A real-time oil yield optimization method for grain and oil processing

Abstract

The application provides a grain oil processing process oil yield real-time optimization method, comprising the following steps: collecting test data of a refining workshop of a grain oil processing enterprise in a preset period, carrying out denoising processing, obtaining standardized characteristic data, and obtaining DCS process parameter data of each time period in the preset period; adjusting the degumming water flow, the phosphoric acid addition amount, the excess alkali and the alkali concentration in the reaction kettle according to a process parameter adjustment scheme, and obtaining the test data and the DCS process parameter of the reaction kettle after adjustment, and predicting the refining yield improvement trend; applying the speed range and power distribution to the grain oil processing process, monitoring and calculating the deviation of the refining yield and the target refining yield in real time, combining the feedback of the test data and the DCS process parameter, and identifying the grain oil processing oil yield.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for real-time optimization of oil yield in grain and oil processing. Background Technology

[0002] Oil refining is a crucial step in the food industry to ensure oil quality and improve economic efficiency. Its core objective is to improve refining yield by optimizing process parameters while ensuring stable product quality. With rising consumer demand and stricter environmental requirements, precise control of the refining process has become a focus of industry attention. However, existing methods often face inefficiencies and instability when dealing with fluctuations in raw oil quality and dynamic changes in process parameters. Many traditional solutions rely on experience to adjust or fix parameters, making it difficult to adapt to real-time changes in key indicators such as acid value and phosphorus content of raw oils. This leads to decreased efficiency and significant resource waste in processes such as degumming and desoaping. In this field, fluctuations in stirring power have become a core technical factor affecting refining yield and process stability. Fluctuations in raw oil quality or process parameters often trigger abnormal stirring, such as excessive soap content in desoaped oil or abnormal phosphorus content in washed oil. These problems stem from an imbalance between shear force and mixing uniformity during stirring. Furthermore, this imbalance directly affects the precise addition of degumming water, phosphoric acid, and alkali, leading to decreased oil yield and fluctuations in product quality. The key issue in optimizing refining yield and stability is how to construct a mapping relationship between shear force and mixing uniformity by dynamically analyzing the fluctuation characteristics of stirring power, combining standardized test data, and generating a process scheme for real-time adjustment of degummed water, phosphoric acid, and alkali content. Summary of the Invention

[0003] This invention provides a method for real-time optimization of oil yield in grain and oil processing, mainly including:

[0004] Test data from the refining workshop of a grain and oil processing enterprise are collected over a predetermined period. The test data is then denoised to obtain standardized feature data. DCS process parameter data for each time period within the predetermined period are acquired. Based on the standardized feature data and the DCS process parameter data, a stirring power fluctuation spectrum distribution characteristic is generated. Abnormal fluctuation signals of the stirring power fluctuation spectrum distribution characteristic are identified, and the dynamic trend of abnormal stirring power fluctuation amplitude is determined. Based on the dynamic trend and the standardized feature data, combined with the soap content of desoaped oil, phosphorus content of washed oil, excessive alkali, degummed water, and alkali concentration... Upper and lower limit constraints are used to construct an optimization model for the refining process. The standardized feature data, the dynamic trend, and the DCS process parameter data are input, and the process parameter combination is output to generate a process parameter adjustment scheme. The refining yield improvement trend is predicted based on the process parameter adjustment scheme. The refining yield improvement trend is analyzed to identify insufficient mixing uniformity or uneven shear force distribution, determine the target shear force distribution parameters, generate the blade angle adjustment range, determine the speed range and power distribution of the stirring system, apply it to the grain and oil processing process, calculate the deviation between the refining yield and the target refining yield, and identify the oil yield of grain and oil processing.

[0005] Furthermore, the test data collected from the refining workshop of the grain and oil processing enterprise during a predetermined period are processed to remove noise and obtain standardized feature data, including:

[0006] Collect and denoise test data from the refining workshop of a grain and oil processing enterprise for a preset period to obtain standardized feature data; acquire DCS process parameter data for each time period within the preset period; calculate the mean and variance of stirring frequency, liquid level and back pressure for each time period based on the denoised process parameter data, and perform normalization processing; integrate the normalized test data and process parameter data, align them according to the time series, and form standardized feature data including crude oil acid value, crude oil phosphorus content, degummed oil acid value, degummed oil phosphorus content, stirring frequency, liquid level and back pressure.

[0007] Furthermore, the step of generating a stirring power fluctuation spectrum distribution feature based on the standardized feature data and the DCS process parameter data, and determining the dynamic trend of the abnormal fluctuation amplitude of the stirring power based on the abnormal fluctuation signal of the stirring power fluctuation spectrum distribution feature, includes:

[0008] The system acquires the power data of the stirring motor, uses spectral decomposition to generate the amplitude distribution of each frequency component, and identifies the dominant frequency component. It extracts the frequency value of the dominant frequency component as the peak value of the spectrum, calculates the reciprocal of the peak value as the fluctuation period, and statistically analyzes the amplitude distribution to form the spectral distribution characteristics of the stirring power fluctuation. The system then processes the power data using a sliding window method, calculates the mean and standard deviation of the power data within the window, and extracts abnormal fluctuation signals. Finally, it calculates the fluctuation amplitude of the abnormal fluctuation signals, arranges the fluctuation amplitudes for each time period, calculates the rate of change, and determines the dynamic trend of the abnormal fluctuation amplitude of the stirring power.

[0009] Furthermore, based on the dynamic trend and the standardized feature data, combined with the upper and lower limits constraints of soap content in desoaped oil, phosphorus content in washed oil, excessive alkali, degummed water, and alkali concentration, a refining process optimization model is constructed. The standardized feature data, the dynamic trend, and the DCS process parameter data are input, and a combination of process parameters is output to generate a process parameter adjustment scheme, including:

[0010] Upper and lower limits are set for the soap content of desoaped oil, phosphorus content of washed oil, excess alkali, degummed water, and alkali concentration. Using the standardized characteristic data and the dynamic trend of abnormal fluctuations as inputs, and the excess alkali, degummed water, alkali concentration, and reactor stirring frequency as decision variables, an optimization model for the refining process is constructed to solve for the numerical combination of decision variables. Based on the numerical combination, the adjustment amounts for excess alkali, degummed water, alkali concentration, and reactor stirring frequency are calculated to generate the process parameter combination. Based on the adjustment amounts of each parameter in the process parameter combination, combined with the response characteristics of the refining equipment, the adjustment sequence and adjustment range of each parameter are determined, generating a process parameter adjustment scheme that includes parameter name, adjustment direction, adjustment amount, and execution sequence.

[0011] Furthermore, the prediction of the refining yield improvement trend based on the process parameter adjustment scheme includes:

[0012] According to the process parameter adjustment plan, the degumming water flow rate, the amount of phosphoric acid added, the excess alkali, and the alkali concentration are adjusted. The adjusted test data and DCS process parameters are collected. The refining yield is calculated, and time series data is constructed. The time series data is fitted to predict the improvement trend of the refining yield.

[0013] Furthermore, the analysis of the refining yield improvement trend, identification of insufficient mixing uniformity or uneven shear force distribution, determination of target shear force distribution parameters, and generation of blade angle adjustment range include:

[0014] The refining yield improvement trend was analyzed, and the coefficient of variation of yield values ​​in adjacent time periods was calculated to determine insufficient mixing uniformity. The flow field inside the reactor was simulated, and the velocity gradient value at each point in the flow field was obtained as a shear force index to generate shear force spatial distribution data. Based on the shear force spatial distribution data, the difference between the maximum and minimum shear force values ​​was calculated, and the tilt angle of the stirring blades was adjusted to obtain the shear force difference values ​​corresponding to multiple angles. The angle corresponding to the minimum difference value was selected as the optimized stirring blade tilt angle. Based on the optimized stirring blade tilt angle, the concentration value of the mixture of feed oil and additives was measured, the standard deviation of the concentration was calculated, the relationship between the shear force difference and the mixing uniformity was determined, the target shear force distribution parameters to meet the mixing uniformity were calculated, and the blade angle adjustment range was generated.

[0015] Furthermore, determining the speed range and power distribution of the stirring system includes:

[0016] Based on the refining yield improvement trend and the standardized characteristic data, calculate the material characteristic index and generate the speed matching coefficient; based on the speed matching coefficient and the blade angle adjustment range, calculate the speed range; based on the speed range and motor power, divide the speed segment and generate the speed adjustment command and power allocation.

[0017] Furthermore, the calculation of the deviation between the refining yield and the target refining yield, and the identification of the oil yield in grain and oil processing, includes:

[0018] Using the aforementioned speed range and power distribution, the crude oil feed rate and finished oil output rate are collected to calculate the refining yield; the deviation between the refining yield and the target refining yield is calculated; based on the test data and the DCS process parameter data, the impurity removal index and oil retention index are calculated to generate the grain and oil processing oil yield.

[0019] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0020] This invention discloses a real-time optimization method for oil yield in grain and oil processing. Addressing the complex correlation between laboratory data and process parameters in refining workshops, it solves the comprehensive problems of fluctuating stirring power, insufficient mixing uniformity, and yield improvement. By collecting laboratory data such as acid value and phosphorus content of crude oil and degummed oil, noise reduction and standardization are performed. Combined with DCS process parameters such as stirring frequency, liquid level, and centrifuge back pressure, the spectral characteristics of stirring power fluctuations are extracted to identify abnormal fluctuation trends. Based on these trends, an optimization model is constructed, incorporating constraints such as soap content in degummed oil and phosphorus content in washed oil, to output process parameter adjustment schemes, dynamically adjusting degumming water flow rate and phosphoric acid addition. Simultaneously, time series analysis is used to predict yield improvement trends, identify insufficient mixing uniformity, and optimize the stirring blade angle and rotation speed using fluid dynamics to determine the target shear force distribution and generate rotation speed adjustment commands. This invention achieves precise improvement in oil yield by real-time monitoring of yield deviations and parameter feedback, significantly improving refining efficiency and product quality. Attached Figure Description

[0021] Figure 1 This is a flowchart of a method for real-time optimization of oil yield in grain and oil processing according to the present invention.

[0022] Figure 2 This is a schematic diagram of a real-time optimization method for oil yield in grain and oil processing according to the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0024] like Figure 1-2 This embodiment of a method for real-time optimization of oil yield in grain and oil processing may specifically include:

[0025] Step S101: Collect test data from the refining workshop of a grain and oil processing enterprise within a preset period, perform noise reduction processing to obtain standardized feature data, and obtain DCS process parameter data for each time period within the preset period.

[0026] By collecting laboratory data from the refining workshops of grain and oil processing enterprises during a predetermined period, including crude oil acid value, crude oil phosphorus content, degummed oil acid value, and degummed oil phosphorus content, the raw data were denoised. Outliers deviating from the mean by more than three standard deviations were removed using a moving average method. The denoised data was then normalized to obtain standardized feature data. DCS process parameter data for each time period within the predetermined period were acquired, including stirring frequency, liquid level, and centrifuge back pressure. This process parameter data was also denoised using a moving average method, removing outliers deviating from the mean by more than three standard deviations. Based on the denoised process parameter data, the mean and variance of stirring frequency, liquid level, and back pressure for each time period were calculated and normalized to ensure the process parameter data and laboratory data were within the same numerical range. The normalized laboratory data and process parameter data were integrated and aligned according to time series to form standardized feature data including crude oil acid value, crude oil phosphorus content, degummed oil acid value, degummed oil phosphorus content, stirring frequency, liquid level, and back pressure.

[0027] Specifically, in the refining workshops of grain and oil processing enterprises, crude oil needs to undergo multiple processes to become finished oil that meets edible standards. Impurities such as free fatty acids and phospholipids in crude oil can affect the stability and taste of the oil, and therefore need to be removed through refining processes. Acid value reflects the content of free fatty acids in the oil, while phosphorus content represents the content of phospholipids and other colloidal substances; these two indicators are key parameters for evaluating oil quality.

[0028] In one possible implementation, the data collection frequency is typically every 2 hours, including sampling from crude oil storage tanks to determine the acid value and phosphorus content of crude oil, and sampling from the degumming process to determine the corresponding indicators of the degummed oil. Due to the complex on-site environment, the collected raw data often contains noise interference. The moving average method smooths the curve by calculating the average of data at adjacent time points. When a data point deviates from the mean of that period by more than 3 times the standard deviation, it is identified as an outlier and removed. This processing method can effectively remove abnormal data caused by equipment failure or operational errors, improving the accuracy of subsequent analysis.

[0029] Specifically, normalization is the process of converting data with different dimensions to the same numerical range. The unit for acid value is mgKOH / g, typically varying between 0.5 and 5; the unit for phosphorus content is mg / kg, potentially ranging from 10 to 200. By mapping these data to the 0-1 interval using a normalization formula, comparisons and comprehensive analysis between different indicators can be achieved. This standardization lays the foundation for subsequent data fusion.

[0030] It should be noted that the DCS process parameters are production process data collected in real time through the distributed control system. The stirring frequency of the reaction tank directly affects the mixing effect of grease and degumming agent; too low a frequency will lead to incomplete reaction, while too high a frequency will generate excessive foam, affecting the separation effect. The liquid level in the reaction tank reflects the residence time of the material; too high a level will prolong the reaction time but reduce the throughput, while too low a level may lead to incomplete reaction. The back pressure of the centrifuge is a key parameter in the separation process; appropriate back pressure can improve oil-water separation efficiency, but excessive back pressure will increase grease loss.

[0031] In one embodiment, these process parameters also require denoising and normalization. The stirring frequency is typically adjusted within the range of 30-60 Hz, the liquid level is controlled between 40% and 80% of the tank height, and the back pressure varies within the range of 0.2-0.8 MPa. After processing using the same moving average denoising method, normalization is performed to convert all parameters to the standard range of 0-1.

[0032] Preferably, time series alignment is a key step in data integration. Laboratory data is collected intermittently, while DCS parameters are recorded continuously. By matching timestamps, the DCS parameters closest to the laboratory test time are associated with the corresponding test data, forming complete standardized feature data.

[0033] Step S102: Based on the standardized feature data and the DCS process parameter data for the corresponding time period, obtain the spectral distribution characteristics of the stirring power fluctuation, extract the abnormal fluctuation signal of the power curve generated by time series analysis and spectral decomposition, and determine the dynamic trend of the abnormal fluctuation amplitude of the stirring power.

[0034] Based on standardized feature data and DCS process parameter data for the corresponding time period, real-time power data of the stirring motor is obtained. Fast Fourier Transform (FFT) is used to perform spectral decomposition on the power data to obtain the amplitude distribution of each frequency component. Frequency components with amplitudes exceeding a preset multiple of the average amplitude are identified as the dominant frequency components. For the dominant frequency components, their frequency values ​​are extracted as spectral peaks. The reciprocal of the spectral peak is calculated to obtain the corresponding fluctuation period. The amplitudes of all frequency components are statistically analyzed and sorted from largest to smallest to obtain amplitude distribution characteristics, forming a stirring power fluctuation spectral distribution feature that includes spectral peaks, fluctuation periods, and amplitude distribution. Based on the fluctuation period in the stirring power fluctuation spectral distribution feature, a sliding window method is used to segment the original power data. The window length is set to an integer multiple of the fluctuation period. The mean and standard deviation of the power data within each window are calculated. If the difference between the power value and the mean within the window exceeds three times the standard deviation, the power curve for that time period is extracted as an abnormal fluctuation signal. For the extracted abnormal fluctuation signal, the difference between its maximum value and the average normal power is calculated as the fluctuation amplitude. The fluctuation amplitude values ​​of each time period are arranged in chronological order. The rate of change is obtained by dividing the difference of the fluctuation amplitude of adjacent time periods by the time interval. The dynamic trend of the abnormal fluctuation amplitude of the stirring power is determined according to the sign and magnitude of the rate of change.

[0035] Specifically, the power data of the stirring motor is an important indicator reflecting the stability of the refining process. In the grain and oil refining process, the stirrer is responsible for thoroughly mixing the oil and degumming agent, and its power fluctuations directly reflect the uniformity of the mixing process and the stability of the reaction. The Fast Fourier Transform (FFT) is a mathematical method that converts a time-domain signal into a frequency-domain signal. Through this transformation, the periodic components implicit in the power signal can be identified.

[0036] Specifically, the power of the stirring motor is relatively stable during normal operation, but it will fluctuate periodically when the viscosity of the material changes, the stirring blades wear, or local agglomeration occurs in the reaction vessel. By performing spectral decomposition on the power data over a period of time, the amplitude distribution of different frequency components can be obtained. The frequency components with larger amplitudes represent the main components of the power fluctuation, and the frequency values ​​of these main frequency components are the spectral peaks.

[0037] In one possible implementation, after the power data undergoes spectral decomposition, a series of frequency-amplitude pairs are obtained. Assuming the amplitude of a certain frequency component is 50W, and the average amplitude of all frequency components is 10W, this frequency component is identified as the dominant frequency when the preset multiplier is 3. The reciprocal of the spectral peak value directly corresponds to the fluctuation period. If the spectral peak value is 0.1Hz, the corresponding fluctuation period is 10 seconds, meaning that the power completes a periodic change every 10 seconds.

[0038] It should be noted that obtaining the amplitude distribution characteristics involves statistical analysis of the amplitudes of all frequency components. By sorting the amplitudes from largest to smallest, we can understand the distribution of power fluctuation energy at different frequencies.

[0039] This distribution characteristic helps determine whether fluctuations are caused by a single reason or the result of multiple factors. The sliding window method is a commonly used technique in time series analysis. The choice of window length is crucial; setting it to an integer multiple of the fluctuation period ensures that each window contains the complete fluctuation period. Within each window, the statistical characteristics of the power are calculated. When the power value for a certain period deviates significantly from the normal range, it indicates abnormal fluctuations during that period. Three standard deviations is a commonly used anomaly detection threshold in engineering; based on the normal distribution theory, data points exceeding this range have an extremely low probability of occurrence.

[0040] Preferably, dynamic trend analysis of abnormal fluctuation signals is of great significance for predicting equipment status and optimizing process parameters. The fluctuation amplitude is calculated using the difference between the maximum value and the normal average value; this method can intuitively reflect the degree of abnormality. By calculating the rate of change of fluctuation amplitude between adjacent time periods, it can be determined whether the abnormality is intensifying or mitigating. A positive rate of change indicates that the fluctuation amplitude is increasing, which may foreshadow the development of equipment failure; a negative rate of change indicates that the fluctuation amplitude is decreasing, indicating that the abnormal situation is improving.

[0041] Step S103: Based on the dynamic trend of abnormal fluctuation amplitude, and simultaneously satisfying the upper and lower limits of soap content in desoaped oil, phosphorus content in washed oil, and soap content in washed oil, as well as the upper and lower limits of excess alkali, degummed water, and alkali concentration, construct a refining process optimization model, input standardized feature data, dynamic trend of stirring power, and DCS process parameters, output process parameter combinations, and generate process parameter adjustment schemes based on the process parameter combinations.

[0042] Based on the dynamic trend of abnormal fluctuations, upper and lower limits for soap content in desoaped oil, phosphorus content in washed oil, and soap content in washed oil are set as quality constraints. Upper and lower limits for excess alkali addition, degumming water addition, and alkali concentration are set as process constraints, forming a constraint set containing six types of upper and lower limit constraints. The upper and lower limits of each item in the constraint set are used as constraints for linear programming. The objective function is to minimize the deviation between the oil quality indicators and the target value. Standardized characteristic data and the dynamic trend of stirring power are used as input parameters. The values ​​of excess alkali addition, degumming water addition, alkali concentration, and stirring frequency are used as decision variables to construct an optimization model for the refining process, and the optimal numerical combination of each decision variable is obtained. Based on the optimal numerical combination, the difference between the current value and the optimal value of each parameter (excess alkali addition, degumming water addition, alkali concentration, and stirring frequency) is calculated to determine the adjustment direction and amount for each parameter, forming a process parameter combination that includes parameter name, current value, target value, and adjustment amount. Based on the adjustment amount of each parameter in the process parameter combination and combined with the response characteristics of the refining equipment, the adjustment sequence and adjustment range of each parameter are determined, and a process parameter adjustment scheme including parameter name, adjustment direction, adjustment amount and execution sequence is generated.

[0043] Specifically, setting the constraints during the refining process is a crucial step in ensuring product quality. The soap content of desoaped oil directly affects the workload of subsequent washing processes; too high a soap content leads to incomplete washing, while too low a content indicates over-refining with alkali, resulting in oil loss. The phosphorus content of washed oil reflects the degumming effect; residual phospholipids can affect the oxidative stability of the oil. The soap content of washed oil is an important indicator for evaluating the washing effect; residual soap will hydrolyze during storage, producing free fatty acids.

[0044] In one possible implementation, quality constraints are set based on national standards and internal enterprise control standards. The upper limit for soap content in desoaped oil is typically set at 500 mg / kg, and the lower limit at 200 mg / kg; the upper limit for phosphorus content in washed oil is 10 mg / kg, and the lower limit at 3 mg / kg; the upper limit for soap content in washed oil is 50 mg / kg, and the lower limit at 10 mg / kg. These constraints ensure that the processing effect of each step is within a reasonable range.

[0045] Specifically, process constraints involve controlling the amount of raw materials added. Excess alkali refers to the amount added exceeding the theoretical neutralization amount, typically 10%-30% of the theoretical amount. Too little alkali will not adequately neutralize free fatty acids, while too much will cause oil saponification losses. The amount of degumming water added affects the hydration effect of phospholipids and is generally controlled at 1%-3% of the oil weight. Alkali concentration determines the speed and thoroughness of the alkali refining reaction; too high a concentration will exacerbate oil loss, while too low a concentration will slow the reaction. The linear programming method is applied here to find the optimal solution by establishing an objective function and constraints. The objective function is set as the weighted sum minimization of the deviations between oil quality indicators and target values, with weight coefficients determined according to the importance of each indicator. Decision variables include excess alkali addition, degumming water addition, alkali concentration, and stirring frequency; these parameters directly affect the refining effect. Standardized characteristic data and the dynamic trend of stirring power serve as input parameters to the model, reflecting the current production status.

[0046] It should be noted that the process of finding the optimal numerical combination takes into account the interrelationships of multiple constraints. When abnormal fluctuations occur in the stirring power, it may be necessary to reduce the stirring frequency to avoid equipment damage, but this will affect the mixing effect. Therefore, it is necessary to adjust the alkali concentration or extend the reaction time accordingly to compensate. The model finds the optimal parameter combination that satisfies all constraints by balancing various factors.

[0047] In one embodiment, the adjustment of process parameters needs to consider the response characteristics of the equipment. Adjusting the alkali concentration is achieved through dilution or concentration, which has a relatively long response time; the stirring frequency is adjusted via a frequency converter, providing a rapid response; and the addition of excess alkali and degumming water is controlled by a metering pump, requiring high precision. The adjustment sequence is determined based on the ease of adjustment of each parameter and its impact on product quality. Parameters with a fast response and significant impact, such as the stirring frequency, are adjusted first, followed by the addition amount parameters, and finally the concentration parameters.

[0048] Step S104: Adjust the degumming water flow rate, phosphoric acid addition amount, excess alkali and alkali concentration in the reactor according to the process parameter adjustment plan, and obtain the test data and DCS process parameters of the reactor after adjustment to predict the trend of refining yield improvement.

[0049] According to the process parameter adjustment plan, the flow rate of degumming water in the reactor is adjusted to the target value by controlling the valve opening, the amount of phosphoric acid added is adjusted to the specified value by the metering pump, the amount of excess alkali added is adjusted by the alkali transfer pump, and the alkali concentration is adjusted to the set value by the online mixing device. The time points when each parameter reaches the target value are recorded. After each parameter reaches the target value, the test data of the reactor output are collected, including the acid value of crude oil, the phosphorus content of crude oil, the acid value of degummed oil, and the phosphorus content of degummed oil. At the same time, the actual values ​​of the adjusted DCS process parameters are obtained, including the flow rate of degumming water, the amount of phosphoric acid added, the amount of excess alkali added, and the alkali concentration, forming an adjusted parameter data set. Based on the test data in the adjusted parameter data set, the ratio of the weight of refined oil to the weight of oil before refining is calculated as the refining yield at the current moment. The current refining yield and the refining yield at each historical moment are arranged in chronological order to construct a time series data containing time tags and yield values. An autoregressive moving average model was used to fit the time series data. The historical refining yield series was used as the model input. The model parameters were used to estimate the predicted refining yield for future periods. The slope of the predicted yield series was calculated, and the trend of refining yield improvement was determined based on the sign and magnitude of the slope.

[0050] Specifically, precise adjustment of the reactor process parameters is a key step in optimizing the refining process. Adjusting the valve opening requires consideration of fluid characteristics and pipeline pressure. When the target flow rate is 100 L / h, the valve opening is gradually adjusted using a PID control algorithm based on the current flow meter feedback value until the set value is reached. The use of a metering pump ensures the accuracy of phosphoric acid addition. Its working principle is to achieve quantitative delivery through the reciprocating motion of a plunger. The displacement of each stroke is fixed, and the addition rate is controlled by adjusting the stroke frequency.

[0051] In one possible implementation, online adjustment of the alkali concentration is achieved by mixing dilution water and concentrated alkali in a specific ratio. When the alkali concentration needs to be adjusted from 12% to 10%, the required amount of dilution water is calculated, and the target concentration is achieved by adjusting the dilution water flow rate. This online mixing method avoids the tedious process of pre-preparing alkali solutions of different concentrations, improving production flexibility. Recording the time points when each parameter reaches the target value is crucial, as this time information is used for subsequent analysis of the lag effect of parameter adjustments on product quality.

[0052] Specifically, the timing of data collection is crucial. Refining reactions typically require time to reach equilibrium; sampling too early can result in data that doesn't accurately reflect the effects of adjustments. Crude oil acid value and phosphorus content reflect the quality of the feedstock, while degummed oil acid value and phosphorus content demonstrate the effectiveness of the degumming process. The degree of acid value reduction directly affects alkali consumption in the alkali refining process, while the phosphorus removal rate influences subsequent decolorization and deodorization. Refining yield is a core indicator for measuring refining economics; its calculation method, though seemingly simple, contains significant information. Assuming 1000 kg of crude oil yields 950 kg of finished oil after refining, the refining yield is 95%. This 5% loss includes removed impurities, neutral oil carried away by soapstock, and washing losses. Continuously monitoring changes in refining yield allows for the assessment of the effectiveness of process adjustments. The construction of time-series data must ensure data continuity and comparability; each data point should include a timestamp and the corresponding yield value.

[0053] It should be noted that the autoregressive moving average model is a classic method for time series forecasting. This model assumes a linear relationship between current and historical values ​​while also considering the impact of random disturbances. The autoregressive part of the model uses historical yield values ​​to predict future values, while the moving average part handles the correlation of prediction errors. By fitting historical data, the model can identify the regular and trend components of yield changes.

[0054] In one embodiment, the determination of the refining yield improvement trend is based on slope analysis of the predicted sequence. A positive slope indicates an upward trend in yield, and the magnitude of the slope reflects the rate of improvement. When the slope is 0.1% / day, it means that the refining yield increases by an average of 0.1 percentage points per day. This quantitative trend analysis provides a scientific basis for evaluating the effectiveness of process optimization and also provides guidance for further parameter adjustments.

[0055] Step S105: Identify the phenomenon of insufficient mixing uniformity or uneven shear force distribution in the improvement trend, adaptively optimize the tilt angle of the stirring blade and the stirring efficiency according to the fluid dynamic characteristics, identify the correlation between shear force distribution and the mixing uniformity of the raw oil, determine the target shear force distribution parameters according to the relationship, and obtain the blade angle adjustment range.

[0056] By analyzing the fluctuation characteristics in the trend data of refining yield improvement, the ratio of the standard deviation to the average yield value of adjacent time periods is calculated as the coefficient of variation. If the coefficient of variation exceeds a preset threshold, it is determined that there is insufficient mixing uniformity. Computational fluid dynamics is used to simulate the flow field inside the reactor, and the velocity gradient value at each point inside the reactor is obtained as a shear force index to obtain shear force spatial distribution data. Based on the shear force spatial distribution data, the difference between the maximum and minimum shear force values ​​is calculated, and this difference is used to assess the degree of distribution unevenness. Multiple flow field simulations are performed by changing the tilt angle parameter of the stirring blades, and the shear force difference corresponding to each angle is recorded. The angle corresponding to the smallest difference is selected as the optimized blade tilt angle. Based on the optimized blade tilt angle, actual stirring tests are conducted, and the concentration values ​​of the mixture of feedstock oil and additives at different time points are measured. The standard deviation of the concentration is calculated as a mixing uniformity index, and the relationship curve between the shear force difference and the mixing uniformity index is plotted to determine the range of shear force difference corresponding to the achievement of mixing uniformity. Based on the range of shear force difference, the set of blade tilt angles that meet this range is calculated in reverse. Combined with the structural limitations of the mixing equipment, the adjustable upper and lower limits of the blade tilt angle are determined, thus obtaining the blade angle adjustment range.

[0057] Specifically, the coefficient of variation, as a measure of relative dispersion, eliminates the influence of dimensions by using the ratio of the standard deviation to the mean. When the refining yields of five consecutive batches are 94.5%, 95.2%, 93.8%, 95.5%, and 94.0%, the calculated average is 94.6%, the standard deviation is 0.7%, and the coefficient of variation is 0.74%. If the preset threshold is 0.5%, it is determined that there is uneven mixing, and this fluctuation often stems from an unreasonable flow field distribution inside the reactor.

[0058] In one possible implementation, computational fluid dynamics (CFD) methods simulate the flow field distribution within a reactor by numerically solving the fluid motion equations. The velocity gradient, as a characterization of shear force, reflects the relative motion intensity between different layers within the fluid. During stirring, the fluid velocity is highest near the blades and lower in areas farther from the blades; this velocity difference generates shear force. By setting virtual monitoring points within the reactor, the velocity gradient values ​​at each location can be obtained, forming a three-dimensional shear force distribution map.

[0059] Specifically, the uniformity of shear force distribution directly affects the contact efficiency between oils and chemical reagents. When the maximum shear force occurs at the blade edge (500 / s) and the minimum occurs near the vessel wall (50 / s), the difference reaches 450 / s, indicating a severe uneven distribution. The tilt angle of the stirring blades is a key parameter affecting the flow field distribution; too small an angle leads to insufficient axial flow, while too large an angle increases power consumption and may cause cavitation.

[0060] It should be noted that the optimal configuration can be found by systematically changing the blade tilt angle through flow field simulation. Simulation calculations were performed every 5 degrees as the angle gradually increased from 30 degrees to 60 degrees. The results show that the shear force difference drops to a minimum of 200 / s at 45 degrees. This optimization process considers the fluid circulation pattern, ensuring both good axial circulation and maintaining appropriate radial flow.

[0061] In one embodiment, the design of the actual stirring test requires strict control of variables. At an optimized 45-degree blade angle, mixture samples are collected every 30 seconds from three sampling ports (top, middle, and bottom) of the reactor to determine the additive concentration. Initially, the concentrations at different points vary significantly, but gradually become more uniform as stirring continues. The mixing effect can be quantitatively assessed by calculating the change in the standard deviation of the concentration. When the standard deviation drops below 5% of the initial value, a fully mixed state is considered achieved.

[0062] Preferably, establishing a curve showing the relationship between shear force difference and mixing uniformity is of significant guiding importance. Experimental data indicates that when the shear force difference is controlled within the range of 150-250 g / s, the mixing time is shortest and the final mixing effect is optimal. Based on this finding, the corresponding blade angle range can be calculated in reverse. Considering the limitations of machining accuracy and adjustment devices, the blade angle adjustment range is determined to be 40-50 degrees.

[0063] Step S106: Dynamic trend calculation of standardized feature data and abnormal fluctuation amplitude of stirring power to obtain speed matching coefficient. Combined with the adaptively optimized stirring power and blade angle adjustment range, generate speed adjustment command for reactor stirring, and determine the speed range and power distribution of the stirring system.

[0064] Based on the dynamic change rate of the refining yield improvement trend and the changes in acid value and phosphorus content in standardized characteristic data, the weighted average of the acid value change rate and phosphorus content change rate is calculated as a material characteristic index. The abnormal fluctuation range of the stirring power is divided by the material characteristic index to obtain a speed matching coefficient reflecting the degree of adaptation between the stirring load and the speed. The speed matching coefficient is multiplied by the current stirring speed to obtain a baseline speed value. Based on the relationship that stirring power is proportional to the cube of speed, and combined with the adaptively optimized stirring power and the upper and lower limits of the blade angle adjustment range, the upper speed limit corresponding to the upper blade angle and the lower speed limit corresponding to the lower blade angle are calculated. Based on the upper and lower speed limits, the rated power of the reactor motor is obtained, and the available power margin is obtained by subtracting the current actual power consumption. If the power required for the upper speed limit exceeds the available power margin, a new upper speed limit is calculated based on the available power margin, forming a speed range. According to the speed range, the speed values ​​within the range are divided into several speed segments at preset intervals. The power demand value corresponding to each speed segment according to the speed cube relationship is calculated, generating speed adjustment instructions containing each speed segment and its corresponding power value, thus determining the speed range and power distribution of the stirring system.

[0065] Specifically, during the refining process, changes in acid value and phosphorus content directly affect the viscosity and flowability of the material. When the acid value decreases from 5.0 mg KOH / g to 3.0 mg KOH / g, the change rate is 40%; when the phosphorus content decreases from 150 mg / kg to 50 mg / kg, the change rate is 66.7%. Using weighted averages with coefficients of 0.6 and 0.4, a material characteristic index of 50% is obtained. The higher this index, the more drastic the change in material properties, and the higher the requirements for agitation.

[0066] In one possible implementation, the introduction of a speed matching coefficient solves the problem of mismatch between the stirring load and the speed. When the abnormal fluctuation range of the stirring power is 30%, and the material characteristic index is 50%, the speed matching coefficient is 0.6. This means that the current speed is too high and needs to be appropriately reduced. If the current speed is 100 rpm, multiplying it by the matching coefficient of 0.6 yields a base speed of 60 rpm, which is more suitable for the current material state.

[0067] Specifically, the cubic relationship between stirring power and rotational speed is a fundamental principle of fluid mechanics. When the rotational speed doubles, power consumption increases eightfold. This relationship plays a crucial role in determining the rotational speed range. Assuming the optimal rotational speed corresponding to an upper limit of 50 degrees for the blade angle is 80 rpm, and the optimal speed corresponding to a lower limit of 40 degrees is 60 rpm, combined with a baseline speed of 60 rpm, the initial rotational speed range is determined to be 60-80 rpm. Each angle corresponds to a different flow pattern and mixing efficiency; the impact of angle changes is compensated for by adjusting the rotational speed.

[0068] It should be noted that the power margin calculation takes into account the safe operation of the equipment. The rated power of the reactor motor is typically 30kW, and the current actual consumption is 15kW, so the available power margin is 15kW. If the power required for 80rpm is 20kW, exceeding the available margin, a recalculation is necessary. Based on the cubic relationship, substituting the available power of 15kW into the recalculation yields a new upper speed limit of approximately 72rpm. This power constraint ensures that the equipment will not operate under overload.

[0069] In one embodiment, the speed range is divided using equal intervals. The 60-72 rpm range is divided into four speed segments: 60, 64, 68, and 72 rpm, with each segment requiring 4 rpm intervals. Power is calculated for each segment based on cubic relationships: 8 kW is required at 60 rpm, 9.7 kW at 64 rpm, 11.6 kW at 68 rpm, and 13.8 kW at 72 rpm. This segmentation method allows operators to select the appropriate operating point based on actual working conditions.

[0070] Preferably, the generated speed adjustment command includes complete operating parameters. The command clearly indicates the path from the current speed to the target speed, the dwell time for each speed segment, and the corresponding expected power value. This detailed command avoids arbitrary operation and ensures the stability of the refining process. By precisely matching speed and power, both the requirements for mixing effect are met, and energy consumption is optimized, thus improving the economic efficiency of the refining workshop.

[0071] Step S107: Apply the rotation speed range and power distribution to the grain and oil processing process. By real-time monitoring and calculation of the deviation between the refining yield and the target refining yield, and combining the feedback from the test data and DCS process parameters, identify the oil yield of the grain and oil processing.

[0072] The determined speed range and power distribution scheme are input into the reactor control device. The stirring equipment operates according to the set speed and corresponding power, and the crude oil feed rate and finished oil output rate data are collected in real time during the refining process. The actual refining yield of the current batch is calculated by dividing the finished oil output rate by the crude oil feed rate. The actual refining yield is compared with the preset target refining yield, and the difference between the two is calculated as the refining deviation value. At the same time, the batch's test data, including the phosphorus content of degummed oil and the soap content of washed oil, as well as DCS process parameters, including the actual stirring speed and stirring power consumption, are obtained to form a complete batch dataset. Based on the batch dataset, the percentage of phosphorus removal and the percentage of soap removal are calculated as impurity removal indicators. The ratio of the actual refining yield to the theoretical maximum yield is used as the oil retention indicator. The oil yield value, which reflects the efficiency of grain and oil processing, is obtained by multiplying the impurity removal indicator by the oil retention indicator, thus completing the identification of the oil yield of grain and oil processing.

[0073] Specifically, the implementation of the speed range and power distribution scheme is a key aspect of refining process optimization. After receiving the speed command, the reactor control device adjusts the motor speed via a frequency converter while simultaneously monitoring power consumption. When the set speed is 65 rpm and the corresponding power is 10 kW, the control system adjusts in real time to maintain this operating state. Accurate metering of the feed and discharge rates directly affects the accuracy of refining yield calculations, and is typically monitored continuously using mass flow meters.

[0074] In one possible implementation, the calculation of the actual refining yield needs to take into account the moisture content of the material. Suppose a batch of 1000 kg of crude oil is input, and after processes such as degumming, deacidification, and washing, 920 kg of finished oil is obtained, with an apparent refining yield of 92%. However, if the crude oil has a moisture content of 2% and the finished oil has a moisture content of 0.1%, a dry basis conversion is required, and the actual refining yield will be slightly different. This precise calculation provides a reliable basis for subsequent deviation analysis.

[0075] Specifically, the target refining yield is set based on the quality characteristics of the feedstock oil. The target yield for high-quality crude oil can reach over 95%, while the target yield for crude oil with a higher acid value may only be 90%. When the actual yield is 92% and the target yield is 94%, the deviation is -2%. This negative deviation indicates the possibility of over-refining or improper operation. Simultaneous acquisition of laboratory data and process parameters provides comprehensive information for problem diagnosis. The construction of batch datasets reflects a systematic approach to process control. Data for each batch includes not only the final results but also process parameters. For example, the phosphorus content in degummed oil decreased from 150 mg / kg to 10 mg / kg, achieving a removal rate of 93.3%; the soap content in washed oil decreased from 300 mg / kg to 30 mg / kg, achieving a removal rate of 90%. These removal rates reflect the processing effectiveness of each step, while the actual stirring speed and power consumption reveal the correlation between equipment operating status and process effectiveness.

[0076] It should be noted that the theoretical maximum yield is an important reference indicator. It represents the ideal yield under the condition that impurities are completely removed and there is no loss of neutral oil. Calculated based on the initial impurity content of the crude oil, if the crude oil contains 2% non-oil substances, the theoretical maximum yield is 98%. When the actual refining yield is 92% and the theoretical maximum yield is 98%, the oil retention index is 93.9%, indicating that 6.1% of neutral oil is lost during the refining process.

[0077] In one embodiment, the calculation of oil yield comprehensively considers both impurity removal effectiveness and oil retention. The impurity removal index is taken as the average of phosphorus removal rate and soap removal rate, such as 91.7%. Multiplying this by the oil retention index of 93.9% yields an oil yield of 86.1%. This value comprehensively reflects the overall efficiency of the refining process, taking into account both product quality and economic benefits. By continuously monitoring and analyzing the trend of oil yield changes, process deviations can be detected in a timely manner, and adjustment measures can be taken to achieve dynamic optimization of the refining process.

[0078] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time optimization of oil yield in a grain oil processing process, characterized by, The method includes: collecting laboratory data from the refining workshop of a grain and oil processing enterprise within a preset period; denoising the laboratory data to obtain standardized feature data; acquiring DCS process parameter data for each time period within the preset period; generating a stirring power fluctuation spectrum distribution feature based on the standardized feature data and the DCS process parameter data; identifying abnormal fluctuation signals of the stirring power fluctuation spectrum distribution feature; and determining the dynamic trend of abnormal fluctuation amplitude of the stirring power based on the dynamic trend and the standardized feature data, combined with factors such as soap content in desoaped oil, phosphorus content in washed oil, excessive alkali, and degumming water. The upper and lower limits of alkali concentration are constrained to construct an optimization model for the refining process. The standardized feature data, the dynamic trend, and the DCS process parameter data are input, and the process parameter combination is output to generate a process parameter adjustment scheme. The refining yield improvement trend is predicted based on the process parameter adjustment scheme. The refining yield improvement trend is analyzed to identify insufficient mixing uniformity or uneven shear force distribution, determine the target shear force distribution parameters, generate the blade angle adjustment range, determine the speed range and power distribution of the stirring system, apply it to the grain and oil processing process, calculate the deviation between the refining yield and the target refining yield, and identify the oil yield of grain and oil processing.

2. The method according to claim 1, wherein, The process of collecting test data from the refining workshop of a grain and oil processing enterprise within a predetermined period, and then denoising the test data to obtain standardized feature data, includes: collecting and denoising test data from the refining workshop of a grain and oil processing enterprise within a predetermined period to obtain standardized feature data; acquiring DCS process parameter data for each time period within the predetermined period, and calculating the mean and variance of stirring frequency, liquid level, and back pressure for each time period based on the denoised process parameter data, and performing normalization processing; integrating the normalized test data and process parameter data, aligning them according to the time series, to form standardized feature data including crude oil acid value, crude oil phosphorus content, degummed oil acid value, degummed oil phosphorus content, stirring frequency, liquid level, and back pressure.

3. The method according to claim 1, wherein, The step of generating a stirring power fluctuation spectrum distribution feature based on the standardized feature data and the DCS process parameter data, and determining the dynamic trend of the abnormal fluctuation amplitude of the stirring power, includes: acquiring stirring motor power data, using spectral decomposition to generate the amplitude distribution of each frequency component, and identifying the dominant frequency component; extracting the frequency value of the dominant frequency component as the peak value of the spectrum, calculating the reciprocal of the peak value of the spectrum as the fluctuation period, and statistically analyzing the amplitude distribution to form a stirring power fluctuation spectrum distribution feature; processing the power data using a sliding window method, calculating the mean and standard deviation of the power data within the window, and extracting the abnormal fluctuation signal; calculating the fluctuation amplitude of the abnormal fluctuation signal, arranging the fluctuation amplitude of each time period, calculating the rate of change, and determining the dynamic trend of the abnormal fluctuation amplitude of the stirring power.

4. The method according to claim 1, wherein, Based on the dynamic trend and the standardized feature data, combined with upper and lower limit constraints on the soap content of desoaped oil, phosphorus content of washed oil, excess alkali, degummed water, and alkali concentration, a refining process optimization model is constructed. The model inputs the standardized feature data, the dynamic trend, and the DCS process parameter data, and outputs a combination of process parameters to generate a process parameter adjustment scheme. This includes setting upper and lower limit constraints on the soap content of desoaped oil, phosphorus content of washed oil, excess alkali, degummed water, and alkali concentration; using the standardized feature data and the dynamic trend of abnormal fluctuation amplitude as input. Using the excess alkali, the degummed water, the alkali concentration, and the stirring frequency of the reaction tank as decision variables, an optimization model for the refining process is constructed, and the numerical combination of the decision variables is solved. Based on the numerical combination, the adjustment amounts of the excess alkali, the degummed water, the alkali concentration, and the stirring frequency of the reaction tank are calculated to generate the process parameter combination. Based on the adjustment amounts of each parameter in the process parameter combination, combined with the response characteristics of the refining equipment, the adjustment sequence and adjustment range of each parameter are determined, and a process parameter adjustment scheme containing parameter name, adjustment direction, adjustment amount, and execution sequence is generated.

5. The method for real-time optimization of oil yield in grain and oil processing according to claim 1, characterized in that, The method of predicting the improvement trend of refining yield based on the process parameter adjustment scheme includes: adjusting the degumming water flow rate, phosphoric acid addition amount, excess alkali, and alkali concentration according to the process parameter adjustment scheme; collecting the adjusted test data and DCS process parameters; calculating the refining yield and constructing time series data; fitting the time series data and predicting the improvement trend of refining yield.

6. The method of claim 1, wherein the method is characterized by, The analysis of the refining yield improvement trend, identification of insufficient mixing uniformity or uneven shear force distribution, determination of target shear force distribution parameters, and generation of blade angle adjustment range includes: analyzing the refining yield improvement trend, calculating the coefficient of variation of yield values ​​in adjacent time periods, and determining insufficient mixing uniformity; simulating the flow field inside the reactor, obtaining the velocity gradient value of each point in the flow field as a shear force index, and generating shear force spatial distribution data; calculating the difference between the maximum and minimum shear force values ​​based on the shear force spatial distribution data, adjusting the tilt angle of the stirring blades, obtaining the shear force difference values ​​corresponding to multiple angles, and selecting the angle corresponding to the minimum difference as the optimized stirring blade tilt angle; based on the optimized stirring blade tilt angle, measuring the concentration value of the mixture of feedstock oil and additives, calculating the concentration standard deviation, determining the relationship between the shear force difference and mixing uniformity, estimating the target shear force distribution parameters that satisfy mixing uniformity, and generating the blade angle adjustment range.

7. The method of claim 1, wherein the method is characterized by, The determination of the speed range and power distribution of the stirring system includes: calculating the material characteristic index and generating a speed matching coefficient based on the refining yield improvement trend and the standardized characteristic data; calculating the speed range based on the speed matching coefficient and the blade angle adjustment range; and dividing the speed range into speed segments and generating the speed adjustment command and power distribution based on the speed range and the motor power.

8. The method of claim 1, wherein the method is characterized by, The calculation of the deviation between the refining yield and the target refining yield, and the identification of the oil yield in grain and oil processing, includes: applying the speed range and power distribution to collect the crude oil feed rate and the finished oil output rate, and calculating the refining yield; calculating the deviation between the refining yield and the target refining yield; and calculating the impurity removal index and oil retention index based on the test data and the DCS process parameter data, and generating the oil yield in grain and oil processing.

Images