Oil stock early warning method based on real-time perception

Abstract

The invention provides an oil inventory early warning method based on real-time perception, and the method comprises the steps: obtaining the suction pressure data of an oil and gas recovery device in real time, monitoring the real-time change trend of the suction pressure, and determining that the suction pressure has abnormal fluctuation if the pressure change exceeds a normal fluctuation range; if the air suction pressure abnormal fluctuation degree exceeds a preset fluctuation threshold value, the real-time pressure value of each storage tank is obtained from a pressure sensor of the multiple storage tanks, the difference value between the current pressure and the target pressure is calculated, and the distribution characteristics of pressure fluctuation are determined; oil flow data of each storage tank is obtained through an oil metering device, the change before and after the valve opening is adjusted is combined, the increase and decrease of loss and the change of loss rate after the valve opening is adjusted are analyzed, and the mapping relation between the valve opening and loss is established.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to an oil inventory warning method based on real-time perception. Background Art

[0002] During the operation of an oil depot, when multiple storage tanks are operating simultaneously, the pressure fluctuation characteristics of the oil-gas balance system become a key issue. When the suction pressure of the oil-gas recovery device fluctuates, this fluctuation will directly affect the gas-phase pressure balance inside each storage tank. Due to differences in oil types, liquid levels, and operating states among different storage tanks, the response speed and sensitivity to pressure are different. When propagating pressure fluctuations within the system, this difference will cause pressure imbalance among the storage tanks, thereby affecting the oil-gas recovery efficiency. The edge computing unit undertakes the responsibilities of real-time monitoring and coordinated control during this process. It needs to dynamically adjust the opening degree of the pressure release valve according to the pressure data of each storage tank to ensure the stability of the overall system pressure. However, this adjustment is not a simple linear relationship. When the opening degree of the valve of a certain storage tank increases to release excess pressure, the pressure changes of other storage tanks may be further disturbed, resulting in increased pressure fluctuations within the system. This chain reaction may trap the control strategy of the edge computing unit in a dilemma of repeated adjustments. In addition, the change in the opening degree of the pressure release valve directly affects the evaporation loss of the oil product. When the valve opening is too large, the amount of oil-gas emissions increases, resulting in increased oil product loss, thus affecting inventory management; while when the valve opening is too small, the pressure inside the storage tank may exceed the safety threshold, increasing the risk. Therefore, when the edge computing unit coordinates the valve opening, it needs to find an optimal solution between pressure balance and oil product loss. This complex dynamic balance relationship places higher requirements on the real-time performance and accuracy of the system. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides an oil inventory warning method based on real-time perception, mainly including:

[0004] Obtain the suction pressure data of the oil-gas recovery device in real time, monitor the real-time change trend of the suction pressure. If the pressure change exceeds the normal fluctuation range, it is determined that there is an abnormal fluctuation in the suction pressure;

[0005] If the abnormal fluctuation degree of the suction pressure exceeds the preset fluctuation threshold, obtain the real-time pressure values of each storage tank from the pressure sensors of multiple storage tanks, calculate the difference between the current pressure and the target pressure, and determine the distribution characteristics of the pressure fluctuation;

[0006] According to the distribution characteristics of the pressure fluctuation, adopt a collaborative control algorithm based on pressure balance to determine the opening adjustment value of the pressure release valve for each storage tank, and send the valve opening adjustment value to the pressure release valve control system of each storage tank to adjust the valve opening in real time;

[0007] Monitor the pressure values of each storage tank after adjustment, and determine whether the preset pressure balance state standard is reached. If not, obtain the corresponding temperature gradient distribution according to the magnitude and change trend of the pressure difference, calculate the gas expansion coefficient differences in each storage tank, and combine the pressure linkage relationship between the storage tanks to re-determine the opening adjustment values of the pressure relief valves of each storage tank;

[0008] Obtain the oil product flow data of each storage tank through the oil product metering device, analyze the increase or decrease of the loss amount and the change of the loss rate after the valve opening adjustment in combination with the changes before and after the valve opening adjustment, and establish the mapping relationship between the valve opening and the loss;

[0009] Based on the mapping relationship between the valve opening and the loss, adjust the valve opening, including adjusting the step size, adjustment frequency and adjustment timing of the valve opening change, monitor the oil product loss data after adjustment in real time, judge the effectiveness of the optimization strategy according to the loss amount and loss rate indicators, and dynamically adjust the opening adjustment value according to the monitoring results;

[0010] Combine with the inventory warning system, integrate the pressure fluctuation, valve opening and oil product loss data into the inventory management platform, cooperate to control the operation status of multiple storage tanks and monitor the whole process. If the loss exceeds the preset loss threshold, an early warning signal will be sent through the inventory management platform.

[0011] Furthermore, obtain the suction pressure data of the oil and gas recovery device in real time, monitor the real-time change trend of the suction pressure. If the pressure change exceeds the normal fluctuation range, it is determined that there is an abnormal fluctuation in the suction pressure, including: according to the pressure value records collected at the pressure sensor detection points, obtain the suction pressure data with a sampling accuracy within the range of kilopascals through the pressure sampling frequency, correct and compensate the pressure data with the pressure compensation coefficient obtained based on historical data statistics to obtain the real-time pressure value, store the data according to the pressure monitoring time period, calculate the pressure change rate by using the ratio of the pressure change amount per unit time to the pressure value at the previous moment, draw the pressure fluctuation trend curve through time series data processing, calculate the pressure difference sequence within two adjacent monitoring time periods, obtain the pressure fluctuation amplitude through the sliding window method, and judge the pressure fluctuation state from the preset pressure fluctuation range threshold; if the pressure fluctuation amplitude exceeds the upper limit of the preset threshold, determine the abnormal pressure level according to the deviation value between the real-time pressure value and the preset reference pressure, record the abnormal pressure data by using the data correlation method based on the time stamp, construct the pressure fluctuation model by using the polynomial regression method, extract the fluctuation characteristic parameters through the model coefficients and fitting errors, and establish the pressure fluctuation characteristic vector; classify and identify the abnormal pressure fluctuations by using the support vector machine algorithm according to the pressure fluctuation characteristic vector, and obtain the characteristic features of different types of abnormal pressure fluctuation modes through cluster analysis.

[0012] Further, if the abnormal fluctuation degree of the suction pressure exceeds the preset fluctuation threshold, obtain the real-time pressure values of each storage tank from the pressure sensors of multiple storage tanks, calculate the difference between the current pressure and the target pressure, and determine the distribution characteristics of the pressure fluctuation, including: obtain the pressure time-series data from the pressure sensors of multiple storage tanks according to the storage tank pressure sampling interval, calculate the abnormal degree of fluctuation through comparison with the preset pressure fluctuation threshold range, record the fluctuation trigger moment for the data with the abnormal degree exceeding the threshold, obtain the real-time pressure values of each storage tank, calculate the pressure deviation sequence using the preset target pressure reference value, and statistically obtain the mean value, maximum value, and minimum value of the fluctuation amplitude from the pressure deviation sequence; perform spectrum analysis on the pressure deviation sequence through fast Fourier transform, extract the main frequency component from the spectrum data to obtain the pressure fluctuation frequency characteristics, and divide the high-frequency fluctuation interval and the low-frequency fluctuation interval according to the fluctuation frequency characteristics; count the start and end moments of the pressure fluctuation for different fluctuation intervals, calculate the fluctuation duration through the time stamp difference, and perform data smoothing processing on the fluctuation duration sequence using the exponential smoothing method; construct a feature vector based on the mean value of the fluctuation amplitude, the fluctuation frequency characteristics, and the smoothed duration sequence, establish a pressure fluctuation feature model through the support vector regression algorithm; classify the pressure fluctuation feature model using the clustering algorithm, and extract the fluctuation distribution characteristics from the clustering results, including the statistical characteristics of the fluctuation amplitude distribution, frequency distribution, and duration distribution.

[0013] Further, according to the distribution characteristics of the pressure fluctuation, adopt a collaborative control algorithm based on pressure balance to determine the opening adjustment value of the pressure release valve of each storage tank, and send the valve opening adjustment value to the pressure release valve control system of each storage tank to adjust the valve opening in real time, including: normalize the pressure data of the storage tank group according to the distribution characteristics of the pressure fluctuation, divide the fluctuation interval through the preset pressure fluctuation threshold, and calculate the mean value of the pressure deviation for the fluctuation interval to obtain the pressure target value of the storage tank group; for the deviation amount between the real-time pressure of each storage tank and the pressure target value of the storage tank group, construct a pressure adjustment rule using the fuzzy neural network algorithm to obtain the pressure release rate and adjustment priority of each storage tank; calculate the initial adjustment amount of the valve opening of each storage tank according to the pressure release rate, determine the maximum opening limit value of the valve through the pressure difference ratio between the pressure in the storage tank and the pipeline back pressure, obtain the actual adjustment amount of the valve, arrange the change of the valve opening of each storage tank in time sequence according to the adjustment priority, and discretize the valve opening adjustment amount through the preset unit adjustment period; generate a valve control sequence according to the discretized opening adjustment amount, calculate the valve opening adjustment compensation value using the pressure gradient between adjacent storage tanks, obtain the valve opening adjustment instruction, send a control signal to the pressure release valve of each storage tank according to the adjustment period, and judge the adjustment execution state through the feedback of the valve opening execution value collected.

[0014] Furthermore, a collaborative control algorithm based on pressure equilibrium is adopted. According to the difference between the real-time pressure value and the target pressure value of each storage tank, combined with the amplitude, frequency, and duration of the pressure fluctuation, the opening adjustment percentage of the pressure release valve of each storage tank is calculated, and a valve opening adjustment value is generated, including: normalizing the deviation between the real-time pressure and the target pressure of each storage tank according to the storage tank pressure difference, quantifying the fluctuation intensity with the pressure fluctuation amplitude, calculating the number of fluctuations per unit time through the fluctuation frequency, and recording the fluctuation interval from the fluctuation duration; constructing a pressure adjustment matrix for the fluctuation intensity, calculating the adjustment urgency through the number of fluctuations per unit time, determining the adjustment action time using the fluctuation interval, obtaining the pressure adjustment priority order, dividing the adjustment levels, calculating the initial adjustment amount by the product of the pressure release rate and the adjustment level, and obtaining the valve opening change curve from the adjustment action time; performing piecewise linearization on the valve opening change curve, sampling the opening change through the valve response time, and calculating the opening value of the sampling point using the linear interpolation method; generating a valve adjustment sequence according to the sampling point opening value, discretizing the adjustment sequence using the quantization interval to obtain a valve opening adjustment step sequence; associating the opening adjustment step sequence with the time stamp, calculating the execution time of each adjustment step through the adjustment period, and generating a valve opening adjustment value from the execution time sequence.

[0015] Furthermore, monitor the pressure values of each storage tank after adjustment, and judge whether the preset pressure balance state standard is reached. If not, according to the magnitude and change trend of the pressure difference, obtain the corresponding temperature gradient distribution, calculate the difference in gas expansion coefficients in each storage tank, and combine the pressure linkage relationship between the storage tanks to re-determine the opening adjustment value of the pressure release valve of each storage tank, including: obtaining the pressure time series data from the storage tank pressure sensor according to the monitoring time period, calculating the pressure deviation value through comparison with the preset pressure balance threshold, calculating the pressure change rate using the sliding window method, and judging the balance state from the pressure deviation value and the change rate. For the storage tank group that has not reached the balance state, use the temperature sensor array to obtain the temperature field distribution data, calculate the temperature gradient field through spatial interpolation, and obtain the temperature change rate of each storage tank from the temperature change trend. Calculate the gas expansion coefficient of each storage tank according to the ideal gas state equation, establish a gas thermodynamic characteristic curve using the storage tank pressure value and the temperature change rate, and obtain the pressure-temperature compensation coefficient from the characteristic curve. Construct a pressure transfer function for the pressure deviation value between adjacent storage tanks, correct the pressure transfer gain through the pressure-temperature compensation coefficient, and obtain the storage tank pressure linkage coefficient from the corrected transfer function. Calculate the pressure adjustment weight according to the storage tank pressure linkage coefficient, classify the weight using the fuzzy control rule base, and determine the pressure adjustment priority of each storage tank from the classification result. For the corresponding relationship between the pressure adjustment priority and the gas expansion coefficient, use a proportional-integral controller to calculate the valve opening compensation amount, and generate a valve opening adjustment value from the compensation amount sequence.

[0016] Furthermore, obtain the oil flow data of each storage tank through the oil metering device. Combine the changes before and after the valve opening adjustment, analyze the increase or decrease in the loss amount and the change in the loss rate after the valve opening adjustment, and establish the mapping relationship between the valve opening and the loss, including: obtain the flow data sequence from the oil metering device according to the flow sampling frequency, use the valve response characteristic curve to record the time points before and after the opening adjustment, and obtain the real-time flow change data during the valve opening adjustment through time series data processing. Calculate the loss amount per unit time for the real-time flow change data, perform data calibration using the preset reference loss value, and statistically analyze the cumulative change in the loss amount through a time window of a fixed size. Construct a time series feature vector based on the cumulative change in the loss amount, calculate the loss change rate using the time-weighted average method, and obtain the loss trend characteristics of each time interval through piecewise linear fitting. Extract the valve opening change interval for the loss trend characteristics, calculate the discrete sampling points during the opening adjustment process using the linear interpolation method, and fit the opening adjustment curve using the least squares method. Establish the corresponding relationship based on the opening adjustment curve and the loss trend characteristics, construct the loss prediction function using the polynomial regression method, and determine the polynomial order through cross-validation. Calculate the loss increment corresponding to the opening adjustment amount for the loss prediction function, optimize the parameters of the prediction function using the gradient descent method, and obtain the mapping relationship between the valve opening and the loss from the optimized function.

[0017] Furthermore, based on the mapping relationship between the valve opening and the loss, adjust the valve opening, including adjusting the step size, adjustment frequency, and adjustment timing of the valve opening change, monitor the oil loss data after the adjustment in real time, judge the effectiveness of the optimization strategy according to the loss amount and loss rate indicators, and dynamically adjust the opening adjustment value according to the monitoring results, including: discretize the opening adjustment value of the valve according to the preset opening change step size, divide the time interval using a fixed adjustment period, screen the opening adjustment moment through the minimum adjustment interval, and determine the effective adjustment interval from the preset upper and lower limits of the opening adjustment. Extract the opening adjustment curve for the effective adjustment interval, record the loss amount per unit time using the loss data acquisition device, calculate the integral value of the loss amount through a continuous monitoring window, and judge the current loss state from the preset loss threshold. Construct the oil loss feature matrix based on the loss state data, calculate the loss change trend using the time-weighted method, determine the opening adjustment priority through the trend prediction function, evaluate the adjustment effect from the combined criterion of the loss amount and the loss rate, divide the loss control area, construct the opening adjustment compensation function using the piecewise linearization method, update the compensation coefficient through the adaptive optimization algorithm, and extract the adjustment correction amount from the compensated opening value sequence. Generate a new opening control instruction according to the adjustment correction amount, sort the instruction execution moment using the response time-weighted method, and obtain the real-time opening adjustment value through dynamic feedback correction. Calculate the valve response characteristic curve for the real-time opening adjustment value, predict the loss change after the opening adjustment using the polynomial fitting method, and dynamically update the opening adjustment sequence from the prediction result.

[0018] Furthermore, combined with the inventory warning system, the pressure fluctuation, valve opening and oil loss data are integrated into the inventory management platform, and the operation status of multiple tanks is collaboratively controlled and monitored throughout the process. If the loss exceeds the preset loss threshold, an early warning signal is issued through the inventory management platform, including: obtaining pressure data from the tank pressure sensor according to the data acquisition cycle, using the valve position feedback device to record the opening value, collecting oil loss data through the loss meter, and establishing a multi-parameter association sequence from the data time tag. For the multi-parameter association sequence, data segmentation is performed according to a fixed time window, abnormal data is eliminated using the time series verification rule, and the operation process feature vector is generated through data standardization processing. The tank operation state matrix is ​​constructed, and the inventory change relationship of adjacent tanks is calculated using the state correlation function. The operation stability of the tank group is judged by the inventory fluctuation rate. The loss influencing factor is calculated for the operation status of the tank group, and the linkage control command is generated using the multi-tank collaborative operation rule, and the operation parameters of each tank are adjusted through the control command actuator. According to the comparison between the real-time loss data and the preset loss threshold, the graded warning rule is used to generate the warning level, and the warning information is output through the warning signal generator. Key status parameters are extracted from the warning information, and emergency response instructions are generated using the inventory warning processing function, which are then sent to each tank control unit through the instruction distribution mechanism.

[0019] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0020] The present invention discloses an oil inventory early warning method based on real-time perception. The method monitors the suction pressure of multiple storage tanks in real time. When abnormal fluctuations are detected, the pressure distribution characteristics are quickly analyzed, and the opening of the pressure release valve of each storage tank is dynamically adjusted using a collaborative control algorithm based on pressure balance. Combined with the temperature gradient and the difference in gas expansion coefficient, the present invention can accurately control the pressure balance between multiple storage tanks. At the same time, by establishing a correlation model between valve opening and oil loss, the present invention can minimize oil loss while ensuring pressure balance. In addition, the present invention also combines pressure control with inventory management to achieve full-process monitoring and early warning of the operating status of multiple storage tanks. This intelligent pressure control method not only improves the operating efficiency of the oil and gas recovery device, but also effectively reduces oil loss, providing an innovative tank management solution for the petrochemical industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The present invention is a flowchart of an oil inventory early warning method based on real-time perception.

[0022] Figure 2 This is another flow chart of an oil inventory early warning method based on real-time perception according to the present invention. DETAILED DESCRIPTION

[0023] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] As Figure 1-2 , a method for warning of oil product inventory based on real-time perception in this embodiment may specifically include:

[0025] S101 If the vapor recovery device is operating, the edge computing unit collects the suction pressure data in real time and monitors its change trend. When the pressure change exceeds the preset normal range, it is determined that there is an abnormal fluctuation, and the fluctuation characteristics are extracted to determine the abnormal level.

[0026] S1011 The edge computing unit collects the suction pressure data of the vapor recovery pipeline through a pressure sensor, sets the sampling frequency to 10 times per second, controls the accuracy within 0.1 kPa, corrects the pressure value using the pressure compensation coefficient based on historical data, and obtains the real-time pressure data. For the real-time pressure data, it is stored in a 15-minute cycle, the ratio of the pressure change amount per unit time to the value at the previous moment is calculated to generate the pressure change rate, and the pressure fluctuation trend curve is drawn through time series processing. The difference between the maximum and minimum pressures within the window is calculated using a 60-second sliding window to obtain the fluctuation amplitude. If the amplitude exceeds the preset range of plus or minus 2 kPa, the abnormal data is recorded.

[0027] S1012 According to the deviation between the real-time pressure value and the preset reference pressure of minus 3 kPa, the abnormal level is determined. If the deviation is greater than 1 kPa, it is a mild abnormality; if it is greater than 2 kPa, it is a moderate abnormality; if it is greater than 3 kPa, it is a severe abnormality. For the abnormal data, a 3rd-order polynomial regression is used to construct a fluctuation model, and the model coefficients and root mean square error are extracted as characteristic parameters. The model is confirmed to be effective when the error is less than 0.5 kPa. Based on the characteristic parameters, a fluctuation feature vector is constructed, and the support vector machine algorithm is used to classify the abnormal types, and three mode feature libraries of pipeline blockage, leakage, and equipment failure are established in combination with cluster analysis.

[0028] In the embodiment of the present application, S101 realizes the real-time perception of the suction pressure through high-frequency sampling and time series analysis to ensure the timely identification of abnormal fluctuations. In an alternative embodiment, the sampling frequency can be adjusted according to the actual scenario, and the feature extraction algorithm can also be replaced with a neural network model to improve the classification accuracy.

[0029] S102 If it is detected that the abnormal fluctuation of the suction pressure exceeds the preset threshold, the real-time pressure data of each storage tank is collected from the pressure sensors of multiple storage tanks, the deviation value from the target pressure is calculated and a pressure deviation sequence is generated. Based on this sequence, the fluctuation distribution characteristics are analyzed, including the fluctuation amplitude, frequency, and duration, and a pressure fluctuation model is constructed through the feature vector and the distribution characteristics are extracted.

[0030] S1021 Use multi-tank pressure sensors to obtain pressure time-series data at a sampling frequency of 10 times per second. Combine with a preset fluctuation threshold of plus or minus 2 kPa to compare with the real-time pressure value, determine the degree of abnormal fluctuation and record the trigger moment. For example, when the pressure of the 3rd tank rises from -3.5 kPa to -1.2 kPa, the trigger time is 10:15:30 am. At the same time, collect the real-time pressures of other tanks, such as -2.8 kPa for the 1st tank, -2.5 kPa for the 2nd tank, and -2.9 kPa for the 4th tank. Calculate the pressure deviation sequences of each tank using the target pressure reference value of -3 kPa. Among them, the deviation of the 3rd tank reaches 1.8 kPa, and the deviations of other tanks are between 0.2 and 0.5 kPa. Statistically analyze the sequences to obtain the mean fluctuation amplitude of 1.2 kPa, the maximum value of 1.8 kPa, and the minimum value of 0.3 kPa.

[0031] S1022 Apply the fast Fourier transform to the pressure deviation sequence for spectral analysis. Extract the main frequency components from the spectral data, such as 0.5 Hz and 2 Hz. Divide the low-frequency fluctuation interval and the high-frequency fluctuation interval respectively to reflect slow changes and fast perturbations. Statistically analyze the start and end times of the fluctuations. For example, from 10:15 to 10:25 for a total of 600 seconds. Use the exponential smoothing method with a smoothing coefficient of 0.3 to process the duration sequence. After eliminating random noise, construct a 12-dimensional feature vector in combination with the mean fluctuation amplitude and frequency characteristics, including amplitude standard deviation, peak factor, frequency bandwidth, and duration variance, etc. Generate a pressure fluctuation model through the support vector regression algorithm, and classify it into the "rapid pressure rise with high-frequency perturbation" mode through cluster analysis.

[0032] In the embodiment of the present application, the pressure fluctuation characteristics are accurately captured through high-frequency sampling and spectral analysis. In an alternative embodiment, the sampling interval can be adjusted according to the scenario, and the clustering algorithm can be replaced by other methods to improve the feature extraction efficiency.

[0033] S103 If the distribution characteristics of the pressure fluctuation are obtained, adopt a cooperative control algorithm based on pressure balance. Calculate the adjustment value of the pressure release valve opening according to the deviation between the real-time pressure of each tank and the target pressure and the fluctuation characteristics, and send the adjustment value to the valve control system of each tank to achieve real-time opening adjustment.

[0034] S1031 Normalize the pressure data of the storage tank group, map the real-time pressure values such as -2.8 kPa, -3.2 kPa, -2.5 kPa, -3.5 kPa, -2.9 kPa, -3.1 kPa to the interval from 0 to 1, divide the high and low pressure fluctuation intervals through the preset fluctuation threshold of ±0.5 kPa, calculate the average pressure deviation to obtain the target pressure of -3.0 kPa, establish adjustment rules using the fuzzy neural network algorithm based on the deviation amount of each storage tank from the target value, output the pressure release rate from 0.2 kPa per minute to 1.0 kPa per minute and the five-level adjustment priority, calculate the initial adjustment amount such as 40% by combining the release rate and the pressure difference ratio of 0.625 and limit the maximum opening to 60%, obtain the actual adjustment amount of 35%, sort by priority and discretize it into a 7-step sequence with a 1-minute cycle and a 5% step size.

[0035] S1032 Construct a pressure adjustment matrix according to the fluctuation amplitude of 0.8 kPa, the frequency of 4 times per minute and the duration of 600 seconds, calculate the adjustment urgency coefficient of 0.9 and determine the priority order, obtain the initial adjustment amount of 50% by multiplying the release rate and the adjustment level, generate an opening change curve and sample it in segments, quickly adjust 40% in the first 300 seconds and finely adjust 10% in the last 300 seconds, calculate the sampled point values through linear interpolation, discretize it into a 12-step sequence with a 5% quantization interval, associate the time stamps to generate an adjustment instruction set to ensure that the No. 3 storage tank is adjusted first and gradually approaches the target pressure.

[0036] S1033 Send the valve opening adjustment instruction to the pressure release valve control system of each storage tank. Taking the No. 1 storage tank as an example, the opening gradually increases from 15% in 7% steps to the target value, the feedback execution value of 21.8% has a deviation from the instruction of less than 1%, confirm that the adjustment is normal, compensate 2% by combining the pressure gradient of 0.3 kPa between adjacent storage tanks, and achieve the dynamic balance of the pressure of the storage tank group.

[0037] In the embodiment of the present application, the valve opening adjustment is optimized through the cooperative control algorithm. In the optional embodiment, the sampling frequency or fuzzy rules can be adjusted to adapt to different scenarios, ensuring pressure balance while improving the execution efficiency.

[0038] S104 Monitor the adjusted pressure values of each storage tank and compare them with the preset balance standard. If the balance state is not reached, obtain the temperature gradient distribution and the difference in gas expansion coefficients according to the pressure difference and the change trend, and recalculate the opening adjustment value of the pressure release valve in combination with the pressure linkage characteristics between the storage tanks.

[0039] S1041 Collect the time-series data of the storage tank pressure at a frequency of once per second through a pressure sensor. For example, the real-time pressures of 5 storage tanks are -2.8 kPa, -3.2 kPa, -2.5 kPa, -3.4 kPa, and -2.7 kPa. Calculate the deviation from the target pressure of -3.0 kPa using the preset balance threshold of plus or minus 0.3 kPa. Analyze the pressure change rate using a 60-second sliding window. After determining that the storage tank group is not in a balanced state, call the temperature sensor array to collect temperature field data, such as 28 degrees Celsius at the top and 22 degrees Celsius at the bottom of storage tank No. 2. Generate a temperature gradient field through Kriging spatial interpolation and calculate the temperature change rate of 1.5 degrees Celsius per hour.

[0040] S1042 Calculate the gas expansion coefficient based on the ideal gas state equation, which is 1.0 at -3.0 kPa and 25 degrees Celsius, and increases to 1.15 after the temperature of storage tank No. 2 rises. Combine the pressure data and the temperature change rate to construct a thermodynamic characteristic curve, and obtain that the pressure compensation coefficient increases by 0.02 for every 1-degree Celsius increase in temperature. Construct a pressure transfer function for adjacent storage tanks, such as No. 1 and No. 2. The initial gain of 0.8 is corrected to 0.92 by the compensation coefficient. Calculate the coupling coefficient to reflect the coupling. Use fuzzy control rules to divide the weights into 5 levels according to the deviation and temperature influence. The weight of storage tank No. 2 is the highest at 0.9, and those of No. 1 and No. 3 are 0.6, and the rest are from 0.3 to 0.4.

[0041] S1043 Determine the adjustment priority according to the weight classification. Use a proportional-integral controller to calculate the valve opening compensation amount with a proportional coefficient of 0.8 and an integral time constant of 30 seconds. The initial compensation of storage tank No. 2 is 25% and is corrected to 28.75%. Generate a sequence containing 6 incremental adjustment values of 4% to 6% and send it to the valve control system to achieve stable pressure regulation.

[0042] S1044 Optimize the opening adjustment through the collaborative analysis of pressure and temperature. In the optional embodiment, the window length or interpolation method can be adjusted to improve the accuracy and ensure the dynamic balance of the pressure of the storage tank group.

[0043] S105 Collect the oil flow data of each storage tank through an oil metering device. Combine the change characteristics before and after the valve opening adjustment, analyze the increasing and decreasing trends and rate changes of the loss amount after adjustment, and establish the relationship between the loss and the opening to optimize the valve control strategy.

[0044] S1051 Obtain the flow data sequence from the oil metering device at a sampling frequency of once per second. For example, it takes 300 seconds for the valve opening to be adjusted from 15% to 35%, and the flow rate drops from 180 cubic meters per hour to 160 cubic meters per hour. Use the valve response characteristic curve to record the adjustment time point, process the time-series data to generate the real-time flow change trend, calculate the loss amount per unit time using a 30-second time window, and compare the cumulative change with the reference loss value of 2 cubic meters per hour. The initial loss increases to 3.5 cubic meters per hour and stabilizes at 2.8 cubic meters per hour later.

[0045] S1052 Construct a time-series feature vector based on cumulative loss data, including 10 sampling periods with weights decreasing from 1.0 to 0.1. Use the time-weighted average method to calculate the loss change rate, which is 0.15 cubic meters per hour in the first 10 minutes and then decreases to 0.05 cubic meters per hour. Divide the rapid rise section and the slow adjustment section by piecewise linear fitting. Extract 30 discrete sampling points within the opening change range, where the opening increases from 15% to 35%. Use linear interpolation to calculate the adjustment curve, which rapidly rises to 28% in the first 120 seconds and then slowly approaches the target value.

[0046] S1053 The error between the opening adjustment curve and the actual curve fitted by the least squares method is less than 1%. Combine the loss trend characteristics and use a third-order polynomial regression to construct a prediction function. Divide 500 groups of data into training and test sets according to 8:2, and optimize the parameters through cross-validation. It shows that when the opening is between 20% and 30%, the loss increment increases by about 0.1 cubic meters per hour for every 1% increase. Use the gradient descent method to iterate 500 times with a learning rate of 0.01 to optimize the function, and obtain the mapping relationship that the inflection point of loss is at 25% opening and the subsequent growth rate slows down.

[0047] In the embodiment of the present application, loss prediction is realized through the correlation analysis of flow rate and opening. In alternative embodiments, the time window or the polynomial order can be adjusted to improve the accuracy.

[0048] S106 Optimize the valve opening adjustment strategy based on the relationship between loss and opening. Adjust the step size, frequency, and timing and monitor the oil product loss data in real time. Use the loss amount and rate to evaluate the effect, and dynamically correct the opening adjustment value to improve the control accuracy.

[0049] S1061 Discretize the valve opening adjustment value with a 5% step size, set a 300-second adjustment period and a 60-second minimum interval to divide it into 6 adjustment units. The effective range is from 15% to 45%. Through the loss data acquisition device, sample and record the loss change when the opening is adjusted from 25% to 35% every 10 seconds. Calculate the integral value of the loss amount in a 120-second monitoring window. When it exceeds the threshold of 3 cubic meters per hour, it is determined as a high-loss state.

[0050] S1062 Construct a feature matrix based on the loss state, including three-dimensional data of loss amount, rate, and duration. Analyze the trend of 4-hour data with exponentially decaying weights. The loss rises rapidly when the opening is between 35% and 40%, and becomes gentle when it is between 30% and 35%. Based on this, divide the high, medium, and low adjustment priorities, and use piecewise linearization to generate a compensation function. The coefficient is 0.8 when the rate exceeds 0.5 cubic meters per hour, otherwise it is 1.2. The adaptive algorithm updates the correction amount to plus or minus 2% every 60 seconds.

[0051] S1063 generates control instructions based on the correction amount and sorts them by weighted response time. Fast instructions less than 5 seconds have a weight of 0.8, slow instructions greater than 5 seconds have a weight of 0.5, and feedback correction deviations are 1% and 2% respectively. A third-order polynomial is used to fit the response characteristic curve, with an error less than 0.5%. It is predicted that when the opening increases by 2%, the loss will increase by 0.2 cubic meters per hour. The large adjustment will be divided into 8 small steps, with dynamic changes at intervals of 60 to 120 seconds.

[0052] S1064 optimizes the opening adjustment through dynamic feedback and prediction. In an optional embodiment, the sampling frequency or weight coefficient can be adjusted to adapt to the scenario to ensure smooth loss control.

[0053] S107 integrates pressure fluctuation, valve opening and oil loss data into the inventory management platform through the inventory early warning system, realizing coordinated control and full-process monitoring of the operating status of multiple storage tanks. When the loss exceeds the preset threshold, it sends out an early warning signal to optimize management efficiency.

[0054] S1071 collects pressure data such as negative 3.2 kPa from pressure sensors of 8 storage tanks at a cycle of 1 second, the valve position feedback device records the opening value such as 35%, and the loss meter obtains the loss volume such as 2.5 cubic meters per hour. It uses time tags to generate multi-parameter correlation sequences, adopts 5-minute window segmentation and uses verification rules to eliminate abnormal data with pressure changes exceeding 1 kPa, opening changes exceeding 10%, and loss changes exceeding 1 cubic meter per hour. After standardization, a feature vector containing volatility, adjustment rate and change rate is formed.

[0055] S1072 constructs an operation state matrix based on the eigenvector and calculates the inventory change relationship between adjacent tanks. For example, the fluctuation rate of tank No. 3 is 2% per hour, which is 0.5% higher than that of other tanks. The operation abnormality is determined by the state association function, and the linkage control instruction is generated by combining the pressure, opening and loss influence factor 0.8. The opening of tank No. 3 is reduced by 5% and the adjacent tanks are compensated. The instruction is adjusted by the actuator to adjust the operating parameters.

[0056] S1073 compares the real-time loss with the threshold of 3 cubic meters, 2.5 cubic meters and 2 cubic meters per hour to divide the warning into level one to level three. The signal generator outputs information including level, time and parameters. According to the rules of level one closing the valve, level two reducing the opening and level three encrypting monitoring, emergency instructions are generated and sent to the control unit through the distribution mechanism to realize coordinated regulation.

[0057] S1074 optimizes storage tank management through multi-parameter coordination and graded warning. In an optional embodiment, the collection cycle or threshold can be adjusted to suit the scenario.

[0058] The above embodiments are only one of the preferred embodiments of the present invention and should not be used to limit the protection scope of the present invention. Any meaningless changes or polish made on the main design concept and spirit of the present invention, as long as the technical problems solved are still consistent with those of the present invention, should be included within the protection scope of the present invention.

Claims

1. A method for early warning of oil inventory based on real-time perception, characterized in that: The method comprises: Acquire the suction pressure data of the oil and gas recovery device in real time, monitor the real-time change trend of the suction pressure, and if the pressure change exceeds the normal fluctuation range, it is determined that the suction pressure has abnormal fluctuations; If the abnormal fluctuation of the suction pressure exceeds the preset fluctuation threshold, the real-time pressure value of each tank is obtained from the pressure sensors of the multiple tanks, the difference between the current pressure and the target pressure is calculated, and the distribution characteristics of the pressure fluctuation are determined; According to the distribution characteristics of pressure fluctuations, a collaborative control algorithm based on pressure balance is used to determine the opening adjustment value of the pressure release valve of each storage tank, and the valve opening adjustment value is sent to the pressure release valve control system of each storage tank to adjust the valve opening in real time; Monitor the adjusted pressure values ​​of each storage tank to determine whether the preset pressure balance state standard is reached. If not, obtain the corresponding temperature gradient distribution according to the size and change trend of the pressure difference, calculate the difference in gas expansion coefficients in each storage tank, and re-determine the opening adjustment value of the pressure release valve of each storage tank in combination with the pressure linkage relationship between the storage tanks; The oil flow data of each storage tank is obtained through the oil metering device. Combined with the changes before and after the valve opening is adjusted, the increase and decrease of the loss amount and the change of the loss rate after the valve opening is adjusted are analyzed to establish the mapping relationship between the valve opening and the loss. Based on the mapping relationship between valve opening and loss, adjust the valve opening, including adjusting the step size, frequency and timing of valve opening change, monitor the oil loss data after adjustment in real time, judge the effectiveness of the optimization strategy based on the loss amount and loss rate indicators, and dynamically adjust the opening adjustment value based on the monitoring results; Combined with the inventory early warning system, the pressure fluctuation, valve opening and oil loss data are integrated into the inventory management platform, and the operating status of multiple storage tanks is coordinated and monitored throughout the process. If the loss exceeds the preset loss threshold, a warning signal will be issued through the inventory management platform.

2. The method according to claim 1, characterized in that The real-time acquisition of the suction pressure data of the oil and gas recovery device, monitoring the real-time change trend of the suction pressure, and determining that the suction pressure has abnormal fluctuations if the pressure change exceeds the normal fluctuation range, includes: The pressure value is collected and recorded according to the pressure sensor detection point, and the real-time pressure value is obtained by correction and compensation through the pressure compensation coefficient; The pressure change rate is calculated based on the ratio of the pressure change per unit time to the pressure value at the previous moment, and a pressure fluctuation trend curve is obtained by time series data processing; According to the pressure fluctuation trend curve, a sliding window method is used to obtain a pressure fluctuation amplitude. If the pressure fluctuation amplitude exceeds a preset threshold upper limit, the pressure abnormality level is determined according to a deviation between the real-time pressure value and a preset reference pressure. A polynomial regression method is used to construct a pressure fluctuation model for the pressure anomaly level, and fluctuation characteristic parameters are extracted through the model coefficients and fitting errors of the pressure fluctuation model. A support vector machine algorithm is used according to the fluctuation characteristic parameters to obtain the abnormal pressure fluctuation classification result.

3. The method according to claim 1, characterized in that If the abnormal fluctuation of the suction pressure exceeds the preset fluctuation threshold, the real-time pressure value of each tank is obtained from the pressure sensors of the multiple tanks, the difference between the current pressure and the target pressure is calculated, and the distribution characteristics of the pressure fluctuation are determined, including: Comparing and calculating the tank pressure sampling data according to a preset pressure fluctuation threshold range, obtaining a fluctuation abnormality degree value and a tank pressure value at the fluctuation triggering moment; Using a preset target pressure reference value to perform deviation calculation on the tank pressure value to obtain a pressure deviation sequence; According to the pressure deviation sequence, the mean, maximum and minimum values ​​of the fluctuation amplitude are calculated; The pressure deviation sequence is analyzed by fast Fourier transform to extract the main frequency component and divide the high-frequency fluctuation range into low-frequency fluctuation range. A feature vector is constructed according to the fluctuation amplitude mean, fluctuation frequency characteristics and fluctuation duration sequence, a pressure fluctuation feature model is established through a support vector regression algorithm, and a cluster analysis is performed on the pressure fluctuation feature model to obtain fluctuation distribution characteristics.

4. The method according to claim 1, characterized in that According to the distribution characteristics of the pressure fluctuation, a collaborative control algorithm based on pressure balance is adopted to determine the opening adjustment value of the pressure release valve of each storage tank, send the valve opening adjustment value to the pressure release valve control system of each storage tank, and adjust the valve opening in real time, including: The pressure data of the storage tank group is normalized to obtain the pressure fluctuation distribution characteristics, and the fluctuation interval is divided according to the pressure fluctuation distribution characteristics by a preset pressure fluctuation threshold, and the pressure deviation mean is calculated to obtain the target pressure value of the storage tank group; According to the deviation between the target pressure value of the tank group and the real-time pressure of each tank, a fuzzy neural network algorithm is used to construct a pressure regulation rule to obtain the pressure release rate and regulation priority of each tank; The pressure release rate is used to calculate the initial adjustment amount of the valve opening of each storage tank, and the maximum opening limit of the valve is determined according to the pressure difference ratio between the pressure in the storage tank and the back pressure of the pipeline network to obtain the actual adjustment amount of the valve; The actual valve adjustment amount is time-sequenced according to the adjustment priority, and the valve opening adjustment amount is discretized through a preset unit adjustment cycle to obtain a valve control sequence.

5. The method according to claim 4, characterized in that Also includes: A collaborative control algorithm based on pressure balance is used to calculate the opening adjustment percentage of the pressure release valve of each tank according to the difference between the real-time pressure value of each tank and the target pressure value, combined with the fluctuation amplitude, frequency and duration of the pressure fluctuation, and generate the valve opening adjustment value, which includes: The deviation between the real-time pressure of the storage tank and the target pressure is normalized, and the pressure fluctuation amplitude is obtained from the normalization to quantify the fluctuation intensity; a pressure regulation matrix is ​​constructed according to the fluctuation intensity, and the number of fluctuations per unit time is calculated through the pressure regulation matrix to obtain the regulation urgency, and the pressure regulation priority is obtained from the regulation urgency; the regulation level is divided according to the pressure regulation priority, and the initial regulation amount is calculated by multiplying the regulation level with the pressure release rate, and the valve opening change curve is obtained from the initial regulation amount; the valve opening change curve is sampled, and the sampling point opening value is calculated by a linear interpolation method, and a valve adjustment sequence is generated from the sampling point opening value, and the valve opening adjustment step sequence is obtained by discretizing the adjustment sequence through a quantization interval.

6. The method according to claim 1, characterized in that The pressure values ​​of each storage tank after monitoring and adjustment are judged whether they reach the preset pressure balance state standard. If not, the corresponding temperature gradient distribution is obtained according to the size and change trend of the pressure difference, the difference in gas expansion coefficient in each storage tank is calculated, and the opening adjustment value of the pressure release valve of each storage tank is re-determined in combination with the pressure linkage relationship between the storage tanks, including: For the pressure time series data obtained by the tank pressure sensor, the pressure deviation value is calculated according to the preset pressure balance threshold, the pressure change rate is obtained by using the sliding window method, and the equilibrium state is determined from the pressure deviation value and the pressure change rate; If the storage tank group has not reached a state of equilibrium, a temperature sensor array is used to obtain temperature field distribution data, and the temperature gradient field is calculated by a spatial interpolation method, and the temperature change rate of the storage tank is obtained from the temperature gradient field; Establishing a gas thermodynamic characteristic curve according to the tank temperature change rate and the pressure time series data, and obtaining a pressure-temperature compensation coefficient from the gas thermodynamic characteristic curve; A pressure transfer function is constructed for the pressure deviation values ​​of adjacent storage tanks, the pressure transfer gain is corrected using the pressure temperature compensation coefficient, a storage tank pressure linkage coefficient is obtained from the pressure transfer function, and the pressure regulation priority is determined according to the storage tank pressure linkage coefficient.

7. The method according to claim 1, characterized in that The oil flow data of each storage tank is obtained through the oil metering device, and the increase or decrease of the loss amount and the change of the loss rate after the valve opening is adjusted are analyzed in combination with the changes before and after the valve opening is adjusted, and a mapping relationship between the valve opening and the loss is established, including: Obtaining the opening adjustment time point recorded in the valve response characteristic curve, and collecting real-time flow change data output by the flow metering device at the time point; Calculate the loss per unit time according to the real-time traffic change data, and use a preset time window to perform cumulative statistics on the loss per unit time to obtain loss accumulation data; The loss change rate is calculated using the time-weighted average method for the loss accumulation data to obtain the loss trend characteristics; The valve opening variation interval is extracted according to the loss trend characteristics, the opening adjustment curve is constructed by using linear interpolation and least square method, and the mapping relationship between valve opening and loss is obtained by polynomial regression.

8. The method according to claim 1, characterized in that: The valve opening is adjusted based on the mapping relationship between the valve opening and the loss, including adjusting the step size, frequency and timing of the valve opening change, real-time monitoring of the oil loss data after adjustment, judging the effectiveness of the optimization strategy according to the loss amount and loss rate indicators, and dynamically adjusting the opening adjustment value according to the monitoring results, including: Discretize the valve opening adjustment value according to the preset opening change step length, and obtain the opening adjustment time sequence through a fixed adjustment cycle; The opening adjustment time sequence is sampled, and the loss integral value is calculated from the sampled data; Constructing an oil product loss characteristic matrix according to the loss amount integral value, and obtaining a loss change trend curve by a time weighted method; The loss control area is divided according to the loss change trend curve, and the opening adjustment compensation function is constructed by adopting a piecewise linearization method, and the opening adjustment correction amount is obtained from the compensation function.

9. The method according to claim 1, characterized in that: The combined inventory early warning system integrates the pressure fluctuation, valve opening and oil loss data into the inventory management platform, coordinates the operation status of multiple storage tanks and monitors the whole process. If the loss exceeds the preset loss threshold, an early warning signal is issued through the inventory management platform, including: Obtain the pressure data collected by the tank pressure sensor, the opening value recorded by the valve position feedback device, and the oil loss data collected by the loss meter, and generate a multi-parameter correlation sequence through the data time tag; According to the multi-parameter association sequence, data is segmented according to a fixed time window, and a time series verification rule is used to remove abnormal data to obtain a feature vector of the operation process; The operation process characteristic vector is used to construct a storage tank operation state matrix, and the inventory change relationship of adjacent storage tanks is calculated through a state correlation function to obtain the inventory fluctuation rate; The loss impact factor is calculated based on the inventory fluctuation rate, and the multi-tank collaborative operation rules are used to generate linkage control instructions, and the operating parameters of each tank are adjusted through the control instruction executor.

Images