Real-time energy and LSTM oil well working fluid level depth measuring method

Abstract

The invention relates to the technical field of liquid level measurement, in particular to a real-time energy and LSTM (Long Short Term Memory) oil well working fluid level depth measurement method, which comprises the following steps of: acquiring and classifying real-time data by adopting an energy monitoring device, analyzing and screening the data according to a consumption threshold value, marking and generating an energy state classification result. According to the invention, real-time measurement and dynamic prediction are carried out on the oil well liquid level depth through the long and short term memory artificial neural network, complex dynamic data of the oil well liquid level are effectively processed, screening and difference analysis of a liquid level change area are realized through accumulation calculation and segmentation processing, and liquid level depth spatial features are generated. The accuracy and real-time performance of liquid level depth prediction are improved, through the space-time diagram convolutional network, the space-time correlation characteristics of the liquid level depth are matched with actual operation data, the problems of accurate judgment of the liquid level change trend and dynamic model generation are solved, and the overall operation efficiency and safety of an oil well are optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of liquid level measurement, and in particular to a method for measuring the dynamic liquid level depth of an oil well using real-time energy and LSTM. Background Art

[0002] The field of liquid level measurement technology is used to measure and monitor the depth and flow status of liquids. In oil well engineering, liquid level measurement is an important means to ensure the normal operation of oil wells, improve oil production efficiency and avoid accidents. Liquid level measurement technology can be implemented in a variety of ways, including using sound waves, radar or pressure sensors for real-time monitoring and measurement to accurately obtain liquid depth and flow information.

[0003] The real-time energy and LSTM oil well dynamic liquid level depth measurement method aims to accurately predict and measure the depth of the oil well dynamic liquid level through an intelligent algorithm. The purpose is to use the long short-term memory artificial neural network model to process dynamic data in a complex oil well environment, achieve high-precision prediction of oil well liquid level changes, improve the efficiency and safety of oil well operation, achieve timely response and effective control of liquid level changes, and optimize the overall operation of the oil well.

[0004] Traditional liquid level measurement methods lack effective intelligent algorithms and rely on real-time monitoring of a single sensor and simple logical judgment. They cannot fully cope with the complex dynamic changes in the oil well environment. When dealing with multi-variable interactions, they have great limitations and are often unable to accurately predict liquid level changes, resulting in lags and errors in liquid level monitoring, which increases the risk of oil well operations. When multiple factors work together, traditional methods are difficult to effectively capture changes in the liquid level, affecting the stability and safety of oil well operations. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a real-time energy and LSTM oil well dynamic liquid level depth measurement method.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: a real-time energy and LSTM oil well dynamic liquid level depth measurement method, comprising the following steps: Step 1: Use energy monitoring devices to collect and classify real-time data, analyze and filter the data according to consumption thresholds, and mark them to generate energy status classification results; Step 2: Based on the energy state classification result, a long short-term memory artificial neural network is used to obtain real-time measurement data of the liquid level depth of the oil well, and the data is accumulated and calculated and segmented, and the liquid level change area is screened, and a difference analysis is performed at the same time to determine the key area and generate the liquid level depth space feature; Step 3: Based on the spatial characteristics of the liquid level depth, interactive analysis is performed through the liquid level depth data and temperature fluctuations, detailed calculations are performed on the temperature changes under the oil well operation mode, and the main change trends of the liquid level depth are determined. Combined with the spatial and temporal data characteristics, a spatiotemporal correlation model of the liquid level depth is generated; Step 4: Based on the spatiotemporal correlation model of the liquid level depth, a spatiotemporal graph convolutional network is used to match real-time data between the energy monitoring device and the liquid level depth sensor, determine the liquid level change under the current oil well operation mode, calculate the time series data, compare the liquid level change trends under different states, and generate a real-time liquid level depth dynamic model; Step 5: Based on the real-time liquid level dynamic model, a liquid level sensor is used to record the data of the measuring points in detail, and abnormal points are identified according to the recorded data, and the pressure conduction between the measuring points is adjusted, and the distribution of the measuring points is optimized through the pressure change to generate an optimization plan for the liquid level measuring points; Step 6: Based on the optimization scheme of the liquid level depth measurement points, the liquid level monitoring device is used to integrate and process the optimized measurement point data, and the integrated data is used for verification, and the data of each measurement point is cross-compared and analyzed to generate the full-well liquid level depth monitoring results.

[0007] As a further solution of the present invention, the spatial characteristics of the liquid level depth include key areas of liquid level depth change, liquid level fluctuation amplitude within the area, and associated spatial positions; the real-time liquid level depth dynamic model includes current energy state, liquid level depth change curve, and predicted trend; the liquid level depth measurement point optimization scheme includes layout adjustment of measurement points, optimization of pressure conduction path, and precision of liquid level depth monitoring; the full-well liquid level depth monitoring results include comprehensive data of liquid level depth, spatiotemporal distribution diagram of each measurement point, and liquid level change prediction results.

[0008] As a further solution of the present invention, the specific steps of generating the energy state classification result are: Use energy monitoring devices to collect and classify real-time data, screen the collected energy consumption data, filter according to the set energy threshold, mark the classification of different energy intervals, and annotate the data labels to generate energy interval classification data; Based on the energy interval classification data, the marked energy data is analyzed, the data of different intervals are classified by segmentation operation, and comprehensive screening is performed to obtain marked energy data; Based on the marked energy data, the data under various oil well operation modes are classified and organized, and merged into a classification index table to generate an energy state classification result.

[0009] As a further solution of the present invention, the specific steps of generating the liquid level depth spatial feature are: Based on the energy state classification result, liquid level depth data is obtained, and a data processor is used to perform cumulative calculation, the liquid level depth data is grouped by time period, and the liquid level depth data is classified and integrated according to the time period to generate segmented liquid level data; Based on the segmented liquid level data, a long short-term memory artificial neural network is used to perform regional difference analysis on the data of different time periods, determine the significant area of ​​liquid level depth change by detecting the change amplitude, aggregate the data with obvious change amplitude, and generate the liquid level change area; Based on the liquid level change area, the key liquid level depth change position in each area is calculated, the spatial difference is compared, the information of the key area is integrated, the corresponding liquid level depth map is established, and the liquid level depth spatial feature is generated.

[0010] As a further solution of the present invention, the long short-term memory artificial neural network is according to the formula:

[0011] Among them: Among them, For the moment The implicit state of is the input segmented liquid level data, is the volatility of the liquid level data, is the environmental interference parameter, is the weight matrix, is an implicit state, is the liquid level data currently input, is the volatility, is the environmental interference parameter, is the adaptive weight coefficient, is the bias vector, is the timing dependent deviation, is the activation function.

[0012] As a further solution of the present invention, the specific steps of generating the liquid level depth spatiotemporal correlation model are: Based on the liquid level depth spatial characteristics, the liquid level depth data is extracted and the temperature fluctuation information is collected, and interactive processing is performed at the same time to generate a temperature and liquid level correlation data set; Based on the temperature and liquid level correlation data set, temperature change calculation is performed, liquid level depth change trend is analyzed, and liquid level depth trend analysis results are generated; Based on the liquid level depth trend analysis results, combined with time series data and spatial characteristics, a liquid level depth spatiotemporal correlation model is established to generate a liquid level depth spatiotemporal correlation model.

[0013] As a further solution of the present invention, the specific steps of generating the real-time liquid level depth dynamic model are: Based on the spatiotemporal correlation model of liquid level depth, an energy monitoring device and a liquid level depth sensor are used to perform real-time data matching and liquid level depth data collection through a spatiotemporal graph convolutional network to generate the current liquid level depth state; Based on the current liquid level depth state, calculate the liquid level depth change trend, compare the change situation under the oil well operation mode, and generate liquid level change trend data; Based on the liquid level change trend data, a real-time liquid level depth dynamic model is updated and generated to obtain a real-time liquid level depth dynamic model.

[0014] As a further solution of the present invention, the spatiotemporal graph convolutional network is based on the formula:

[0015] in: The updated liquid level state matrix, is a nonlinear activation function, is the dynamic adjustment coefficient, is the adjacency matrix Power, is the initial liquid level depth data matrix, For the The convolution kernel weight matrix of order, is the time attenuation coefficient, For the time difference, is the time matrix, is the environmental pressure coefficient, is the pressure matrix.

[0016] As a further solution of the present invention, the specific steps of generating the optimization solution for the liquid level depth measurement points are: Based on the real-time liquid level depth dynamic model, a liquid level depth sensor is used to record data of the measuring point, and the data is screened by recording the data multiple times, and the effective data is integrated to generate a measuring point data set; Based on the measurement point data set, abnormal points are identified by detecting abnormal fluctuations in the data, pressure conduction between the measurement points is adjusted, the adjustment effect is confirmed by re-measurement, and a measurement point distribution optimization result is generated; Based on the measurement point distribution optimization result, the optimized distribution is verified by checking the pressure changes of the measurement points one by one, and the measurement point data is recorded again to generate an optimization plan for the liquid level depth measurement points.

[0017] As a further solution of the present invention, the specific steps of generating the full well liquid level depth monitoring result are: Based on the optimization scheme for the liquid level depth measurement points, the optimized measurement point data are integrated by using a liquid level monitoring device, and a comprehensive data set is generated by successively superimposing the measurement data to generate an optimized measurement point data set; Based on the optimized measurement point data set, by comparing the data change trends between different measurement points, cross-comparison and analysis are performed to unify the relevant data and generate measurement point cross-analysis results; Based on the cross analysis results of the measuring points, the data of all measuring points are integrated to form the final liquid level depth distribution, generate the full-well liquid level depth monitoring result, and obtain the full-well liquid level depth monitoring result.

[0018] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, the long short-term memory artificial neural network is used to measure and dynamically predict the liquid level depth in the oil well in real time, and the complex dynamic data of the liquid level in the oil well is effectively processed. Through cumulative calculation and segmented processing, the screening and difference analysis of the liquid level change area are realized, and the spatial characteristics of the liquid level depth are generated. The accuracy and real-time performance of the liquid level depth prediction are improved. Through the spatiotemporal graph convolution network, the spatiotemporal correlation characteristics of the liquid level depth are matched with the actual operation data, and the problems of accurate judgment of the liquid level change trend and dynamic model generation are solved, ensuring timely response and adjustment to the liquid level depth changes under different operation modes, and optimizing the overall operation efficiency and safety of the oil well. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the main steps of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0021] Embodiment 1 See also Figure 1 The present invention provides a technical solution: a real-time energy and LSTM oil well dynamic liquid level depth measurement method, comprising the following steps: Step 1: Use energy monitoring devices to collect and classify real-time data, analyze and filter the data according to consumption thresholds, and mark them to generate energy status classification results; Step 2: Based on the energy state classification results, a long short-term memory artificial neural network is used to obtain real-time measurement data of the liquid level depth of the oil well, accumulate and calculate the data, process the data in segments, and screen the liquid level change area. At the same time, a difference analysis is performed to determine the key area and generate the liquid level depth spatial characteristics; Step 3: Based on the spatial characteristics of the liquid level depth, the liquid level depth data and temperature fluctuations are interactively analyzed, and the temperature changes under the oil well operation mode are calculated in detail. The main change trend of the liquid level depth is determined, and the spatiotemporal correlation model of the liquid level depth is generated by combining the spatial and temporal data characteristics; Step 4: Based on the spatiotemporal correlation model of liquid level depth, a spatiotemporal graph convolutional network is used to match real-time data between the energy monitoring device and the liquid level depth sensor, determine the liquid level change under the current oil well operation mode, calculate the time series data, compare the liquid level change trends under different states, and generate a real-time liquid level depth dynamic model; Step 5: Based on the real-time liquid level dynamic model, the liquid level sensor is used to record the data of the measuring points in detail, and the abnormal points are identified according to the recorded data, and the pressure conduction between the measuring points is adjusted. The distribution of the measuring points is optimized through the pressure change, and the optimization plan of the liquid level measurement points is generated; Step 6: Based on the optimization plan of the liquid level depth measurement points, the liquid level monitoring device is used to integrate and process the optimized measurement point data, and the integrated data is used for verification. The data of each measurement point are cross-compared and analyzed to generate the full-well liquid level depth monitoring results.

[0022] The spatial characteristics of the liquid level depth include the key areas of liquid level depth change, the liquid level fluctuation amplitude in the area, and the associated spatial position. The real-time liquid level depth dynamic model includes the current energy state, the liquid level depth change curve, and the predicted trend. The optimization plan for the liquid level depth measurement points includes the layout adjustment of the measurement points, the optimization of the pressure transmission path, and the precision of the liquid level depth monitoring. The full-well liquid level depth monitoring results include the comprehensive data of the liquid level depth, the spatiotemporal distribution map of each measurement point, and the prediction results of the liquid level change.

[0023] The specific steps to generate energy state classification results are: Use energy monitoring devices to collect and classify real-time data, screen the collected energy consumption data, filter according to the set energy threshold, mark the classification of different energy intervals, and annotate the data labels to generate energy interval classification data; Based on the energy interval classification data, the marked energy data is analyzed, the data of different intervals are classified by segmentation operation, and comprehensive screening is performed to obtain the marked energy data; Based on the marked energy data, the data under various oil well operation modes are classified and organized, and merged into a classification index table to generate energy state classification results; Based on the energy consumption data collected by the energy monitoring device, a data classification algorithm is used, specifically the K-Means clustering algorithm, with parameters including the number of cluster centers k=3, the maximum number of iterations max_iter=300, and the initial cluster center selection method of k-means++. The collected data is classified, and the Euclidean distance between each data point and the cluster center is calculated to assign the data point to the nearest cluster center, and the energy consumption data is clustered and marked. After the classification is completed, according to the set energy threshold, specifically the high energy consumption threshold is 1000J, the medium energy consumption threshold is 500J, and the low energy consumption threshold is 100J, the classified data is threshold filtered, and during filtering, the data exceeding the high energy consumption threshold is marked as high energy consumption, the data below the low energy consumption threshold is marked as low energy consumption, and the remaining data is marked as medium energy consumption, generating energy interval classification data; Based on the energy interval classification data, the marked energy data is analyzed, and a segmentation operation algorithm is used, specifically a segmented linear regression algorithm. The number of segments is set to 3, and the number of data points in each segment is set to 1000 points. The regression model adopts a linear model, and the loss function is the least squares method. The data in different intervals are classified, and the energy data is divided into 3 segments according to the number of data points. Each segment of data is fitted with a linear regression model, and the regression coefficient after fitting is analyzed. The regression coefficients of each segment of data are comprehensively screened, and the segment data with a goodness of fit R² value greater than 0.8 are retained, and the data segments with low goodness of fit are eliminated to obtain the marked energy data; Based on the marked energy data, the data under various oil well operation modes are classified and organized. The data merging algorithm is used, specifically an index-based merging algorithm. The merging mode is set to inner join, the index field is the oil well ID, and the merging strategy is set to deduplication merging. The data under different modes are indexed and matched. When matching, the data is inner joined and merged through the index field oil well ID. After removing duplicate data, the merged data is organized into a classification index table to generate the energy status classification result.

[0024] The specific steps for generating the liquid surface depth spatial features are: Based on the energy state classification result, the liquid level depth data is obtained, and the data processor is used for cumulative calculation, the liquid level depth data is grouped by time period, and the liquid level depth data is classified and integrated according to the time period to generate segmented liquid level data; Based on the segmented liquid level data, a long short-term memory artificial neural network is used to analyze the regional differences of data in different time periods. The significant areas of liquid level depth change are determined by detecting the change amplitude. The data with obvious change amplitude are aggregated to generate the liquid level change area. Based on the liquid level change area, calculate the key liquid level depth change position in each area, compare the spatial differences, integrate the information of the key areas, establish the corresponding liquid level depth map, and generate the liquid level depth spatial characteristics; Based on the energy state classification result, the liquid level depth data is obtained, and a data processor is used for cumulative calculation, specifically a cumulative summation algorithm, with parameters set as cumulative summation for each time period, the time period length is 1 hour, and the data unit is meter. The liquid level depth data is grouped by time period, and the data is divided according to the set time period length. Then, the data in each time period is cumulatively summed, and the liquid level depth data of each time period is integrated into a single data point, and the integration is classified by time period to generate segmented liquid level data; Based on the segmented liquid level data, a long short-term memory artificial neural network is used. The parameters are set as 128 neurons in the input layer, 64 neurons in the hidden layer, 1 neuron in the output layer, tanh as the activation function, MSE as the loss function, Adam optimizer as the optimization algorithm, and a learning rate of 0.001. Regional difference analysis is performed on data in different time periods. The segmented liquid level data is input into the long short-term memory artificial neural network. The time series data is processed through the network to detect the change amplitude of the liquid level depth, and the significant area of ​​the liquid level depth change is determined according to the change amplitude. Then, the data with obvious change amplitude is aggregated, and the significant change data in the same area are merged to generate the liquid level change area. Based on the liquid level change area, the key liquid level depth change position in each area is calculated. The spatial difference analysis method is adopted, specifically the spatial interpolation algorithm. The parameters are set to an interpolation radius of 10 meters and the interpolation method is Kriging interpolation. The depth data of each liquid level change area is spatially interpolated to determine the key liquid level depth change position. By comparing the spatial differences between different areas, the information of the key change position of each area is integrated, and the corresponding liquid level depth map is established to generate the liquid level depth spatial characteristics.

[0025] Long short-term memory artificial neural network, according to the formula:

[0026] Among them: Among them, For the moment The implicit state of is the input segmented liquid level data, is the volatility of the liquid level data, is the environmental interference parameter, is the weight matrix, is an implicit state, is the liquid level data currently input, is the volatility, is the environmental interference parameter, is the adaptive weight coefficient, is the bias vector, is the timing dependent deviation, is the activation function; Execution process: First, receive the input data at the current moment , that is, the measured oil well liquid depth data, while considering the implicit state of the previous moment , represents the memory and integration of the previous liquid level data, and then, through the weight matrix Perform a linear combination and then introduce the liquid level fluctuation rate , reflects the intensity of local changes in liquid depth, captures short-term drastic fluctuations caused by pumping or other operations, and introduces environmental interference parameters , correct the influence of external environmental factors on the liquid depth data, and add an adaptive weight coefficient , adjust the bias vector The influence of the model is ensured to be adaptable under different measurement conditions. Finally, the timing dependency deviation is considered. , used to compensate for systematic deviations caused by changes in time sequence or operating conditions, the activation function Perform nonlinear transformation on the calculation results to generate the implicit state at the current moment .

[0027] The specific steps for generating the spatiotemporal correlation model of liquid level depth are as follows: Based on the spatial characteristics of liquid level depth, the liquid level depth data is extracted and the temperature fluctuation information is collected. At the same time, interactive processing is performed to generate a data set related to temperature and liquid level. Based on the temperature and liquid level correlation data set, calculate the temperature change, analyze the liquid level depth change trend, and generate the liquid level depth trend analysis results; Based on the results of the liquid level depth trend analysis, combined with the time series data and spatial characteristics, a liquid level depth spatiotemporal correlation model is established to generate a liquid level depth spatiotemporal correlation model; Based on the spatial characteristics of the liquid level depth, the liquid level depth data is extracted and the temperature fluctuation information is collected. An interactive processing algorithm is used, specifically a multiple linear regression algorithm. The parameters are set as the independent variables of the liquid level depth and temperature, and the dependent variable is the interactive influence value of the temperature and the liquid level depth. The regression coefficient is estimated by the least squares method, and the learning rate is 0.01. The liquid level depth data and the temperature fluctuation information are aligned, and then the correlation between the liquid level depth and the temperature is calculated by the multiple linear regression algorithm to generate a temperature and liquid level correlation data set. Based on the temperature and liquid level correlation data set, a temperature change calculation method is adopted, specifically, the exponential smoothing method, with the parameter set to the smoothing coefficient α=0.3 and the time window of 1 hour. The temperature change is calculated, the temperature data in the temperature and liquid level correlation data set is grouped into time series, and then the temperature change in each time window is exponentially smoothed. By calculating the temperature smoothing value in each time window, the liquid level depth change trend is analyzed, and the liquid level depth trend analysis results are generated; Based on the results of liquid level depth trend analysis, combined with time series data and spatial features, a spatiotemporal correlation modeling algorithm was adopted. The parameters were set as 3 graph convolution layers, 3×3 convolution kernel size, ReLU activation function, Adam optimizer optimization algorithm, and learning rate 0.001. The results of liquid level depth trend analysis were integrated with the time series data, the data was processed, the spatiotemporal features were extracted, and the correlation modeling was performed in combination with the spatial graph structure information to generate a spatiotemporal correlation model of liquid level depth.

[0028] The specific steps to generate a real-time liquid level depth dynamic model are: Based on the spatiotemporal correlation model of liquid level depth, an energy monitoring device and a liquid level depth sensor are used to perform real-time data matching and liquid level depth data collection through a spatiotemporal graph convolutional network to generate the current liquid level depth status. Based on the current liquid level depth status, calculate the liquid level depth change trend, compare the changes under the oil well operation mode, and generate liquid level change trend data; Based on the liquid level change trend data, a real-time liquid level depth dynamic model is updated and generated to obtain a real-time liquid level depth dynamic model; Based on the spatiotemporal correlation model of liquid level depth, an energy monitoring device and a liquid level depth sensor are used to perform real-time data matching and liquid level depth data collection through a spatiotemporal graph convolutional network. Specifically, the spatiotemporal graph convolutional network GCN is used. The parameters are set to 3 convolution layers and 3×3 convolution kernel size. The input data is the data collected by the energy monitoring device and the liquid level depth sensor. The optimization algorithm is the Adam optimizer, and the learning rate is set to 0.001. The spatiotemporal features of the data are extracted through the graph convolution layer, and the correlation between the liquid level depth and the energy state is matched through the network layer. The matched data is merged and processed to generate the current liquid level depth state. Based on the current liquid level depth state, a trend analysis algorithm is used, specifically a time series model, with parameters set as autoregressive order p=2, differential order d=1, and moving average order q=2. The model optimizes parameters through the maximum likelihood estimation MLE method, calculates the liquid level depth change trend, performs time series modeling on the current liquid level depth state, calculates the future change trend of the liquid level depth, compares and analyzes the calculation results with historical data under different operating modes of the oil well, and generates liquid level change trend data; Based on the liquid level change trend data, a real-time dynamic modeling method is adopted. The parameters are set as 64 neurons in the input layer, 32 neurons in the hidden layer, 1 neuron in the output layer, 10 steps in the time step, and the optimization algorithm is Adam optimizer. The learning rate is 0.001. The liquid level change trend data is used as input, the parameters are updated in real time, and the model weights are adjusted through the dynamic learning mechanism to generate a real-time liquid level depth dynamic model.

[0029] Spatiotemporal graph convolutional network, according to the formula:

[0030] in: The updated liquid level state matrix, is a nonlinear activation function, is the dynamic adjustment coefficient, is the adjacency matrix Power, is the initial liquid level depth data matrix, For the The convolution kernel weight matrix of order, is the time attenuation coefficient, For the time difference, is the time matrix, is the environmental pressure coefficient, is the pressure matrix; Execution process: First, collect the initial liquid level depth data through the liquid level depth sensor to generate the initial liquid level depth data matrix , construct the spatiotemporal correlation model of the oil well liquid level depth, and obtain the adjacency matrix , and calculate its first Power , which represents the correlation between the liquid level depth at different time points and spatial positions, by dynamically adjusting the coefficient Adjust the influence of different order adjacency matrices on the final measurement results and dynamically adjust the coefficients Determined by experiment or cross-validation, combined with the time matrix and the time decay coefficient , calculate the attenuation effect of liquid depth over time, and introduce the environmental pressure matrix and environmental pressure coefficient , considering the effect of the ambient pressure around the oil well on the liquid level depth, the ambient pressure coefficient It is obtained by fitting the experimental data. Finally, through the nonlinear activation function Process the calculation results to generate an updated liquid level depth state matrix .

[0031] The specific steps to generate the optimization solution for the liquid level depth measurement points are: Based on the real-time liquid level dynamic model, the liquid level sensor is used to record the data of the measuring point. The data is screened through multiple recordings, and the effective data is integrated to generate a measuring point data set. Based on the measurement point data set, abnormal points are identified by detecting abnormal fluctuations in the data, the pressure conduction between the measurement points is adjusted, the adjustment effect is confirmed by re-measurement, and the measurement point distribution optimization result is generated; Based on the measurement point distribution optimization results, the optimized distribution is verified by checking the pressure changes of the measurement points one by one, and the measurement point data is recorded again to generate the optimization plan for the liquid level depth measurement points; Based on the real-time liquid level dynamic model, a liquid level sensor is used to record the data of the measuring point, and the data is screened through multiple recorded data. Specifically, a time series data screening algorithm is used, and the parameters are set as a data sampling frequency of once per second. The data screening condition is to remove abnormal data points with a variance greater than 0.5, and the data recorded each time is sorted in time series. The multiple recorded data are compared and analyzed, and the abnormal data points are removed through the screening algorithm. The valid data with a small variance is retained, and then the screened valid data is integrated into a unified data set to generate a measurement point data set; Based on the measurement point data set, abnormal points are identified by detecting abnormal fluctuations in the data. Specifically, an abnormal detection algorithm based on statistics is used. The parameters are set to a standard deviation threshold of 3. The Z-Score method is used to calculate the abnormal score of each data point. The measurement point data set is subjected to abnormal detection calculation. Each data point is standardized by the Z-Score method, and its deviation from the mean is calculated. Data with a deviation greater than 3 times the standard deviation is marked as an abnormal point. Then, the pressure conduction between the measurement points is adjusted. The specific adjustment method is to adjust the pressure conduction coefficient of the adjacent measurement points according to the location of the abnormal point, confirm the adjustment effect by re-measurement, and generate the measurement point distribution optimization result. Based on the optimization results of the measurement point distribution, the optimized distribution is verified by checking the pressure changes of the measurement points one by one. Specifically, the pressure conduction optimization algorithm has parameters set to an adjustment range of the conduction coefficient of ±10%. The iterative optimization method is used for multiple iterative adjustments. The pressure changes of each measurement point are checked one by one, and the conduction coefficient is adjusted according to the actual measurement results. The accuracy of the distribution adjustment is confirmed through multiple iterative optimizations, and then the optimized measurement point data is recorded again. The final data records of all measurement points are integrated and analyzed to generate an optimization plan for the liquid level depth measurement points.

[0032] The specific steps to generate the full well liquid level depth monitoring results are: Based on the optimization scheme of the liquid level depth measurement points, the liquid level monitoring device is used to integrate the optimized measurement point data, and a comprehensive data set is generated by successively superimposing the measurement data to generate an optimized measurement point data set; Based on the optimized measurement point data set, cross-comparison and analysis are performed by comparing the data change trends between different measurement points, unifying the relevant data and generating measurement point cross-analysis results; Based on the cross analysis results of the measurement points, the data of all measurement points are integrated to form the final liquid level depth distribution, generate the full-well liquid level depth monitoring results, and obtain the full-well liquid level depth monitoring results; Based on the optimization scheme of the liquid level depth measurement points, the liquid level monitoring device is used to integrate the optimized measurement point data, and a comprehensive data set is generated by successively superimposing the measurement data. Specifically, the data superposition algorithm is used, and the parameters are set to a superposition step of 5 minutes. The superposition method is the weighted average method, and the weights are set to 0.5, 0.3 and 0.2 according to the importance of the measurement points. The optimized data of each measurement point are arranged in time series, and then superimposed in turn. During superposition, the weighted average of the measurement data at each time point is calculated according to the weight, and the result is added to the comprehensive data set to generate an optimized measurement point data set; Based on the optimized measurement point data set, by comparing the data change trends between different measurement points, the time series analysis algorithm is used, the parameters are set to a window size of 10, and the distance metric uses the Euclidean distance. The data of each measurement point is standardized, and the time series data of different measurement points are aligned and compared. The data change trend is analyzed by calculating the similarity between the measurement points, and a cross comparison is performed. The data with higher similarity is processed uniformly to generate the measurement point cross analysis results; Based on the cross-analysis results of the measuring points, the data of all measuring points are integrated. Specifically, the data fusion algorithm adopts the Bayesian reasoning method. The parameters are set as the prior probability according to the historical data and calculated according to the current measurement data. The data in the cross-analysis results of the measuring points are processed by Bayesian reasoning. The data is updated and integrated through the combination of the prior probability and the likelihood function. The data of all measuring points are fused into a unified liquid level depth distribution to generate the full-well liquid level depth monitoring results.

[0033] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A real-time energy and LSTM oil well dynamic liquid level depth measurement method, characterized in that: The following steps are involved: Step 1: Use energy monitoring devices to collect and classify real-time data, analyze and filter the data according to consumption thresholds, and mark them to generate energy status classification results; Step 2: Based on the energy state classification result, a long short-term memory artificial neural network is used to obtain real-time measurement data of the liquid level depth of the oil well, and the data is accumulated and calculated and segmented, and the liquid level change area is screened, and a difference analysis is performed at the same time to determine the key area and generate the liquid level depth space feature; Step 3: Based on the spatial characteristics of the liquid level depth, interactive analysis is performed through the liquid level depth data and temperature fluctuations, detailed calculations are performed on the temperature changes under the oil well operation mode, and the main change trends of the liquid level depth are determined. Combined with the spatial and temporal data characteristics, a spatiotemporal correlation model of the liquid level depth is generated; Step 4: Based on the spatiotemporal correlation model of the liquid level depth, a spatiotemporal graph convolutional network is used to match real-time data between the energy monitoring device and the liquid level depth sensor, determine the liquid level change under the current oil well operation mode, calculate the time series data, compare the liquid level change trends under different states, and generate a real-time liquid level depth dynamic model; Step 5: Based on the real-time liquid level dynamic model, a liquid level sensor is used to record the data of the measuring points in detail, and abnormal points are identified according to the recorded data, and the pressure conduction between the measuring points is adjusted, and the distribution of the measuring points is optimized through the pressure change to generate an optimization plan for the liquid level measuring points; Step 6: Based on the optimization scheme of the liquid level depth measurement points, the liquid level monitoring device is used to integrate and process the optimized measurement point data, and the integrated data is used for verification, and the data of each measurement point is cross-compared and analyzed to generate the full-well liquid level depth monitoring results.

2. The method for measuring the dynamic liquid level of an oil well using real-time energy and LSTM according to claim 1, characterized in that: The spatial characteristics of the liquid level depth include key areas of liquid level depth change, liquid level fluctuation amplitude in the area, and associated spatial positions. The real-time liquid level depth dynamic model includes current energy state, liquid level depth change curve, and predicted trend. The liquid level depth measurement point optimization plan includes layout adjustment of measurement points, optimization of pressure conduction path, and precision of liquid level depth monitoring. The full-well liquid level depth monitoring results include comprehensive data on liquid level depth, spatiotemporal distribution diagram of each measurement point, and liquid level change prediction results.

3. The method for measuring the dynamic liquid level depth of an oil well using real-time energy and LSTM according to claim 1, characterized in that: The specific steps of generating the energy state classification result are: Use energy monitoring devices to collect and classify real-time data, screen the collected energy consumption data, filter according to the set energy threshold, mark the classification of different energy intervals, and annotate the data labels to generate energy interval classification data; Based on the energy interval classification data, the marked energy data is analyzed, the data of different intervals are classified by segmentation operation, and comprehensive screening is performed to obtain marked energy data; Based on the marked energy data, the data under various oil well operation modes are classified and organized, and merged into a classification index table to generate an energy state classification result.

4. The method for measuring the dynamic liquid level depth of an oil well using real-time energy and LSTM according to claim 1, characterized in that: The specific steps of generating the liquid level depth spatial feature are: Based on the energy state classification result, liquid level depth data is obtained, and a data processor is used to perform cumulative calculation, the liquid level depth data is grouped by time period, and the liquid level depth data is classified and integrated according to the time period to generate segmented liquid level data; Based on the segmented liquid level data, a long short-term memory artificial neural network is used to perform regional difference analysis on the data of different time periods, determine the significant area of ​​liquid level depth change by detecting the change amplitude, aggregate the data with obvious change amplitude, and generate the liquid level change area; Based on the liquid level change area, the key liquid level depth change position in each area is calculated, the spatial difference is compared, the information of the key area is integrated, the corresponding liquid level depth map is established, and the liquid level depth spatial feature is generated.

5. The method for measuring the dynamic liquid level depth of an oil well using real-time energy and LSTM according to claim 1, characterized in that: The long short-term memory artificial neural network is according to the formula: Among them: Among them, For the moment The implicit state of is the input segmented liquid level data, is the volatility of the liquid level data, is the environmental interference parameter, is the weight matrix, is an implicit state, is the liquid level data currently input, is the volatility, is the environmental interference parameter, is the adaptive weight coefficient, is the bias vector, is the timing dependent deviation, is the activation function.

6. The method for measuring the dynamic liquid level depth of an oil well using real-time energy and LSTM according to claim 1, characterized in that: The specific steps of generating the liquid level depth spatiotemporal correlation model are as follows: Based on the liquid level depth spatial characteristics, the liquid level depth data is extracted and the temperature fluctuation information is collected, and interactive processing is performed at the same time to generate a temperature and liquid level correlation data set; Based on the temperature and liquid level correlation data set, temperature change calculation is performed, liquid level depth change trend is analyzed, and liquid level depth trend analysis results are generated; Based on the liquid level depth trend analysis results, combined with time series data and spatial characteristics, a liquid level depth spatiotemporal correlation model is established to generate a liquid level depth spatiotemporal correlation model.

7. The method for measuring the dynamic liquid level depth of an oil well using real-time energy and LSTM according to claim 1, characterized in that: The specific steps of generating the real-time liquid level depth dynamic model are: Based on the spatiotemporal correlation model of liquid level depth, an energy monitoring device and a liquid level depth sensor are used to perform real-time data matching and liquid level depth data collection through a spatiotemporal graph convolutional network to generate the current liquid level depth state; Based on the current liquid level depth state, calculate the liquid level depth change trend, compare the change situation under the oil well operation mode, and generate liquid level change trend data; Based on the liquid level change trend data, a real-time liquid level depth dynamic model is updated and generated to obtain a real-time liquid level depth dynamic model.

8. The method for measuring the dynamic liquid level depth of an oil well using real-time energy and LSTM according to claim 1, characterized in that: The spatiotemporal graph convolutional network is based on the formula: in: The updated liquid level state matrix, is a nonlinear activation function, is the dynamic adjustment coefficient, is the adjacency matrix Power, is the initial liquid level depth data matrix, For the The convolution kernel weight matrix of order, is the time attenuation coefficient, For the time difference, is the time matrix, is the environmental pressure coefficient, is the pressure matrix.

9. The method for measuring the dynamic liquid level depth of an oil well using real-time energy and LSTM according to claim 1, characterized in that: The specific steps of generating the optimization scheme of the liquid level depth measurement point are as follows: Based on the real-time liquid level depth dynamic model, a liquid level depth sensor is used to record data of the measuring point, and the data is screened by recording the data multiple times, and the effective data is integrated to generate a measuring point data set; Based on the measurement point data set, abnormal points are identified by detecting abnormal fluctuations in the data, pressure conduction between the measurement points is adjusted, the adjustment effect is confirmed by re-measurement, and a measurement point distribution optimization result is generated; Based on the measurement point distribution optimization result, the optimized distribution is verified by checking the pressure changes of the measurement points one by one, and the measurement point data is recorded again to generate an optimization plan for the liquid level depth measurement points.

10. The method for measuring the dynamic liquid level depth of an oil well using real-time energy and LSTM according to claim 1, characterized in that: The specific steps of generating the full well liquid level depth monitoring result are: Based on the optimization scheme for the liquid level depth measurement points, the optimized measurement point data are integrated by using a liquid level monitoring device, and a comprehensive data set is generated by successively superimposing the measurement data to generate an optimized measurement point data set; Based on the optimized measurement point data set, by comparing the data change trends between different measurement points, cross-comparison and analysis are performed to unify the relevant data and generate measurement point cross-analysis results; Based on the cross analysis results of the measuring points, the data of all measuring points are integrated to form the final liquid level depth distribution, generate the full-well liquid level depth monitoring result, and obtain the full-well liquid level depth monitoring result.

Images