate the feed and discharge flow (adjustment accuracy ±0.5m³ / h), temperature gradient (control resolution ±0.1℃) and flocculant dosage (measurement error <0.2%) etc., and execute the optimal decision output;

[0105] (3) Man-machine interaction unit: support three-dimensional visualization reconstruction and abnormal condition early warning of the settling process, and provide self-tuning PID control curve of process parameters;

[0106] (4) Backtracking verification unit: used to verify the effectiveness of the executed decision, through dynamic monitoring and analysis of the key parameters of the settling tank, the deviation between the theoretical settling effect and the actual effect is evaluated, and the machine learning model is iteratively optimized accordingly, to improve the accuracy and practicality of the system decision.

[0107] Example 2:

[0108] In this embodiment, step two is specifically:

[0109] Predict the performance of the actual tank (output X exp ) from the experimental tank data (input Y real ), and ensure that the prediction conforms to the physical law.

[0110] Step 1, build a physical information neural network architecture

[0111] Input layer: parameters obtained from experimental settling tank operation experiment; output layer: performance parameters of actual settling tank, particle settling efficiency, turbidity removal rate, etc. Define loss function:

[0112]

[0113] wherein: data For data matching term, take a small amount of actual settling tank calibration data when calculating Y real ;

[0114] ;

[0115] phys For physical constraint term;

[0116] scale For similarity criterion constraint term;

[0117]

[0118] Adjust the allocation of each term dynamically through the validation set.

[0119] Step 2, data preprocessing and model training

[0120] Enhance the experimental data, and perform dimensionless processing on the experimental parameters. At the same time, interpolate the experimental data to match the actual tank time axis. Only use experimental tank data to train the basic PINN, and learn F : X exp →Y exp ; Add 5%~10% actual tank monitoring data, freeze the first 3 layers of network, and only update the last 2 layers of weight to fine-tune the model and adjust the network parameters.

[0121] After training, perform model validation and uncertainty quantification. Use cross-validation method of time and space segmentation to ensure that the training / test set contains different flow stages (start, steady state, shutdown) and check the physical reasonableness of the experimental equipment. Use Monte Carlo Dropout (keep 20% Dropout during training); add confidence interval to the prediction output:

[0122]

[0123] Step 3, dynamic update of model

[0124] Fine-tune the model every 24h with new data (incremental learning, learning rate=1e-6). If the KL divergence detects data drift >5%, trigger retraining.

[0125] Through the above steps, the PINN algorithm not only establishes a high-precision mapping relationship between the experimental tank and the actual tank, but also ensures engineering interpretability through physical constraints, significantly reducing the dependence of traditional pure data-driven methods on a large amount of actual data. Example 3

[0126] The third step in this embodiment is specifically:

[0127] Based on the multi-factor orthogonal test and dynamic disturbance coupling analysis, the settlement system is optimized and controlled, and the working characteristics of the experimental equipment are intelligently analyzed and identified according to the real-time operating parameters, and the subsequent working characteristics are predicted. The specific implementation process of the experimental design and data acquisition technology is combined with the multi-parameter fuzzy comprehensive evaluation method model. The method realizes the precise optimization and robustness verification of the working parameters of the processing system by constructing a multi-parameter coupling matrix and an extreme working condition simulation test system. The specific implementation process is as follows:

[0128] Step 1, parameter definition and action mechanism

[0129] The physical meaning of the parameters is clear, and the action mechanism of each parameter on the settlement process is defined:

[0130] Water flow rate ( Q ): determines the residence time of wastewater in the tank, affects oil droplet coalescence and separation efficiency;

[0131]

[0132] Gas flotation intensity ( G ): through micro-bubble adsorption of oil droplets, accelerate floating, according to the modified equation of Stokes law:

[0133]

[0134] Heating temperature ( T ): reduces the viscosity of wastewater, promotes oil-water separation;

[0135]

[0136] Wastewater turbidity ( C TSS ): reflects the concentration of suspended solids, affects light scattering and settlement resistance;

[0137] Oil content ( C oil ): determines the oil droplet collision probability and coalescence efficiency (Smoluchowski equation).

[0138] Step 2, experimental design and data acquisition

[0139] After the experimental settling tank equipment is built and each data acquisition device is connected with the experimental instrument, the orthogonal experiment design is started, L25(5 6 ) orthogonal table is adopted, for example, 5 levels of working parameters such as water distribution flow (1 / 2 / 3 / 4 / 5 m³ / min), air floatation intensity (0.5 / 1.0 / 1.5 / 2.0 / 2.5 L / min), heating temperature (60 / 70 / 80 / 90 / 100℃) are selected, and the experiment is grouped according to the L25 orthogonal table. Then, dynamic disturbance experiments are carried out, for example, extreme working conditions such as oil content step change (from 5% to 10%) and water distribution flow fluctuation (±30%) are designed, to verify the system robustness. By using PIV and other data acquisition systems, the optimal parameters in 5 levels are compared and analyzed.

[0140] Based on the mass conservation and kinetic equation, a settling efficiency (η) h ) model is constructed:

[0141]

[0142] Subfunction expansion:

[0143] (1) Air floatation enhancement factor Γ:

[0144]

[0145] In the formula: k DAF Air floatation unit efficiency coefficient (related to the space position of the floatation unit); A b Total bubble surface area flux, m² / s; v b Bubble rising velocity, m / s; d 32 Sauter average bubble diameter, mm;

[0146] C oil,crit Oil phase critical emulsification concentration, mg / L.

[0147] (2) Thermal effect correction term T eff :

[0148]

[0149] In the formula: E a Apparent activation energy, J / mol; R Ideal gas constant; T ref Reference temperature, ℃;

[0150] fcoil — Heating coil spatial position impact function (positively correlated with thermal field uniformity)

[0151] (3) Spatial coordination factor ξ space :

[0152]

[0153] In the formula: P i — Spatial coordinates of water distribution port ( i =1), heating coil ( i =2), and flotation unit ( i =3); P opt — Optimal layout position of each component; L char — Characteristic length, taking 20% of the diameter of the tank body; w i — Weight coefficient.

[0154] (4) Foundation settlement rate η0

[0155]

[0156] In the formula: m(T) — Fluid viscosity, Pa·s; A settle — Effective settlement area, m²; Q max — Critical flow, exceeding which causes short flow.

[0157] After sorting:

[0158]

[0159] In the formula: p w — Density of water, kg / m 3 ; p o — Density of oil, kg / m 3 ; g — Acceleration of gravity, m / s 2 ; α — Kinetic coefficient, related to particle-fluid interaction and collision efficiency; Q — Liquid distribution flow, m 3 / s; k DAF — Flotation unit efficiency coefficient (related to the spatial position of the flotation unit); A b — Total bubble surface area flux, m² / s; vb — Bubble rising speed, m / s; d 32 —Sauter's average bubble diameter, mm; C oil —Oil content; C oil,crit —Critical emulsification concentration of the oil phase, mg / L; E a — Apparent activation energy, J / mol; R —Ideal gas constant; T — Heating coil temperature, °C; T ref —Reference temperature, °C; f coil ( x c , y c ,z c )——Spatial location is ( x c , y c ,z c The influence function of the spatial position of the heating coil at point (positively correlated with thermal field uniformity); P i —Water distribution point ( i =1), heating coil ( i =2), flotation unit ( i =3) spatial coordinates; P opt —Optimal layout positions for each component; L char —Characteristic length, taken as 20% of the tank diameter, in meters; w i —Weighting coefficients; m ( T — Fluid viscosity, Pa·s; A settlle —Effective settlement area, m²; Q max —Critical flow rate; exceeding this value will cause a short circuit, m 3 / s; C TSS —Wastewater turbidity; β —Oil phase correction coefficient, reflecting the effect of oil phase on sedimentation efficiency.

[0160] The multi-parameter fuzzy comprehensive evaluation model is used to train the model with experimental data, and the nonlinear coupling effect (such as the nonlinear improvement of gas flotation efficiency at high temperature) is quantified. 100 groups of experimental data (including 75 groups of normal working conditions and 25 groups of disturbance working conditions) are collected, and the training set, validation set and test set are divided according to the ratio of 7:2:1. The multi-parameter fuzzy comprehensive evaluation model is used to input the above five characteristic parameters and dynamic gradient characteristics, and the target variables are the oil content of the discharged water, the water content of the discharged oil and the settling time. After continuous iteration, the optimal experimental parameters are evaluated comprehensively.

[0161] Step 4, data dimension reduction

[0162] An unsupervised machine learning method of principal component analysis is selected to reduce the dimension of the data obtained by the experiment while maximizing the variance of the data. The main variables that best reflect the running state of the equipment are extracted from the original high-dimensional features, and the contribution of each feature in the principal component is sorted to screen out the parameters that have a greater impact on the running state of the equipment. Based on the results of the aforementioned working characteristic analysis and model establishment, the correlation between each working parameter, control parameter and settling tank efficiency can be determined, and the influence weight of each variable can be determined.

[0163] (1) Data standardization

[0164] Each feature is standardized to have a mean of 0 and a standard deviation of 1.

[0165]

[0166] In the formula: m — feature mean; s — standard deviation;

[0167] (2) Calculate the covariance matrix

[0168] Reflect the correlation between features, and the non-diagonal elements represent the covariance. The covariance matrix of the standardized data matrix X scaled is:

[0169]

[0170] 3) Feature decomposition

[0171] The covariance matrix is decomposed to obtain the eigenvalue l i and the corresponding eigenvector v i . The eigenvalue size represents the variance contribution of the corresponding principal component.

[0172] (4) Select the number of principal components

[0173] According to the variance contribution rate, the firstk The maximum eigenvalue is determined such that the cumulative contribution rate is > 95%.

[0174]

[0175] (5) Data projection

[0176] Before use k The projection matrix is composed of the first eigenvectors W The original data is projected into a new space to obtain new data after dimensionality reduction X pca :

[0177]

[0178] Step 5, state recognition and prediction

[0179] An intelligent prediction and recognition of the working condition is made by using a multi-parameter fuzzy comprehensive evaluation model to clarify the specific influence of each variable on the running state of the experimental settling tank. According to the current experimental settling tank running data, the running state of the experimental settling tank is judged in real time, and evaluation indexes are established. Combined with the relationship model and machine learning algorithm established before, the possible situation of the equipment in the current state is predicted, and the future running state changes with time. The current and future development trends are comprehensively considered to identify the equipment state and give a warning of the risk, and the evaluation results are given. After the prediction is completed, new data is included in the database.

[0180] (1) Determine the evaluation factor set

[0181] Input parameters: select the key parameters (such as temperature, pressure, flow rate, flotation intensity, etc.) that affect the running state of the settling tank to form the factor set:

[0182]

[0183] Target output: define the comment set of the running state: “normal”, “good settling effect”, “general settling effect”, “poor settling effect”, “very poor settling effect”, to form the comment set:

[0184]

[0185] (2) According to the weight of each parameter determined by the previous principal component analysis method, a fuzzy relationship matrix is established:

[0186]

[0187] (3) Fuzzy synthesis operation

[0188] The maximum membership degree method is used for fuzzy synthesis to obtain the comprehensive evaluation result B:

[0189]

[0190] Nonlinear and uncertain problems are handled by fuzzy logic, and the influence of multiple parameters on the state of the settling tank is determined, which is suitable for intelligent monitoring and prediction of complex industrial systems. The actual working characteristics of the sewage settling tank are mapped by combining the aforementioned PINN algorithm, and the working characteristics of the actual equipment on the experimental platform in the invention are analyzed.

[0191] Example 4:

[0192] In this embodiment, step four is specifically:

[0193] Based on the dynamic variable monitoring and Markov decision process (MDP) of the sewage settling tank intelligent control architecture, the purpose is to improve the settling efficiency and system stability by high-precision real-time monitoring and fine adjustment of process parameters. The core idea of the intelligent control architecture is to accurately monitor the dynamic variables directly related to the settling process, and combine the PPO (Proximal Policy Optimization) algorithm with migration enhancement to optimize the control decision, so as to realize the automatic adjustment of system performance.

[0194] 1. State space definition and dynamic variable collection: In order to realize the intelligent control of the settling process, the invention first defines the dynamic variables closely related to the settling efficiency as the state space of Markov decision process (MDP). These dynamic variables include but are not limited to: oil layer thickness, turbidity, temperature, pH value and suspended solids concentration, etc. These variables can fully reflect the changing state of the settling process and have an important influence on the settling efficiency. Therefore, accurate monitoring of the real-time changes of these variables is the key to the intelligent control architecture of the invention.

[0195] In order to ensure the high-precision real-time collection of state variables, the invention adopts the particle image velocimetry (PIV) and optical turbidimeter and multi-source sensor fusion technology. Through the data captured by the multi-modal sensor network, the fluid state can be monitored comprehensively, especially in the complex settling process environment, ensuring that the state variables can be collected in real time with high precision, thereby providing a reliable basis for subsequent decision-making.

[0196] 2. Action space mapping and fine control: In the intelligent control architecture, the invention maps adjustable parameters such as water bath electric heating rod power, sludge valve opening, centrifugal pump power, etc. as the action space in the Markov decision process. Each adjustable parameter corresponds to a control action, and the intelligent control system calculates the best control strategy according to the changes of the current state variables, so as to adjust these parameters to optimize the efficiency and stability in the settling process.

[0197] The present application particularly emphasizes the design of discrete-continuous mixed action space, which solves the problem that the traditional discrete action control method cannot meet the fine control requirements. In practical applications, many control parameters are not only discrete values, but also can be continuously adjusted within a certain range. For example, the power of the water bath electric heating rod, the opening of the mud valve, etc. can be finely adjusted according to the real-time state of the system. Through the design of discrete-continuous mixed action space, the control system can ensure control accuracy while being more flexible to adapt to various operating conditions, improving the response speed and control accuracy of the entire system.

[0198] 3. Markov decision process optimization strategy: the present application adopts Markov decision process (MDP) as the optimization framework of control decision. MDP can select the best control strategy according to the current state of the system by defining state space, action space and reward function. Through learning from historical data and real-time feedback, MDP continuously optimizes the control strategy, thereby maximizing the performance of the system. The introduction of MDP makes the intelligent control architecture of the present application have adaptability and learning ability, which can automatically adjust the control strategy under changing process conditions, thereby maintaining the efficiency and stability of the sedimentation process.

[0199] Step 2, Markov decision algorithm implementation

[0200] A multi-objective compound reward function is constructed, for example:

[0201]

[0202] wherein α , β , g are learnable dynamic weight coefficients to realize multi-objective optimization of sedimentation efficiency, reagent consumption and energy consumption.

[0203] The transfer enhanced PPO algorithm combining reinforcement learning framework and transfer learning technology is adopted to realize intelligent optimization of control decisions of oilfield sewage sedimentation tank.

[0204] First, the multi-source perception experimental data is fused, and the features are encoded.

[0205] (1) Basic PPO algorithm

[0206]

[0207]

[0208] In the formula: A t Advantage function at time t, i.e. the degree of good or bad of the advantage or deviation value function of action selection dTD error at time t, i.e., the difference between the current estimate and the actual observation; g Discount factor that weighs the importance of future rewards. γ ∈ [0, 1]; l Decay factor that determines the degree of influence of TD error at past time steps. A larger λ considers rewards over a longer time span; a smaller λ indicates that TD error is mainly influenced by the current time step; V ( s t ) Value function of state s t at time t, i.e., the expected total reward starting from state s t .

[0209] (2) Policy objective function

[0210]

[0211] where: L CLIP ( t ) Objective function of PPO, which represents the target of updating the policy using the clipping method under the current policy. The objective function optimizes the policy parameters t by maximizing this expression; E t Expectation over all time steps t. It represents the expectation over the trajectories (state-action sequences) changed by executing the policy; p t ( t ) Importance sampling ratio, which represents the ratio between the current policy and the old policy, specifically:

[0212]

[0213] where, p t ( a t ∣ s t ) is the probability of the current policy selecting action a t in state s t , clip old ( a t ∣ s t ) is the selection probability of the old policy. This ratio measures the degree of change of the current policy relative to the old policy.

[0214] p ( t t ( , 1 - e, 1 + e ) p ) Clipping operation, which limits(t) t p The value is within the range of [1-ε, 1+ε]. The effect of clipping is to prevent the policy from updating too much, reducing the risk of "overfitting". Specifically:

[0215] If (t) t p It is clipped (i.e., set to the boundary value) if it is outside the range [1-ε, 1+ε];

[0216] If (t) t e It will not be modified within this interval.

[0217] This operation is the key of PPO, which helps to control the magnitude of policy update, prevent the policy from being changed greatly, and thus improve the stability of training.

[0218] g -- A hyperparameter representing the size of the clipping range.

[0219] (3) Migration enhancement implementation

[0220] Train the base policy network in the source MDP decision scheme and the digital twin system, covering multi-condition simulation data. Then implement migration enhancement for PPO, add a migration-related regularization term to the original loss function of PPO, and calculate the migration loss term through KL divergence to obtain the adaptive loss function as follows:

[0221]

[0222] In the formula: L total -- Total objective function, i.e., the goal to be optimized in the PPO training process. It includes the PPO objective function itself (i.e., the part that limits the magnitude of policy update through clipping operation) and an additional KL divergence penalty term. L PPO -- PPO policy objective function; D KL ( π old ‖ π new ) - KL divergence, i.e., the difference between the new policy ( π new ) and the old policy ( π old ):

[0223]

[0224] After the model training is completed, the multi-MDP strategy fusion is performed, and the value evaluation of each candidate strategy under each MDP is calculated a i , and the value evaluation of each candidate strategy under each MDP is calculated

[0225]

[0226] At the same time, the process constraint filtering is applied to the integrated result to ensure the implementability of the preferred strategy, and the optimal decision is selected and executed.

[0227] Embodiment 5:

[0228] In this embodiment, step five is specifically:

[0229] After the implementation of the theoretical optimal decision, the real-time working parameters of the experimental settling tank are continuously monitored to evaluate the improvement effect of the decision execution on the operation efficiency. At the same time, the deviation between the theoretical settling result and the actual observation data is analyzed, and the model algorithm is iteratively optimized based on this to promote the continuous self-learning of the model in practice. Through this process, the prediction accuracy of the model is gradually improved, and the overall working efficiency of the experimental equipment after the execution of the decision is further improved.

[0230] After the execution of the decision proposed by the Markov model, the MDP algorithm will perform a function prediction on the decision value:

[0231]

[0232] In the formula: h The discount factor is 0.95; r The immediate reward function (weighted items such as oil content in effluent and energy consumption).

[0233] After the execution of the decision, the actual efficiency improvement of the experimental settling tank is calculated through the actual operation data:

[0234]

[0235] In the formula: h exp ( t ) is the composite efficiency index after the execution of the decision; t base ( t ) is the efficiency index before the execution of the decision;

[0236] T is the settling experiment time.

[0237] Efficiency difference quantitative analysis:

[0238]

[0239] Step 2, algorithm for dynamic compensation mechanism of deviation between theory and actual efficiency

[0240] Gradient Boosting Decision Tree (GBDT) realizes the continuous optimization of decision model through the closed-loop feedback of theoretical prediction and actual efficiency. Specifically, for the prediction and actual deviation generated by existing decision algorithms under idealized assumptions, a three-stage dynamic correction process is innovatively designed, which is characterized by:

[0241] (1) Establish a multi-dimensional feature space to tensor-fuse the parameter set of the theoretical execution strategy and the experimental environment variables;

[0242] (2) Use the Boosting integration strategy of GBDT to construct a nonlinear mapping model of theoretical-actual deviation through weighted residual approximation method;

[0243] (3) Based on the Bayesian optimization framework, an incremental learning mechanism is constructed to dynamically update the node splitting rules of the decision tree using real-time experimental data. This architecture forms a complete iterative optimization loop of "prediction-execution-feedback-correction" by constructing an interpretable feature importance matrix. The specific implementation process is as follows:

[0244] (1) Input feature set

[0245] Theoretical parameters: theoretical settling effect value output by fuzzy evaluation algorithm E theory ;

[0246] Real-time monitoring parameters: liquid level, sewage turbidity, oil-water separation rate, etc.

[0247] Historical deviation sequence: mean, variance and trend of deviation in sliding window (default length 30 periods);

[0248] Operating parameters: power of external heating rod of coil, speed of centrifugal pump, etc.

[0249] (2) Target variable: actual deviation value

[0250]

[0251] (3) Model establishment and optimization

[0252] Construct a regression-type GBDT model whose prediction target is to minimize the absolute error of deviation:

[0253]

[0254] Where: GBDT( X i ; t) —— represents the prediction of the GBDT model on the sample X i l — the set of parameters in GBDT; h — the regularization strength hyperparameter.

[0255] The optimal hyperparameter combination is determined by the grid search strategy: to effectively suppress the risk of overfitting, the maximum depth of the decision tree is limited to 7 layers; the learning rate of 0.15 is used to control the iteration step of the gradient descent process, balancing the convergence speed and optimization accuracy; the sub-sample sampling ratio of 0.8 is set to enhance the robustness of the model by random feature subset selection. This parameter combination is cross-validated, which significantly improves the model generalization ability while maintaining high prediction accuracy.

[0256] (1) Calculate the contribution of each feature to the bias prediction, select the top 20% features, and dynamically adjust the multi-parameter weight w i :

[0257]

[0258] where: w i — the weight of the i th element; β — learning rate, default value 0.01; P i — the i th standardized parameter; ∂ Δ / ∂P i — the gradient of the prediction result, indicating how the error changes with P i .

[0259] (2) Online incremental learning

[0260] When the bias analysis module triggers the optimization instruction, the following steps are executed:

[0261] a. Extract the new abnormal data samples from the time series database;

[0262] b. Update the GBDT model using the partial fitting algorithm:

[0263] c. If the model performance decreases, restart the global training.

[0264] (3) Collaborative reinforcement learning for model optimization

[0265] Map the key features output by GBDT to the Markov MDP state space, and reconstruct the decision model:

[0266] ​

[0267] Reward function modification:

[0268]

[0269] wherein ​ is an adaptive adjustable bias penalty coefficient.

[0270] Through the deep coupling of GBDT and reinforcement learning, continuous self-optimization for the decision-making process of the MDP model is realized, providing an innovative solution for intelligent control of wastewater treatment.

Claims

1. A method for intelligent prediction and optimized control of the operating status of oilfield wastewater settling tanks, characterized in that... Includes the following steps: Step 1: A modular design for reconfigurable settling system integrating multimodal sensing is used to collect time-series data, including flow rate, particle concentration, and flow field vorticity. The settling system integrates a PIV particle image velocimeter, a distributed fiber optic sensor, a conductivity-turbidity multi-parameter sensor array, and temperature staining technology to construct a multi-dimensional physical field dynamic monitoring network, thereby realizing three-dimensional visualization reconstruction of the flow characteristics inside the experimental settling tank and quantitative analysis of particle transport trajectories. Step 2: Combining the Physical Information Neural Network (PINN) algorithm, construct a mapping relationship between the working characteristics of the actual settling tank and the experimental settling tank: predict the performance of the actual settling tank using data from the experimental settling tank. Step 3: Identification and prediction of the operating status of the experimental settling tank; Establish a multi-parameter fuzzy comprehensive evaluation model, and use the multi-parameter fuzzy comprehensive evaluation model to evaluate the operating status of the experimental settling tank in real time based on the current operating data of the experimental settling tank, and establish evaluation indicators; then combine machine learning algorithms to predict the possible situations that may occur when the settling system continues to operate under the current state, as well as the changes in the future operating status over time. Step four, the settling system operates with intelligent decision-making and automatic control; by accurately monitoring dynamic variables directly related to the settling process, a Markov decision process (MDP) is used as the optimization framework for control decisions, and a migration-enhanced PPO algorithm is combined to optimize control decisions, thereby achieving automatic adjustment of system performance. The dynamic variables include oil layer thickness, turbidity, temperature, pH value, and suspended solids concentration. Step 5: Decision verification and dynamic correction closed-loop optimization control of the Markov Decision Process (MDP). Through deep coupling of GBDT and reinforcement learning, continuous self-optimization of the MDP decision process is achieved.

2. The intelligent prediction and optimized control method for the operating status of oilfield wastewater settling tanks according to claim 1, characterized in that: The multi-parameter fuzzy comprehensive evaluation model in step three is as follows: ； In the formula: ρ w The density of water; ρ o The density of the oil; g It is the acceleration due to gravity; α For kinetic coefficients; Q This refers to the solution flow rate; k DAF The efficiency coefficient of the air flotation unit; A b The total surface area flux of the bubbles; v b This refers to the rising speed of the bubble; d 32 This represents the average bubble diameter of Sauter; C oil Oil content; C oil,crit This is the critical emulsification concentration of the oil phase; E a It is the apparent activation energy; R It is the ideal gas constant; T To heat the coil temperature; T ref For reference temperature; f coil ( x c , y c ,z c ) is the spatial location ( x c , y c ,z c The influence function of the spatial position of the heating coil at point ( ); P i For water distribution outlet i =1. Heating coil i =2. Flotation Unit i =3 spatial coordinates; P opt To determine the optimal layout positions for each component; L char The characteristic length is taken as 20% of the tank diameter; w i These are the weighting coefficients; μ ( T ) represents the fluid viscosity; A settlle For effective settlement area; Q max The critical flow rate; C TSS Turbidity of wastewater; β This is the oil phase correction factor.

3. The intelligent prediction and optimized control method for the operating status of oilfield wastewater settling tanks according to claim 2, characterized in that: The modular design reconfigurable settlement system integrating multimodal perception includes an experimental module, a data acquisition module, and a dataset parsing and control module. These modules work together to form a complete experimental feedback loop. The experimental module includes an experimental settling tank, a power unit, a heating unit, and a flotation unit. The experimental settling tank is equipped with a central column tube and a central reaction cylinder. The flotation unit is a dissolved air pump, which uses a multiphase dissolved air pump to draw in air and wastewater together. The impeller of the dissolved air pump cuts the wastewater and air into fine foam, making them fully mixed. At the same time, the high pressure generated by the high-speed rotation of the impeller fully dissolves the drawn-in gas into the water, and the combined effect makes it form stable dissolved air water. When this dissolved air water is released at normal pressure through a pressure reducing valve, it produces tiny bubbles with a diameter between 20 and 50 μm. The data acquisition module integrates a PIV particle image velocimetry device, a CCD camera array, a multispectral turbidimeter, and a distributed sensor group. The PIV particle image velocimetry device and the CCD camera are synchronously triggered by laser pulses to achieve submicron-level dynamic analysis of the oil droplet particle size distribution. The multispectral turbidimeter uses multi-wavelength scattered light intensity detection technology combined with an embedded temperature and pressure compensation algorithm to achieve online calibration of turbidity changes during the settling process. The distributed sensor group includes a micro differential pressure transmitter and an array of temperature probes, forming a multi-point physical parameter monitoring network for the flow field inside the settling tank. The data parsing and control module is equipped with a motion feature extraction unit, a multi-physics coupling analysis unit, and a settlement efficiency evaluation unit; The motion feature extraction unit uses an improved PTV particle tracking algorithm to perform Gaussian fitting and noise reduction on continuous frame images, and calculates the oil droplet velocity field, acceleration field and fractal dimension of motion trajectory. The multiphysics coupling analysis unit analyzes the dynamic characteristics of oil droplet coalescence and breakup by establishing a correlation matrix model of turbidity-particle size-flow velocity; the sedimentation efficiency evaluation unit constructs a particle group sedimentation prediction model based on the Fokker-Planck equation and outputs key indicators such as sedimentation rate and interface clarity.

4. The intelligent prediction and optimized control method for the operating status of oilfield wastewater settling tanks according to claim 3, characterized in that: Step two specifically involves: Step 1: Construct a physical information neural network architecture. Input layer: Parameters obtained from the experimental settling tank operation; Output layer: Actual settling tank performance parameters, including particle settling efficiency and turbidity removal rate. Define the loss function: ； In the formula: data For data matching, a small amount of actual settling tank calibration data is used in the calculation. Y real ; ； phys For physical constraints; scale For similarity criterion constraints; The allocation of each item is dynamically adjusted through the validation set; Step 2: Time series data preprocessing and training of the PINN neural network model for physical information; The time series data is augmented by making it dimensionless and interpolating it to match the actual tank time axis. The experimental tank settling time series data is used to train the basic PINN to establish a high-precision mapping relationship between the experimental settling tank and the actual settling tank. Step 3: Dynamically update and continuously fine-tune the model with new data to optimize the PINN physical information neural network model.

5. The intelligent prediction and optimized control method for the operating status of oilfield wastewater settling tanks according to claim 4, characterized in that: The specific method for step three is as follows: Step 1: Obtain data, including water flow rate, flotation intensity, heating temperature, wastewater turbidity, and oil content; Step 2: Establish a multi-parameter fuzzy comprehensive evaluation model; Settlement efficiency is constructed based on mass conservation and kinetic equations. η Model: ； Subfunction expansion: (1) Air flotation enhancement factor Γ: ； In the formula: k DAF The efficiency coefficient of the non-air flotation unit is related to the spatial location of the flotation unit; A b The total surface area flux of the bubbles is m² / s; v b This refers to the rising speed of the bubble; d 32 This represents the average bubble diameter of Sauter; C oil,crit This is the critical emulsification concentration of the oil phase; (2) Thermal effect correction term T eff : In the formula: E a It is the apparent activation energy; R It is the ideal gas constant; T ref For reference temperature; f coil ( x c , y c ,z c ) is the spatial location ( x c , y c ,z c The influence function of the spatial position of the heating coil at point ( ); (3) Spatial Coordination Factor ξ space : In the formula: P i For water distribution outlet ( i =1), heating coil ( i =2), flotation unit ( i =3) spatial coordinates; P opt To determine the optimal layout positions for each component; L char The characteristic length is taken as 20% of the tank diameter; w i These are the weighting coefficients; (4) Foundation settlement rate η0 ； In the formula: μ(T) For fluid viscosity; A settle For effective settlement area; Q max The critical flow rate; Summarized as follows: ； Using the data obtained in step 1, a multi-parameter fuzzy comprehensive evaluation model is trained. The multi-parameter fuzzy comprehensive evaluation model is used to input the above-mentioned water distribution flow rate, air flotation intensity, heating temperature, sewage turbidity, oil content and dynamic gradient characteristics. The target variables are the oil content of the dewatering water, the water content of the dewatering water and the settling time. After continuous iteration, the optimal experimental parameters are comprehensively evaluated. Step 3: Data dimensionality reduction, determine the correlation between each time series data and the efficiency of the settling tank, and determine the influence weight of each time series data. Step 4, State Identification and Prediction.

Images