Intelligent control method and system for oil processing process parameters for energy consumption optimization

Abstract

The application discloses an intelligent control method and system for oil processing process parameters of energy consumption optimization, and relates to the technical field of process control, comprising: collecting real-time process parameters and raw material quality indexes of each process in the oil refining process, calculating the real-time efficiency of the heat exchanger according to the inlet and outlet temperatures of the heat exchanger, and calculating the temperature gradient matching coefficient between processes according to the temperature of each process and the real-time efficiency of the heat exchanger; determining the quality deviation according to the raw material quality index, and calculating the quality characteristic conflict degree under the current process parameters; taking the minimization of total energy consumption, the maximization of the temperature gradient matching coefficient and the minimization of the quality characteristic conflict degree as multi-objectives, and solving the optimal process parameter combination through an improved optimization algorithm; and adjusting and controlling the execution parameters of each process according to the optimal process parameter combination. The application solves the technical problem that it is difficult to realize the unified optimization of total energy consumption, heat energy cascade utilization and product quality in the existing oil refining process, which leads to high energy consumption or difficulty in effectively controlling quality fluctuations.

Description

Technical Field

[0001] This invention relates to the field of process control technology, specifically to an intelligent control method and system for oil processing parameters for energy consumption optimization. Background Technology

[0002] Oil refining is a crucial step in vegetable oil processing, mainly involving degumming, deacidification, decolorization, and deodorization. Each process requires operation under specific temperature conditions, and steam and electricity consumption account for a significant proportion of refining costs. In current oil refining production processes, energy management and quality control are often considered as two relatively independent dimensions. There is a lack of systematic planning for the utilization of heat energy between processes, and a large amount of waste heat emitted from upstream processes is directly lost without effective recovery, resulting in significant energy waste.

[0003] Meanwhile, crude oil quality exhibits significant batch-to-batch fluctuations, with inherent interrelationships among the three key quality indicators: acid value, peroxide value, and trans fatty acid content. Increasing deodorization temperature to lower acid value often leads to an increase in trans fatty acid content, while lowering deodorization temperature to control trans fatty acid formation may result in acid value failing to meet finished product standards. This trade-off creates a dilemma in setting process parameters. Existing technologies lack intelligent control methods that can simultaneously optimize energy consumption and ensure quality. Adjustments to process parameters often rely on operator experience and judgment, making it difficult to minimize energy consumption or achieve optimal balance when quality indicators conflict. Summary of the Invention

[0004] This invention provides an intelligent control method and system for oil processing parameters for energy consumption optimization. This invention solves the technical problem that existing oil refining processes struggle to achieve unified optimization of total energy consumption, cascaded utilization of heat energy, and finished product quality, leading to high energy consumption or difficulty in effectively controlling quality fluctuations.

[0005] In view of the above problems, the present invention provides an intelligent control method for oil processing parameters for energy consumption optimization, comprising:

[0006] Real-time process parameters and raw material quality indicators for each step in the oil refining process are collected. The real-time process parameters include at least the temperature of each step, the inlet and outlet temperatures of the heat exchanger, and the amount of direct steam used. The raw material quality indicators include at least the acid value, peroxide value, and trans fatty acid content of crude oil.

[0007] The real-time efficiency of the heat exchanger is calculated based on the inlet and outlet temperatures, and the temperature gradient matching coefficient between processes is calculated based on the temperature of each process and the real-time efficiency of the heat exchanger.

[0008] The quality deviation is determined based on the raw material quality indicators, and the quality characteristic conflict degree under the current process parameters is calculated. The quality characteristic conflict degree characterizes the degree of mutual constraint between crude oil acid value, peroxide value and trans fatty acid content.

[0009] With the goals of minimizing total energy consumption, maximizing temperature gradient matching coefficient, and minimizing quality characteristic conflict, an improved optimization algorithm is used to solve for the optimal combination of process parameters, wherein the weight allocation of each objective is dynamically adjusted according to the quality deviation.

[0010] Adjust the execution parameters of each process according to the optimal combination of process parameters.

[0011] This invention also provides an intelligent control system for oil processing parameters for energy consumption optimization, comprising:

[0012] The data acquisition module is used to collect real-time process parameters and raw material quality indicators for each step in the oil refining process. The real-time process parameters include at least the temperature of each step, the inlet and outlet temperatures of the heat exchanger, and the amount of direct steam used. The raw material quality indicators include at least the acid value of crude oil, the peroxide value, and the trans fatty acid content.

[0013] The efficiency calculation and cascade matching module is used to calculate the real-time efficiency of the heat exchanger based on the inlet and outlet temperatures of the heat exchanger, and to calculate the temperature cascade matching coefficient between processes based on the temperature of each process and the real-time efficiency of the heat exchanger.

[0014] The quality evaluation module is used to determine the quality deviation based on the raw material quality indicators and to calculate the quality characteristic conflict degree under the current process parameters. The quality characteristic conflict degree characterizes the degree of mutual constraint between crude oil acid value, peroxide value and trans fatty acid content.

[0015] The multi-objective optimization module is used to solve for the optimal combination of process parameters with multiple objectives, including minimizing total energy consumption, maximizing temperature gradient matching coefficient, and minimizing quality characteristic conflict degree. The weight allocation of each objective is dynamically adjusted according to the quality deviation.

[0016] The execution control module is used to adjust the execution parameters of each process according to the optimal combination of process parameters.

[0017] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0018] This invention achieves precise quantification and graded recovery of process waste heat by calculating heat exchanger efficiency and temperature gradient matching coefficients in real time during the refining process, significantly reducing the amount of external heat source replenishment. By constructing quality deviation and quality feature conflict indices, the optimization process can simultaneously consider the inherent constraints between raw material batch fluctuations and the three quality indicators. By using quality deviation as the basis for weight allocation, the algorithm automatically adjusts the optimization focus under different raw material quality conditions, ensuring that the finished product quality remains under control. By introducing a quality-oriented population reinitialization mechanism in the particle swarm optimization process, the algorithm effectively avoids getting trapped in local optima, ensuring the global optimality of the solution. These methods together achieve the technical effect of synergistic optimization of energy consumption, heat energy utilization, and quality assurance in the oil refining process, effectively reducing total energy consumption while ensuring that all quality indicators of the finished oil meet the standards.

[0019] In summary, this invention solves the technical problem that it is difficult to achieve unified optimization of total energy consumption, cascade utilization of heat energy and finished product quality in existing oil refining processes, resulting in high energy consumption or difficulty in effectively controlling quality fluctuations. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the intelligent control method for oil processing parameters for energy consumption optimization provided in an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of the three-level deviation response mechanism in the intelligent control method for oil processing parameters for energy consumption optimization provided in this embodiment of the invention;

[0023] Figure 3 This is a schematic diagram of the intelligent control system for oil processing parameters for energy consumption optimization provided in an embodiment of the present invention.

[0024] The components represented by each number in the attached diagram are explained below:

[0025] Data acquisition module 11, efficiency calculation and tiered matching module 12, quality evaluation module 13, multi-objective optimization module 14, execution control module 15. Detailed Implementation

[0026] This invention provides an intelligent control method and system for oil processing parameters for energy consumption optimization, which specifically solves the technical problem that it is difficult to achieve unified optimization of total energy consumption, cascade utilization of heat energy and finished product quality in existing oil refining processes, resulting in high energy consumption or difficulty in effectively controlling quality fluctuations.

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0028] Example 1, as Figure 1 As shown, this invention provides an intelligent control method for oil processing parameters for energy consumption optimization, comprising:

[0029] S100: Collect real-time process parameters and raw material quality indicators for each step in the oil refining process. The real-time process parameters include at least the temperature of each step, the inlet and outlet temperatures of the heat exchanger, and the amount of direct steam used. The raw material quality indicators include at least the acid value of crude oil, the peroxide value, and the trans fatty acid content.

[0030] Currently, the oil processing industry generally suffers from fragmented data and inconsistent metering in energy consumption collection and management. Steam consumption, electricity consumption, and cooling energy consumption for each process are often recorded in separate ledgers, lacking a unified energy consumption accounting standard and standardized conversion methods. This makes it impossible to accurately grasp the true energy consumption level of the entire refining process. There is a lack of comparability between different energy sources; steam energy consumption is measured in kilograms or tons, electricity in kilowatt-hours, and cooling energy consumption in kilojoules or tons of water, making comprehensive analysis and horizontal comparison of various energy consumption data difficult. Furthermore, traditional energy consumption data collection relies on manual meter reading and periodic statistics, resulting in poor timeliness and failing to support the needs of real-time energy consumption optimization decisions. Therefore, establishing a standardized, real-time, and unified multi-energy integrated energy consumption calculation method has become an urgent technical problem to be solved.

[0031] Step S100 in the method provided in this embodiment of the invention includes:

[0032] The deodorization temperature, decolorization temperature, deacidification temperature, and degumming temperature are collected by thermocouple temperature sensors installed in each process.

[0033] The inlet and outlet temperatures of the heat exchanger are collected by resistance temperature sensors installed at the inlet and outlet of the hot side and the inlet and outlet of the cold side of the heat exchanger.

[0034] The direct steam consumption is collected by a vortex flow meter installed on the steam pipeline;

[0035] The acid value, peroxide value, and trans fatty acid content of crude oil were collected using an online near-infrared spectrometer installed on the crude oil feed pipeline.

[0036] The steam energy consumption of the deodorization process is calculated based on the direct steam consumption, steam pressure, and steam enthalpy of the deodorization process. The power energy consumption of each process is calculated based on the heating power and processing time of each process.

[0037] Calculate the cooling energy consumption based on the cooling water flow rate, inlet and outlet temperature difference, and specific heat capacity of the cooling system.

[0038] The total energy consumption is obtained by weighting and summing the steam energy consumption, electricity energy consumption, and cooling energy consumption. The weighting coefficients for the weighted summation are determined based on the standard coal conversion coefficients for various energy sources.

[0039] In this embodiment of the invention, during the oil refining process, the temperature of each step is collected using thermocouple temperature sensors installed at each step. One deodorization temperature sensor is installed in the middle and one in the lower part of the tower. This is because there are temperature stratifications at different heights within the deodorization tower; the middle is the main reaction zone, and the lower part is the oil outlet zone. The temperatures at both locations jointly determine the degree of deodorization reaction and the degree of thermal degradation of the oil. A single measuring point cannot represent the overall temperature situation of the entire tower. The sensor insertion depth is set to 300 to 400 mm, which is sufficient for the temperature probe to penetrate the pipe wall insulation layer and the boundary stagnant layer to enter the main fluid region, measuring the true temperature of the oil rather than the pipe wall temperature. The decolorization temperature sensor is installed on the side wall of the reaction tank and extends into the tank at half its diameter. This location is the area within the tank where mixing is most complete and the temperature distribution is most uniform, avoiding interference from the cold wall effect near the wall.

[0040] Both the deacidification temperature and degumming temperature sensors are installed on the outlet pipe because the heat input for these two processes is mainly used to reach the required reaction temperature. The outlet temperature represents the true temperature of the oil after the reaction, indicating the final heat treatment effect of the process. All sensors are kept away from frequency converters and large motors of 50 kW or more, because the high-frequency harmonics of the frequency converter and the strong magnetic field of the motor can introduce broadband noise beyond 50 Hz power frequency interference through signal line coupling, causing measurement fluctuations of more than ±2 degrees Celsius. The connection between the sensors and the signal processing equipment uses shielded cables, with the shielding layer grounded at a single point at the sensor end to suppress electromagnetic interference.

[0041] Furthermore, the inlet and outlet temperatures of the heat exchanger are acquired using resistance temperature sensors (RTS) installed at the hot and cold side inlet and outlet. The hot-side inlet temperature sensor is installed on a straight pipe section at least five times the pipe diameter upstream of the hot-side inlet flange. The hot-side outlet temperature sensor is installed on a straight pipe section at least five times the pipe diameter downstream of the hot-side outlet flange. The cold-side inlet and outlet temperature sensors are installed at their respective locations on the cold-side pipe according to the same principle. After installation, all temperature sensors undergo individual commissioning and calibration. Calibration uses a standard temperature source, and comparisons are made at three temperature points within the measurement range: low, medium, and high. Sensors with deviations exceeding the allowable range are replaced or have their coefficients corrected. Sensors can only be put into use after joint commissioning confirms accurate measurements.

[0042] Furthermore, the direct steam consumption is collected using a vortex flow meter installed on the steam pipeline. The vortex flow meter is installed on the steam inlet pipeline of the deodorization process, with an upstream straight pipe section length of no less than 15 times the pipe diameter and a downstream straight pipe section length of no less than 5 times the pipe diameter, to ensure that the fluid has formed a fully developed turbulent velocity distribution after passing through the flow meter, thus maintaining a constant flow coefficient. The vortex flow meter is installed horizontally on the pipeline, with its axis concentric with the pipeline axis and the flow direction consistent with the fluid direction. No regulating valves or other components that may cause fluid disturbance are installed upstream of the flow meter. If this cannot be avoided due to pipeline layout, the regulating valve is installed downstream of the flow meter. During installation, anti-vibration pads are added at the connection between the flow meter and the pipeline to reduce the impact of pipeline vibration on measurement accuracy. The anti-vibration pads are made of high-temperature resistant rubber gaskets with a thickness of 5 to 10 mm.

[0043] Furthermore, the acid value, peroxide value, and trans fatty acid content of crude oil are collected using an online near-infrared spectrometer installed on the crude oil feed pipeline. The near-infrared analyzer is installed on the pipeline at the outlet of the crude oil feed pump. A dedicated flow cell accessory ensures stable flow of crude oil within the flow cell. The optical path length of the flow cell is set according to the oil's absorption intensity, typically between 1 and 5 millimeters. The main probe of the near-infrared analyzer uses a fixed grating type or Fourier transform type near-infrared spectrometer, performing a rapid scan every 5 seconds. Before use, a calibration model for the current oil type must be established. The collected near-infrared spectral signals are correlated and calibrated with the acid value, peroxide value, and trans fatty acid content determined by standard laboratory methods. The correlation coefficient of the calibration model should be above 0.95 to ensure good consistency between online measurement results and laboratory test results. The calibration model is updated and maintained every three months or when there is a significant change in the type of feedstock oil.

[0044] Furthermore, the steam energy consumption of the deodorization process is calculated based on the direct steam consumption, steam pressure, and steam enthalpy. First, the direct steam mass flow rate per unit time is obtained using a vortex flow meter, in kilograms per hour. Then, the steam pressure is read from the pressure gauge installed on the steam pipeline, and the steam enthalpy at that pressure is determined by combining it with the steam temperature. The enthalpy is obtained by referring to a table of steam thermodynamic properties, in kilojoules per kilogram. Multiplying the steam mass flow rate by the steam enthalpy yields the steam thermal power per unit time, which is then multiplied by the process time to obtain the steam energy consumption for that batch of deodorization, in kilojoules. Considering condensate and heat loss during steam transportation and use, a correction factor is applied in the actual calculation. This correction factor ranges from 0.92 to 0.95 and is periodically calibrated based on the pipeline insulation condition and the operating status of the steam traps.

[0045] Meanwhile, the power consumption of each process is calculated based on the heating power and processing time of each process. The heating power is read by an electric meter installed on the power supply circuit of the heating equipment. The electric meter is a three-phase multi-function energy meter that can simultaneously collect voltage, current, power factor, and active power, in kilowatts. Multiplying the heating power by the process processing time yields the power consumption of that process, in kilowatt-hours.

[0046] Furthermore, cooling energy consumption is calculated based on the cooling water flow rate, inlet and outlet temperature difference, and specific heat capacity of the cooling system. The cooling water flow rate is obtained using an electromagnetic flowmeter installed on the cooling water pipes, measured in kilograms per hour, with an accuracy class of ±0.5%. The cooling water inlet and outlet temperatures are obtained using resistance temperature sensors installed on the inlet and outlet pipes of the cooler, respectively. Multiplying the cooling water flow rate by the inlet and outlet temperature difference, and then by the specific heat capacity of water (taken as 4.18 kJ / kg / degreeC), yields the cooling heat power per unit time. Multiplying this by the cooling system operating time gives the cooling energy consumption, measured in kilojoules.

[0047] The total energy consumption is obtained by weighted summing the calculated steam energy consumption, electricity energy consumption, and cooling energy consumption. The weighting coefficients for the weighted summation are determined based on the standard coal conversion factors for each type of energy. The standard coal conversion factor for electricity energy consumption uses the equivalent value, i.e., 0.1229 kg of standard coal per kilowatt-hour of electricity. The standard coal conversion factor for steam energy consumption is determined based on the steam pressure level. If cooling energy consumption is included in the form of energy consumption from electrically driven cooling equipment, it is processed according to the electricity energy consumption conversion factor; if it is included in the form of cooling water itself as the energy-consuming medium, it is processed according to the corresponding energy-consuming medium conversion standard. The energy consumption of each type of energy is first converted to kilograms of standard coal, and then summed to obtain the total energy consumption measured in kilograms of standard coal.

[0048] This invention addresses several issues in existing oil refining processes. These include: installing thermocouple and resistance temperature sensors at each stage to collect temperature data; installing vortex flow meters on steam pipelines to collect direct steam consumption data; and installing an online near-infrared spectroscopy analyzer on crude oil feed pipelines to collect raw material quality indicators. The invention also addresses the problems of unscientific energy consumption data collection points, insufficient accuracy in temperature measurement due to wall temperature and electromagnetic interference, deviations in steam flow measurement due to insufficient straight pipe sections and pipe vibration, lag in raw material quality testing relying on manual sampling, and the inability to comprehensively optimize steam, electricity, and cooling energy consumption as independent indicators under the same dimension. This allows a multi-objective optimization algorithm to perform global optimization based on accurate and unified energy consumption data, providing reliable data support for intelligent control of process parameters.

[0049] S200: Calculate the real-time efficiency of the heat exchanger based on the inlet and outlet temperatures, and calculate the temperature gradient matching coefficient between processes based on the temperatures of each process and the real-time efficiency of the heat exchanger.

[0050] In the oil refining process, a large amount of waste heat resources exist between various processes. However, as heat exchangers operate for longer periods, scale and polymers gradually accumulate on the tube walls, leading to a continuous decline in heat transfer efficiency. This results in upstream waste heat not being effectively transferred to downstream heat-consuming processes. Current technologies lack real-time quantitative assessment methods for heat exchanger performance. The decrease in heat transfer efficiency caused by scaling is often only discovered during equipment maintenance, and fixed efficiency alarm thresholds cannot adapt to the differentiated needs of different heat exchanger types and different operating stages. Furthermore, the potential for waste heat recovery between processes lacks unified quantitative evaluation indicators, making it impossible to accurately determine the actual degree to which upstream waste heat meets downstream heating needs. This makes it difficult to incorporate the level of cascaded thermal energy utilization into the objective function of process parameter optimization, and upstream waste heat cannot be rationally allocated at the system level. Therefore, establishing a real-time quantitative assessment method for heat exchanger performance and quantitative evaluation indicators for cascaded thermal energy utilization between processes has become an urgent technical problem to be solved.

[0051] Step S200 in the method provided in this embodiment of the invention includes:

[0052] Obtain the hot-side inlet temperature, hot-side outlet temperature, cold-side inlet temperature, and cold-side outlet temperature of the heat exchanger, as well as the hot-side flow rate and the specific heat capacity of the hot-side oil.

[0053] The actual heat transfer of the heat exchanger is calculated based on the temperature difference between the hot side inlet temperature and the hot side outlet temperature, the hot side flow rate, and the specific heat capacity of the hot side oil.

[0054] The theoretical maximum heat transfer of the heat exchanger is calculated based on the temperature difference between the hot side inlet temperature and the cold side inlet temperature, the hot side flow rate, and the specific heat capacity of the hot side oil.

[0055] The real-time efficiency of the heat exchanger is determined based on the degree to which the actual heat transfer of the heat exchanger is maintained relative to the theoretical maximum heat transfer of the heat exchanger.

[0056] Specifically, when the real-time efficiency of the heat exchanger is lower than the preset efficiency threshold, a heat exchanger cleaning reminder signal is generated. The preset efficiency threshold is adaptively set according to the heat exchanger type and running time.

[0057] In this embodiment of the invention, the inlet temperature, outlet temperature, inlet temperature, and outlet temperature of the heat exchanger on the hot side, as well as the mass flow rate of the hot-side fluid and the specific heat capacity of the hot-side oil, are obtained. The mass flow rate of the hot-side fluid is obtained by a mass flow meter installed on the hot-side pipeline of the heat exchanger, and the unit is kilograms per second. The specific heat capacity of the hot-side oil is determined by referring to an oil property handbook based on the type of oil and temperature.

[0058] Furthermore, based on the temperature difference between the hot-side inlet and outlet temperatures, the hot-side mass flow rate, and the specific heat capacity of the hot-side oil, the actual heat transfer of the heat exchanger is calculated. Actual heat transfer = Hot-side mass flow rate × Hot-side oil specific heat capacity × (Hot-side inlet temperature - Hot-side outlet temperature). The actual heat transfer reflects the amount of heat transferred from the hot-side fluid to the cold-side fluid per unit time under actual operating conditions.

[0059] Furthermore, based on the temperature difference between the hot-side and cold-side inlet temperatures, the hot-side mass flow rate, and the specific heat capacity of the hot-side oil, the theoretical maximum heat transfer of the heat exchanger is calculated. Theoretical maximum heat transfer = hot-side mass flow rate × hot-side oil specific heat capacity × (hot-side inlet temperature - cold-side inlet temperature). The theoretical maximum heat transfer reflects the upper limit of the maximum heat that the heat exchanger can transfer under ideal conditions.

[0060] Furthermore, the real-time efficiency of the heat exchanger is determined based on the ratio of the actual heat transfer to the theoretical maximum heat transfer. Real-time efficiency = Actual heat transfer / Theoretical maximum heat transfer. A higher efficiency value indicates better heat transfer performance and more complete heat recovery.

[0061] Furthermore, when the real-time efficiency of the heat exchanger falls below a preset efficiency threshold, a heat exchanger cleaning reminder signal is generated. The preset efficiency threshold is adaptively set based on the heat exchanger type and operating time. The determination of the preset efficiency threshold follows this principle: During long-term operation, as oil, polymers, and coke gradually deposit on the heat exchanger tube walls, the thermal resistance continuously increases, and the heat transfer coefficient gradually declines, resulting in a year-on-year decrease in the heat exchanger's real-time efficiency. This decline process can be approximated using an exponential or linear decay model. This embodiment uses a linear decay model to determine the preset efficiency threshold, and the calculation formula is: Preset efficiency threshold = Initial efficiency × (1 − Annual decay rate × Number of operating years)

[0062] The initial efficiency is the real-time efficiency value of the heat exchanger measured during its first operation after commissioning or cleaning. This value represents the optimal and relatively stable performance during the initial operation phase. It is calculated by averaging the results of 72 hours of continuous operation with measurements taken every two hours. The annual degradation rate is the percentage decrease in heat exchanger efficiency due to fouling accumulation each year. It is determined based on the heat exchanger type and media characteristics, obtained through regression fitting of historical efficiency degradation data from similar heat exchangers. For plate heat exchangers, with their narrow flow channels, high oil flow velocity, and high shear force, fouling is less likely to adhere, resulting in an annual degradation rate ranging from 3% to 5%. When processing oils with high acid values ​​or high gum content, the annual degradation rate tends towards the upper limit of this range. For shell-and-tube heat exchangers, with relatively wide tube and shell flow channels, lower oil flow velocity within the tubes, thicker boundary layers, and easier fouling deposition on the tube walls, the annual degradation rate ranges from 6% to 10%. The number of years of operation is the actual operating time of the heat exchanger since it was put into operation or since the last cleaning was completed, converted into a value in years. Operating time = cumulative actual operating hours of the heat exchanger / annual design operating hours.

[0063] When the real-time efficiency of the heat exchanger falls below the preset efficiency threshold calculated using the above formula, it indicates that the decrease in heat transfer coefficient due to oil deposits on the heat exchanger tube walls has exceeded the allowable range. An automatic cleaning reminder signal is then generated, notifying maintenance personnel to arrange cleaning operations. This adaptive setting method allows the efficiency threshold to gradually decrease over time, avoiding the delayed alarm problem caused by excessively low thresholds during the initial commissioning phase leading to severe fouling accumulation, and also avoiding frequent false alarms caused by excessively high thresholds during long-term operation. This ensures accurate judgment of the cleaning reminder timing.

[0064] Step S200 in the method provided in this embodiment of the invention further includes:

[0065] The outlet temperature of the upstream process is extracted as the waste heat source temperature, and the inlet demand temperature of the downstream process is extracted as the heat consumption temperature.

[0066] Calculate the recoverable heat from the upstream process based on the waste heat source temperature, the oil flow rate from the upstream process, and the specific heat capacity of the oil.

[0067] Calculate the heat demand of the downstream process based on the heat application temperature, the oil flow rate of the downstream process, and the specific heat capacity of the oil.

[0068] Calculate the actual heat transfer capacity of the heat exchanger based on its real-time efficiency, recoverable heat, and required heat.

[0069] The temperature gradient matching coefficient is determined based on the degree to which the actual transferable heat meets the required heat.

[0070] In this embodiment of the invention, the outlet temperature of the upstream process is extracted as the waste heat source temperature, and the inlet demand temperature of the downstream process is extracted as the heat consumption temperature. The waste heat source temperature refers to the actual temperature of the oil discharged from the upstream process at the outlet, representing the energy grade suitable for supplying heat downstream. The heat consumption temperature refers to the inlet temperature required by the downstream process to meet its process requirements. These temperature values ​​are acquired in real time by temperature sensors installed on the outlet pipes of the upstream process and the inlet pipes of the downstream process.

[0071] Furthermore, based on the waste heat source temperature, the oil flow rate of the upstream process, and the specific heat capacity of the oil, the recoverable heat of the upstream process is calculated. Recoverable heat = upstream process oil mass flow rate × upstream process oil specific heat capacity × (upstream process outlet temperature - reference ambient temperature). The reference ambient temperature is taken as 25 degrees Celsius. Recoverable heat represents the total heat that can be released when the oil outlet temperature of the upstream process drops from the current temperature to the reference ambient temperature, reflecting the total amount of waste heat resources in the upstream process.

[0072] Furthermore, based on the heating temperature, the oil flow rate of the downstream process, and the specific heat capacity of the oil, the heat demand of the downstream process is calculated. Heat demand = Oil mass flow rate of the downstream process × Specific heat capacity of the oil in the downstream process × (Required inlet temperature of the downstream process - Reference ambient temperature). The heat demand represents the total heat required to heat the oil at the inlet of the downstream process from the reference ambient temperature to the required temperature.

[0073] Furthermore, based on the heat exchanger's real-time efficiency, recoverable heat, and required heat, the actual transferable heat of the heat exchanger is calculated. Actual transferable heat = real-time efficiency of heat exchanger × recoverable heat. Actual transferable heat represents the amount of heat that can actually be transferred between the upstream waste heat source and the downstream heat-consuming process through the heat exchanger. It is simultaneously limited by the total amount of recoverable heat upstream and the heat exchanger's own heat transfer efficiency.

[0074] Furthermore, the temperature gradient matching coefficient is determined based on the degree to which the actual transferable heat meets the demand for heat. Temperature gradient matching coefficient = Actual transferable heat / Demand for heat. When the temperature gradient matching coefficient is greater than or equal to 1, it indicates that the upstream recoverable heat, after being transferred through the heat exchanger, is sufficient to meet the entire heating demand of the downstream process, and the temperature gradient matching coefficient is set to 1. When the temperature gradient matching coefficient is less than 1, it indicates that the upstream recoverable heat is insufficient to meet the entire heating demand of the downstream process, and the shortfall needs to be supplemented by an external heat source; the temperature gradient matching coefficient is taken as its actual calculated value. The closer the temperature gradient matching coefficient is to 1, the more fully the heat energy is utilized in the gradient between processes, and the smaller the need for external heat source supplementation. When the upstream waste heat source temperature is lower than the downstream heating temperature, it is determined that the temperature gradient is reversed, and the temperature gradient matching coefficient is set to 0.

[0075] This invention achieves real-time quantitative evaluation of heat exchanger performance by using the ratio of actual heat transfer to theoretical maximum heat transfer as real-time efficiency, enabling operators to monitor the operating status of each heat exchanger at any time. By adaptively setting efficiency thresholds using a linear decay model based on heat exchanger type and operating time, the efficiency threshold for any operating time point can be dynamically calculated using two parameters: initial efficiency and annual decay rate. This overcomes the problem of fixed thresholds being unsuitable for different equipment characteristics and operating stages. Newly commissioned heat exchangers have higher thresholds to ensure efficient operation, while long-term operating heat exchangers have gradually lower thresholds to avoid frequent false alarms. Furthermore, the use of annual decay rate values ​​based on historical statistical data provides clear data support for threshold setting rather than subjective determination. By automatically generating cleaning reminder signals when efficiency falls below the threshold, on-demand heat exchanger maintenance is achieved instead of periodic, indiscriminate cleaning, reducing unnecessary downtime and maintenance costs.

[0076] Meanwhile, by calculating the temperature gradient matching coefficient, the upstream recoverable heat, downstream demanded heat, and real-time efficiency of the heat exchanger are uniformly incorporated into a quantitative index, which accurately reflects the actual feasibility and satisfaction of waste heat recovery between processes. This provides a clear and calculable basis for maximizing the thermal energy gradient utilization objective in the multi-objective optimization algorithm, enabling the optimization algorithm to coordinate the heat supply and demand matching of each process at the global level and minimize the supplementary consumption of external heat sources.

[0077] S300: Determine the quality deviation based on the raw material quality indicators and calculate the quality characteristic conflict degree under the current process parameters. The quality characteristic conflict degree characterizes the degree of mutual constraint between crude oil acid value, peroxide value and trans fatty acid content.

[0078] In existing oil refining process parameters, the acid value, peroxide value, and trans fatty acid content of different batches of crude oil exhibit significant fluctuations. However, the lack of quantitative evaluation methods for the degree of raw material quality fluctuation leads to the inability to differentiate process parameters based on the actual quality of the raw materials. High-quality and low-quality raw materials often use the same process parameters, resulting in energy waste or substandard quality. Furthermore, adjustments to process parameters such as deodorization temperature do not have entirely consistent effects on acid value, peroxide value, and trans fatty acid content. Measures to reduce acid value may lead to an increase in trans fatty acid content, while measures to control peroxide value may adversely affect acid value. There is an inherent inverse relationship between these three quality indicators, but existing technologies lack a unified quantitative characterization method for this mutual constraint, making process parameter optimization lack a clear directional basis. Therefore, establishing a quantitative evaluation system for raw material quality fluctuations and a quantitative characterization method for the degree of mutual constraint among quality indicators has become an urgent technical problem to be solved.

[0079] Step S300 in the method provided in this embodiment of the invention includes:

[0080] The benchmark acid value, benchmark peroxide value, and benchmark trans fatty acid content of the same raw material type are extracted from the historical raw material database. The benchmark index is the median of the corresponding index of the historical raw material.

[0081] Calculate the deviation of the current crude oil acid value from the benchmark acid value, the deviation of the current crude oil peroxide value from the benchmark peroxide value, and the deviation of the current crude oil trans fatty acid content from the benchmark trans fatty acid content, and use these as the acid value deviation, peroxide value deviation, and trans fatty acid deviation, respectively.

[0082] In this embodiment of the invention, the benchmark acid value, benchmark peroxide value, and benchmark trans fatty acid content for the same raw material type are first extracted from the historical raw material database. The historical raw material database stores the test data of all batches of crude oil from the oil processing enterprise over the past three years, categorized by raw material type such as soybean crude oil, rapeseed crude oil, and palm crude oil. The benchmark index is the median of the corresponding index in the historical raw materials. The median, rather than the average value, is chosen as the benchmark because the median is not sensitive to extreme values ​​and can more robustly reflect the typical level of raw material quality, avoiding inaccuracies in the benchmark value due to abnormalities in the quality of individual batches of raw materials.

[0083] Further, calculate the deviation between the current crude oil acid price and the benchmark acid price. Acid price deviation = / Base acid value. Similarly, peroxide value deviation = / Baseline peroxide value. Trans fatty acid deviation = / Benchmark trans fatty acid content. The three deviations reflect the degree of fluctuation of each quality indicator of the current batch of crude oil relative to historical normal levels. The larger the deviation, the more abnormal the quality indicator is, and the higher its weight should be given in subsequent optimization.

[0084] Step S300 in the method provided in this embodiment of the invention further includes:

[0085] Obtain the preset values ​​of process parameters for each process step. The preset values ​​of process parameters include at least the deodorization temperature, deodorization time, decolorization temperature, and decolorization time.

[0086] Based on the preset values ​​of process parameters and raw material quality indicators, the predicted oleic acid value, predicted peroxide value and predicted trans fatty acid content after the deodorization process are predicted by a pre-trained quality prediction model.

[0087] The quality standard thresholds for refined oil products are obtained, including the upper limit of acid value, the upper limit of peroxide value, and the upper limit of trans fatty acid content.

[0088] The degree of acid price compliance is determined by the relative deviation between the predicted oleic acid price and the upper limit of the acid price.

[0089] The peroxide value compliance rate is determined based on the relative deviation between the predicted peroxide value and the upper limit of the peroxide value.

[0090] The trans fatty acid compliance rate is determined based on the relative deviation between the predicted trans fatty acid content and the upper limit of trans fatty acid content.

[0091] The deviation of quality characteristics is determined comprehensively based on the degree of deviation of acid value compliance, peroxide value compliance, and trans fatty acid compliance from the standard values.

[0092] The standard deviations among acid value compliance, peroxide value compliance, and trans fatty acid compliance are calculated as the dispersion of quality characteristics.

[0093] The quality feature conflict degree is obtained by weighted summing of the quality feature deviation and the quality feature dispersion.

[0094] In this embodiment of the invention, preset values ​​of process parameters for each process step are first obtained. These preset values ​​include at least deodorization temperature, deodorization time, decolorization temperature, and decolorization time. These preset values ​​of process parameters are derived from candidate solutions corresponding to the current iteration of the optimization algorithm, or from initial values ​​set empirically at the beginning of the optimization algorithm.

[0095] Furthermore, based on the preset values ​​of process parameters and raw material quality indicators, the predicted oleic acid value, predicted peroxide value, and predicted trans fatty acid content after the deodorization process are predicted using a pre-trained quality prediction model.

[0096] The quality prediction model is established by collecting historical production data, including the initial acid value, initial peroxide value, initial trans fatty acid content, and corresponding deodorization temperature, deodorization time, decolorization temperature, and decolorization time process parameters of each batch of crude oil, as well as the actual acid value, actual peroxide value, and actual trans fatty acid content of the corresponding batch of finished oil, with a sample size of no less than 500 groups.

[0097] Secondly, outlier removal is required for the collected raw data. Using the Laida criterion, the mean and standard deviation of each variable are calculated, and sample points deviating from the mean by more than three times the standard deviation are removed. Since the seven input variables have different dimensions and numerical ranges, all input and output variables need to be normalized, mapping the numerical range of each variable to the interval between 0 and 1. The normalized value is calculated as (original value - minimum value) / (maximum value - minimum value).

[0098] Then, a three-layer feedforward BP neural network is used, consisting of an input layer, a hidden layer, and an output layer. The seven neurons in the input layer correspond to the seven input variables, and the three neurons in the output layer correspond to the three output variables. The number of neurons in the hidden layer is determined by trial and error, and the value with the smallest validation set error is selected. The activation function of the hidden layer is a sigmoid function, and the activation function of the output layer is a linear function.

[0099] During model training, all sample data are randomly divided into training and validation sets. The training set accounts for 70-80% of the data and is used to learn the model parameters, while the validation set accounts for 20-30% and is used to evaluate the model's generalization ability. During training, after each training cycle, the current model is used to make a prediction on the validation set, and the validation error is calculated. When the validation error continues to decrease, it indicates that the network is still learning effective features. When the validation error stops decreasing or even starts to increase, it indicates that the network is overfitting, and training should be stopped immediately. After training, the coefficients of determination (CODs) between the predicted and actual values ​​of acid value, peroxide value, and trans fatty acid content on the validation set are calculated. The model can only be used in practice when the CODs for all three indicators reach 0.95 or higher.

[0100] Finally, after the model is put into use, every three months or whenever more than fifty sets of valid production data are added, the new data will be included in the historical database, and the entire process of data preprocessing, normalization, model training and validation will be re-executed to update the model parameters. When the type of raw oil product changes, samples need to be collected again for the new oil product and the model needs to be re-modeled.

[0101] It should be noted that division by zero may occur during the normalization process. In this case, the normalized value of all samples on this variable is uniformly set to a preset fixed value of 0.5, because 0.5 is located in the middle of the 0 to 1 interval after normalization, which will not introduce biased information and will not affect the normalization results of other variables.

[0102] Meanwhile, after model deployment, the value of a variable in new measured data may indeed be lower than the minimum value or higher than the maximum value in the training set. Without processing, the normalized value will be less than 0 or greater than 1, exceeding the effective input range of the neural network activation function, potentially leading to prediction distortion. Therefore, in practical applications, boundary truncation processing is required for normalization calculations: when the original value of the new data is less than the minimum value in the training set, the normalized value is forced to 0; when the original value of the new data is greater than the maximum value in the training set, the normalized value is forced to 1, ensuring that all input values ​​remain within the model's training distribution range and preventing out-of-range inputs from causing prediction failure.

[0103] Furthermore, quality standard thresholds for refined oil products are obtained, including upper limits for acid value, peroxide value, and trans fatty acid content. These thresholds are set based on national food safety standards or enterprise internal control standards. For example, according to GB 5009.229-2016 "National Food Safety Standard - Determination of Acid Value in Food" and GB 5009.227-2016 "National Food Safety Standard - Determination of Peroxide Value in Food," the upper limit for acid value of Grade 1 soybean oil is 0.2 mg potassium hydroxide per gram, and the upper limit for peroxide value is 5.0 mmol per kilogram. The upper limit for trans fatty acid content is set according to GB 5009.257-2016 "National Food Safety Standard - Determination of Trans Fatty Acids in Food" and enterprise internal control standards, and is generally set to 2.0% to 3.0%.

[0104] Furthermore, the acid value compliance rate is determined based on the relative deviation between the predicted oleic acid value and the upper limit of the acid value. The acid value compliance rate equals 1 - predicted acid value / upper limit of acid value. When the predicted acid value is higher than the upper limit of the acid value, the compliance rate is 0. The closer the compliance rate is to 1, the better the quality indicator is compared to the upper limit of the standard, and the better the compliance situation. The closer the compliance rate is to 0, the closer the quality indicator is to or exceeds the upper limit of the standard, and the greater the quality risk. Similarly, the peroxide value compliance rate is determined based on the relative deviation between the predicted peroxide value and the upper limit of the peroxide value, and the trans fatty acid compliance rate is determined based on the relative deviation between the predicted trans fatty acid content and the upper limit of the trans fatty acid content. The three compliance rates reflect the margin of the three quality indicators relative to the standard threshold under given process parameters.

[0105] Furthermore, based on the deviations of the acid value, peroxide value, and trans fatty acid compliance rates from the standard values, the deviation of the quality characteristics is comprehensively determined. The standard value is set to 1, meaning that ideally, all three compliance rates should equal 1. Quality characteristic deviation = The greater the deviation of the quality characteristics, the greater the degree to which the three quality indicators deviate from the ideal standard.

[0106] Furthermore, the standard deviations among the acid value compliance, peroxide value compliance, and trans fatty acid compliance were calculated as the dispersion of quality characteristics. When the three compliance values ​​are close, the dispersion is small, indicating a relatively balanced compliance across quality indicators. When one compliance value is significantly lower than the other two, the dispersion is large, indicating a significant conflict between quality indicators. The larger the dispersion, the more severe the conflict between quality indicators. For example, increasing the deodorization temperature to lower the acid value may lead to an increase in trans fatty acid content, resulting in a significantly lower trans fatty acid compliance than the other two, thus increasing the dispersion.

[0107] Finally, the quality characteristic deviation and quality characteristic dispersion are weighted and summed to obtain the quality characteristic conflict degree. The quality characteristic conflict degree comprehensively reflects the degree to which the three quality indicators deviate from the ideal state as a whole and the degree of imbalance among them. The higher the conflict degree, the more serious the mutual constraints among the three quality indicators under the current process parameters, and the more optimization and adjustment are needed.

[0108] The weighting coefficients in the formula for calculating the conflict degree of quality characteristics are determined using principal component analysis. Specifically, at least three hundred sets of production batch data are extracted from the company's historical database. Each set includes the compliance rates of acid value, peroxide value, and trans fatty acid under specific process parameters for that batch. The deviation and dispersion of quality characteristics for each set are calculated, forming a two-dimensional dataset. Principal component analysis is then performed on this dataset. The variance contribution rate of the first principal component reflects the degree to which the deviation of quality characteristics explains the total variance of the data, while the variance contribution rate of the second principal component reflects the degree to which the dispersion of quality characteristics explains the total variance of the data. After normalizing the variance contribution rates of the two principal components, the weighting coefficients α and β are obtained, corresponding to the weights of the deviation and dispersion of quality characteristics, respectively.

[0109] S400: With multiple objectives of minimizing total energy consumption, maximizing temperature gradient matching coefficient, and minimizing quality characteristic conflict, the optimal combination of process parameters is solved through an improved optimization algorithm, in which the weight allocation of each objective is dynamically adjusted according to the quality deviation.

[0110] In multi-objective optimization, there is an inherent contradiction between the three objectives of minimizing total energy consumption, maximizing the temperature gradient matching coefficient, and minimizing the conflict degree of quality characteristics. Measures to reduce energy consumption may affect quality assurance, and process settings that enhance waste heat recovery may be inconsistent with the optimization direction of quality conflict degree. However, existing technologies lack a dynamic allocation mechanism that can adaptively adjust the weights of each objective according to the fluctuation of raw material quality, resulting in the optimization algorithm being unable to make differentiated optimization decisions for raw materials of different qualities. At the same time, particle swarm optimization algorithms are prone to getting trapped in local optima in the later stages of iteration due to particle convergence, but existing algorithms lack effective strategies to escape local optima, so the combination of process parameters obtained is often a local optimum rather than a global optimum.

[0111] Furthermore, after optimization, deviations may occur between the actual production results and the predicted values ​​due to factors such as model prediction errors, equipment execution deviations, or fluctuations in raw material quality. However, existing technologies lack a closed-loop correction mechanism based on actual detection feedback, making it impossible to automatically trigger re-optimization or safety protection upon detecting deviations. Therefore, establishing a dynamic adaptive allocation mechanism for multi-objective weights, improving the global search capability of optimization algorithms, and constructing a closed-loop control system for prediction-execution-feedback have become urgent technical problems to be solved.

[0112] Step S400 in the method provided in this embodiment of the invention includes:

[0113] Obtain the deviation of acid value, peroxide value, and trans fatty acid;

[0114] The first weighting adjustment coefficient is determined based on the acid value deviation, and the first weighting adjustment coefficient is positively correlated with the acid value deviation.

[0115] The second weighting adjustment coefficient is determined based on the peroxide value deviation, and the second weighting adjustment coefficient is positively correlated with the peroxide value deviation.

[0116] The third weighting adjustment coefficient is determined based on the deviation of trans fatty acids, and the third weighting adjustment coefficient is positively correlated with the deviation of trans fatty acids.

[0117] Obtain the basic weights for the target of minimizing total energy consumption, the basic weights for the target of maximizing temperature gradient matching coefficient, and the basic weights for the target of minimizing quality characteristic conflict.

[0118] The adjusted weights of the total energy consumption minimization target are obtained by dividing the base weights of the total energy consumption minimization target by the arithmetic mean of the first weight adjustment coefficient, the second weight adjustment coefficient, and the third weight adjustment coefficient.

[0119] Multiply the base weight of the temperature gradient matching coefficient maximization target by the first weight adjustment factor to obtain the adjusted temperature gradient matching coefficient maximization target weight;

[0120] The adjusted weights for minimizing quality feature conflict are obtained by multiplying the base weight of the quality feature conflict minimization objective by the average of the second and third weight adjustment coefficients.

[0121] In this embodiment of the invention, the deviation of acid value, deviation of peroxide value and deviation of trans fatty acid are first obtained. The calculation methods for these three deviations have been described in detail in step S300.

[0122] Furthermore, a first weighted adjustment coefficient is determined based on the acid value deviation, and this coefficient is positively correlated with the acid value deviation. The formula for calculating the first weighted adjustment coefficient is k1 = 1 + acid value deviation. When the acid value deviation is 0, k1 equals 1, indicating that no adjustment is made to this coefficient; when the acid value deviation is 0.5, k1 equals 1.5, indicating that the coefficient is increased by 50%. A linear positive correlation is used to set the adjustment coefficient, ensuring that the weight adjustment magnitude changes in the same direction as the degree of quality deviation; the more severe the deviation, the greater the adjustment.

[0123] Simultaneously, a second weighting adjustment coefficient is determined based on the peroxide value deviation, and this second weighting adjustment coefficient is positively correlated with the peroxide value deviation. The second weighting adjustment coefficient k2 = 1 + peroxide value deviation. A third weighting adjustment coefficient is determined based on the trans fatty acid deviation, and this third weighting adjustment coefficient is positively correlated with the trans fatty acid deviation. The third weighting adjustment coefficient k3 = 1 + trans fatty acid deviation.

[0124] Furthermore, the basic weights for the objectives of minimizing total energy consumption, maximizing temperature gradient matching coefficients, and minimizing quality feature conflict are obtained. The basic weights are pre-set initial weight values, and the sum of the three basic weights is 1. The basic weights are set based on the relative importance of the three objectives in multi-objective optimization.

[0125] The basic weights are determined through a combination of expert experience scoring and normalization. Multiple experts, including process engineers, production managers, and quality control personnel, were selected. Each expert independently scored the three objectives based on their importance in actual production. Minimizing total energy consumption, as the core benefit objective, received the highest score; maximizing the temperature gradient matching coefficient, as a key supporting means for energy consumption optimization, received the next highest score; and minimizing the conflict degree of quality characteristics, as a binding objective to ensure quality compliance, received the lowest score. The arithmetic mean of the scores given by each expert for the same objective was calculated to obtain the average score for each objective. The three average scores were then divided by the sum of the three scores to obtain the normalized basic weights, ensuring that the sum of the three scores equals 1. This method eliminates the subjective bias of a single decision-maker through independent scoring by multiple experts, and the normalization process gives the weights a clear probabilistic interpretation, accurately reflecting the priority hierarchy of the three objectives in actual production.

[0126] As a specific implementation method, minimizing total energy consumption is the core objective of energy conservation and emission reduction, and its basic weight is set at 0.4. Maximizing the temperature gradient matching coefficient reflects the level of thermal energy utilization in the gradient and is a key way to reduce the consumption of external heat sources, and its basic weight is set at 0.3. Minimizing the degree of conflict of quality characteristics reflects the ability to ensure the quality of finished products, and its basic weight is set at 0.3.

[0127] Furthermore, the basic weight of the total energy consumption minimization target is divided by the arithmetic mean of the first, second, and third weight adjustment coefficients to obtain the adjusted total energy consumption minimization target weight. The arithmetic mean is (k1 + k2 + k3) / 3. The adjusted total energy consumption minimization target weight = basic weight of the total energy consumption minimization target × 3 / (k1 + k2 + k3). When the deviations in quality are large, k1, k2, and k3 are all greater than 1, and the arithmetic mean is greater than 1. The weight of the total energy consumption minimization target decreases accordingly, indicating that when the quality risk is high, quality should be prioritized rather than pursuing the lowest energy consumption. The principle behind this adjustment method is that the greater the quality deviation, the worse the raw material quality or the higher the quality risk under the current process parameters, requiring the optimization focus to shift towards quality assurance.

[0128] Furthermore, the basic weight of the temperature gradient matching coefficient maximization target is multiplied by the first weight adjustment coefficient to obtain the adjusted weight of the temperature gradient matching coefficient maximization target. The weight of the temperature gradient matching coefficient maximization target = the basic weight of the temperature gradient matching coefficient maximization target × k1. The greater the deviation in acid value, the higher the weight of the temperature gradient matching coefficient. This is because reducing the acid value often requires a higher deodorization temperature, and a higher deodorization temperature means more waste heat can be recovered, thus increasing the importance of temperature gradient matching.

[0129] Simultaneously, the base weight of the quality characteristic conflict minimization target is multiplied by the average of the second and third weight adjustment coefficients to obtain the adjusted weight of the quality characteristic conflict minimization target. The weight of the quality characteristic conflict minimization target = base weight of the quality characteristic conflict minimization target × (k2 + k3) / 2. The greater the deviation of peroxide value and trans fatty acid, the higher the weight of the quality characteristic conflict minimization target. This is because peroxide value and trans fatty acids are sensitive to temperature changes, and the greater their deviation, the more necessary it is to optimize and balance the conflict between the three quality indicators.

[0130] Finally, the three adjusted weights are normalized so that their sum equals 1. Specifically, the adjusted total energy consumption minimization target weight is divided by the sum of the three adjusted weights to obtain the normalized total energy consumption minimization target weight; the adjusted temperature gradient matching coefficient maximization target weight is divided by the sum of the three adjusted weights to obtain the normalized temperature gradient matching coefficient maximization target weight; and the adjusted quality characteristic conflict degree minimization target weight is divided by the sum of the three adjusted weights to obtain the normalized quality characteristic conflict degree minimization target weight.

[0131] Step S400 in the method provided in this embodiment of the invention further includes:

[0132] Obtain the process parameter constraint ranges for each process, including the deodorization temperature range, deodorization time range, decolorization temperature range, and decolorization time range;

[0133] An initial particle swarm is generated within the constraints of process parameters. The position vector of each particle includes the deodorization temperature, deodorization time, decolorization temperature, and decolorization time.

[0134] The initial solution set generation strategy is determined based on the quality deviation.

[0135] A multi-objective fitness function is constructed based on the adjusted weights of each optimization objective. In each iteration, the fitness value of each particle is calculated, and the individual optimal position and global optimal position of each particle are updated.

[0136] Based on the velocity update formula and position update formula of the particle swarm optimization algorithm, the velocity and position of each particle are iteratively updated until the number of iterations reaches the preset upper limit or the fitness value converges.

[0137] In the iteration process, when the improvement of the global optimal fitness value is lower than the preset convergence threshold for several consecutive iterations, the quality feature conflict degree corresponding to the current global optimal solution is calculated. If the quality feature conflict degree is higher than the preset conflict degree threshold, the particles with the worst fitness value in the population are re-initialized with quality guidance.

[0138] The deodorization temperature, deodorization time, decolorization temperature, and decolorization temperature corresponding to the globally optimal position are taken as the optimal combination of process parameters.

[0139] In this embodiment of the invention, the process parameter constraints for each step are first obtained. These constraints include the deodorization temperature range, deodorization time range, decolorization temperature range, and decolorization time range. These constraints are determined comprehensively based on equipment design parameters, process safety requirements, and product quality standards. The deodorization temperature range is set to 220 to 270 degrees Celsius. Below 220 degrees Celsius, the deacidification and deodorization effects are insufficient; above 270 degrees Celsius, the thermal deterioration of the oil is accelerated, and the equipment safety risk increases significantly. The deodorization time range is set to 60 to 180 minutes. Below 60 minutes, the vapor-liquid contact is insufficient; above 180 minutes, energy consumption increases, and the risk of oil quality deterioration rises. The decolorization temperature range is set to 80 to 120 degrees Celsius. Below 80 degrees Celsius, the adsorption activity is insufficient; above 120 degrees Celsius, pigment oxidation is detrimental to decolorization. The decolorization time range is set to 20 to 60 minutes, within which sufficient adsorption is ensured while avoiding overheating.

[0140] Furthermore, an initial particle swarm is generated within the constraints of the process parameters. The position vector of each particle includes four dimensions: deodorization temperature, deodorization time, decolorization temperature, and decolorization time. The population size of the particle swarm is set to 100 particles, which achieves a good balance between computational accuracy and computational efficiency. The positions of the initial particles are generated randomly and uniformly within their respective constraints, and the initial velocities of the particles are randomly generated within 10% of their respective constraints.

[0141] Furthermore, the generation strategy for the initial solution set is determined based on the quality deviation. A quality deviation threshold is set, which is determined using statistical process control methods: quality deviation data for all batches of the same type of raw material over the past year are extracted from the historical raw material database, and their mean and standard deviation are calculated. The mean plus twice the standard deviation is used as the quality deviation threshold. The principle is that when the quality deviation is below this threshold, it indicates that the raw material quality is within the normal statistical fluctuation range; when the quality deviation exceeds this threshold, it indicates that the raw material quality has a statistically significant anomaly, requiring targeted optimization of indicators with quality risks.

[0142] Specifically, when the quality deviation is lower than or equal to the quality deviation threshold, it indicates that the deviations of the three quality indicators relative to their respective benchmark values ​​are all within the normal statistical fluctuation range, and the initial solution set is generated by uniform sampling within the constraints of the process parameters. The uniform sampling generation method ensures that the initial particle swarm is distributed as evenly as possible in the search space, which is beneficial for the algorithm to fully explore all regions in the early stages of iteration.

[0143] When the quality deviation exceeds the quality deviation threshold, it indicates that at least one quality indicator has deviated beyond the normal statistical fluctuation range. The initial solution set is then shifted towards the optimal parameter range corresponding to the quality feature with the largest deviation. Specifically, the method involves comparing the deviations of acid value, peroxide value, and trans fatty acid, extracting the quality feature with the largest deviation as the primary focus, and using the optimal process parameter range corresponding to that quality feature as the shift direction for the initial solution set.

[0144] The correspondence between the quality characteristic with the largest deviation and the optimal range of process parameters is determined according to the following principle. When the acid value deviation is the largest, it indicates that the free fatty acid content of the crude oil is significantly higher than the historical typical level, the deacidification burden is increased, and it is necessary to appropriately increase the deodorization temperature and extend the deodorization time to enhance the deacidification effect. The deacidification reaction follows the esterification reaction kinetics principle, the reaction rate increases with increasing temperature, and the stripping deacidification efficiency increases with the extension of the gas-liquid contact time. According to the statistical distribution of process parameters corresponding to batches that meet the acid value standard in historical data, the compliance rate is highest in the deodorization temperature range of 245 to 265 degrees Celsius and the compliance rate is highest in the deodorization time range of 110 to 150 minutes. Therefore, when the acid value deviation is the largest, the initial solution set is shifted to the optimal parameter range of deodorization temperature 245 to 265 degrees Celsius and deodorization time 110 to 150 minutes, and the remaining variables are randomly and uniformly generated within their respective constraints.

[0145] When the peroxide value deviation is at its maximum, it indicates that the current primary oxidation level of the crude oil is significantly higher than the historical typical level. Peroxide value is a direct indicator of the degree of oil oxidation. High temperature and prolonged heating will accelerate the decomposition of hydroperoxides and the generation of secondary oxidation products. Therefore, it is necessary to appropriately control the decolorization temperature and shorten the high-temperature treatment time. According to the statistical distribution of process parameters corresponding to batches that meet the peroxide value standard in historical data, the highest compliance rate is achieved when the decolorization temperature is in the range of 85 to 100 degrees Celsius and the highest compliance rate is achieved when the deodorization time is in the range of 80 to 120 minutes. Therefore, when the peroxide value deviation is at its maximum, the initial set of solutions will shift towards the optimal parameter range of 85 to 100 degrees Celsius for decolorization temperature and 80 to 120 minutes for deodorization time.

[0146] When the deviation of trans fatty acids is at its maximum, it indicates that the current trans fatty acid content of crude oil has significantly deviated from the historical typical level. The formation of trans fatty acids follows geometric isomerization reaction kinetics, and its formation rate is exponentially positively correlated with temperature. The isomerization reaction rate accelerates significantly when the deodorization temperature exceeds 240 degrees Celsius; therefore, it is necessary to appropriately lower the deodorization temperature to suppress the formation of trans fatty acids. According to the statistical distribution of process parameters corresponding to batches meeting trans fatty acid standards in historical data, the highest compliance rate is achieved within the deodorization temperature range of 225 to 240 degrees Celsius and the highest compliance rate is achieved within the deodorization time range of 60 to 100 minutes. Therefore, when the deviation of trans fatty acids is at its maximum, the initial set of components shifts towards the preferred parameter range of 225 to 240 degrees Celsius for deodorization and 60 to 100 minutes for deodorization time.

[0147] Furthermore, the boundary values ​​of the optimal parameter range were determined using the following statistical method: Production data for all compliant batches of this oil type within the past two years were extracted from the historical database. A compliant batch refers to a batch where the acid value, peroxide value, and trans fatty acid content of the finished oil all meet the quality standard thresholds. For each process parameter, the distribution of the parameter value in the compliant batches was statistically analyzed. The 25th and 75th quantiles of this distribution were taken as the lower and upper limits of the optimal parameter range, respectively. This range covers the middle 50% of the parameter values ​​in the compliant batches. The principle is that the process parameter values ​​in the compliant batches represent the effective range of values ​​for each parameter under the premise of ensuring final quality compliance. The 25th to 75th quantile range eliminates the interference of extreme values ​​and reflects the stable concentration area of ​​the optimal process parameters.

[0148] Furthermore, during the initial solution set shift towards the optimal parameter interval, the positions of 70% of the particles in the population are randomly generated within the optimal parameter interval using a normal distribution. The mean of the normal distribution is the median of the optimal parameter interval, and the standard deviation is one-sixth of the width of the optimal parameter interval. The principle behind this setting is that a normal distribution ensures that the vast majority of particles are concentrated near the center of the optimal interval. The central region represents the most densely populated area of ​​values ​​in historical batches of successful samples, serving as the optimal starting point for the algorithm's search. The standard deviation being one-sixth of the interval width reduces the probability density at the boundary of the optimal interval to approximately 5% of the central peak, ensuring that over 99% of the particles fall within the optimal interval. The positions of the remaining 30% of the particles in the population are uniformly and randomly generated within the complete constraints to maintain a certain level of population diversity.

[0149] Furthermore, a multi-objective fitness function is constructed based on the adjusted weights of each optimization objective. The fitness function is calculated as follows: Fitness value = Normalized weight of the objective that minimizes total energy consumption × Normalized total energy consumption value + Normalized weight of the objective that maximizes temperature gradient matching coefficient × Temperature gradient matching coefficient + Normalized weight of the objective that minimizes quality feature conflict degree × Normalized quality feature conflict degree. Normalized total energy consumption value = (Maximum total energy consumption - Current total energy consumption) / (Maximum total energy consumption - Minimum total energy consumption). The temperature gradient matching coefficient itself ranges from 0 to 1 and does not require normalization. Normalized quality feature conflict degree = (Maximum quality feature conflict degree - Current quality feature conflict degree) / (Maximum quality feature conflict degree - Minimum quality feature conflict degree). A larger fitness function value indicates better overall particle performance.

[0150] It should be noted that both the maximum and minimum total energy consumption are taken from the total energy consumption of all particles in the current particle swarm. Therefore, for any particle in the swarm, its current total energy consumption must be between the minimum and maximum total energy consumption. Thus, the difference between the maximum total energy consumption and the current total energy consumption is always greater than or equal to zero; there is no possibility of a negative value. When the absolute value of the difference between the maximum and minimum total energy consumption is less than a preset minimum value, such as 10 to the power of -6, the normalized value of the total energy consumption of all particles is uniformly set to 1. Similarly, the calculation of the normalized value of the quality characteristic conflict degree also has the same boundary conditions and is handled in the same fault-tolerant manner. When the denominator approaches zero, the normalized value of the quality characteristic conflict degree of all particles is uniformly set to 1.

[0151] Furthermore, in each iteration, the fitness value of each particle is calculated, and the individual optimal position and global optimal position of each particle are updated. The individual optimal position is the position where each particle has the highest fitness value in all iterations, and the global optimal position is the position where all particles have the highest fitness value in all iterations.

[0152] Furthermore, based on the velocity and position update formulas of the particle swarm optimization algorithm, the velocity and position of each particle are iteratively updated. The particle's velocity vector in the (t+1)th iteration = inertia weight × particle's velocity vector in the tth iteration + first acceleration constant × first random number × (particle's optimal position - particle's position in the tth iteration) + second acceleration constant × second random number × (global optimal position - particle's position in the tth iteration). Here, the inertia weight decreases linearly from 0.9 to 0.4 during iteration, and inertia weight = 0.9 - current iteration number × 0.5 / maximum iteration number. Both the first and second acceleration constants are set to 2.0, and the first and second random numbers are random numbers between 0 and 1. The particle's position vector in the (t+1)th iteration = particle's position vector in the tth iteration + particle's velocity vector in the (t+1)th iteration. After each position update, it is checked whether each dimension exceeds the constraint range. If it does, it is pulled back to the nearest boundary value, and the velocity of the corresponding dimension is set to zero.

[0153] During the iteration process, if the improvement in the global optimal fitness value after 10 consecutive iterations is lower than a preset convergence threshold (set to 0.001), indicating a relative improvement rate of less than one-thousandth, the algorithm is considered to be potentially trapped in a local optimum. In this case, the quality feature conflict degree corresponding to the current global optimal solution is calculated. If the conflict degree is higher than a preset conflict degree threshold (set to 0.3), it indicates that while the current solution performs well on some objectives, significant conflicts exist between quality indicators. Therefore, a quality-oriented reinitialization is needed for the bottom 20% of particles with the worst fitness values ​​in the population. The principle behind this strategy is that in the later stages of the Particle Swarm Optimization (PSO) algorithm, the positions of all particles gradually converge towards the global optimum, leading to particle convergence and loss of population diversity. The reinitialization method involves first calculating the attainment rates of the three quality features corresponding to the current global optimal solution, and then identifying the quality indicator with the lowest attainment rate, which is the main cause of the increased quality feature conflict degree. If the lowest acceptable value is acid value, it indicates that insufficient deacidification is the main source of conflict. The particles to be reinitialized are then regenerated in the upper half of the deodorization temperature constraint range to enhance the deacidification effect. If the lowest acceptable value is peroxide value or trans fatty acids, it indicates that oxidation control or isomerization inhibition is the main source of conflict. The particles to be reinitialized are then regenerated in the lower half of the deodorization temperature constraint range to inhibit the increase of peroxide value or the formation of trans fatty acids. Deodorization time, decolorization temperature, and decolorization time are regenerated uniformly and randomly within their respective complete constraint ranges. Simultaneously, the velocities of these particles are all set to zero after reinitialization.

[0154] The criteria for setting the convergence improvement threshold of 0.001 are as follows: the fitness value is obtained by weighted summation of three factors: the normalized value of total energy consumption, the temperature gradient matching coefficient, and the normalized value of quality characteristic conflict. The calculation accuracy of each normalized value is limited by the accuracy of the actual measuring instruments. The accuracy of thermocouples and resistance temperature detectors (RTDs) for temperature measurement is ±0.5 to 1.0 degrees Celsius, and the accuracy of flow meters in energy consumption calculation is ±1.0%. Therefore, the effective resolution in fitness value calculation is on the order of 0.001. Thus, when the fitness value improvement is less than 0.001 after multiple iterations, this improvement is below the physical threshold that the actual measurement system can distinguish, meaning it cannot constitute a meaningful improvement in process parameters and can be used as a convergence criterion.

[0155] The rationale for setting the number of consecutive iterations to 10 is based on runs theory in statistical process control. If the algorithm has converged to near the optimal solution, fluctuations in the fitness value near the optimal value can be considered random noise, and the probability of improvement in each iteration falling below the threshold is approximately 0.5. The probability of 10 consecutive iterations falling below the threshold in a random sequence is 0.5 to the power of 10, approximately one in a thousand, which is statistically considered a low-probability event. Therefore, when there is no significant improvement after 10 consecutive iterations, it is determined that the algorithm has either fallen into a local optimum or has sufficiently converged.

[0156] The basis for setting the quality conflict threshold of 0.3 is as follows: This threshold is derived from the statistical distribution analysis of quality feature conflict in weight 6. The calculation results of the quality feature conflict in historical compliant batches show that its standard deviation is approximately 0.1, the mean is approximately 0.15, and the mean plus twice the standard deviation is 0.35, which is rounded down to 0.3. When the quality feature conflict is higher than 0.3, it indicates that the conflict of the current solution exceeds more than 95% of the cases in historical compliant batches, which is a statistical outlier. Although the energy consumption target is good, the conflict between quality indicators is still at an unacceptable level, requiring a re-initialization to guide the search towards better quality.

[0157] Finally, the deodorization temperature, deodorization time, decolorization temperature, and decolorization temperature corresponding to the globally optimal position are output as the optimal combination of process parameters.

[0158] Step S400 in the method provided in this embodiment of the invention, as follows: Figure 2 As shown, it also includes:

[0159] Based on the temperature and time of each process in the optimal combination of process parameters, the acid value, peroxide value and trans fatty acid compliance of the finished oil are predicted by the quality prediction model.

[0160] During the execution of each process according to the optimal combination of process parameters, finished oil samples are collected at preset intervals to test the actual acid value, actual peroxide value and actual trans fatty acid content, and to calculate the compliance rate of actual acid value, actual peroxide value and actual trans fatty acid.

[0161] Calculate the absolute value of the deviation between the predicted and actual compliance rates for each quality characteristic;

[0162] When the absolute value of the deviation of each quality characteristic is lower than the first threshold, the current process parameters remain unchanged;

[0163] When the absolute value of the deviation of any quality feature is between the first threshold and the second threshold, the quality deviation is updated according to the type of quality feature with the largest deviation, the weight allocation of each optimization objective is recalculated, and the full re-optimization process is triggered.

[0164] When the absolute value of the deviation of any quality characteristic exceeds the second threshold, the process parameter is determined to be abnormal, an early warning signal is generated, the temperature and time of each process are restored to the baseline safe parameters, and the manual intervention process is initiated.

[0165] In this embodiment of the invention, based on the temperature and time of each process in the optimal combination of process parameters, the acid value compliance, peroxide value compliance, and trans fatty acid compliance of the finished oil are predicted by a quality prediction model, and these are used as the predicted compliance.

[0166] Furthermore, during the execution of each process according to the optimal combination of process parameters, finished oil samples are collected at a preset cycle. The preset cycle is set based on the stability of the production line and the response speed of the testing equipment, with sampling every 2 hours. The collected samples are tested according to national standard methods for actual acid value, actual peroxide value, and actual trans fatty acid content. The compliance rate of actual acid value, actual peroxide value, and actual trans fatty acid is calculated. The calculation method is the same as the calculation method for predicted compliance rate, that is, actual acid value compliance rate = 1 - actual acid value divided by the upper limit of acid value. When the actual acid value is higher than the upper limit of acid value, the compliance rate is 0. The compliance rates of peroxide value and trans fatty acid are calculated in the same way.

[0167] Further, calculate the absolute value of the deviation between the predicted and actual achievement rates of each quality characteristic. Absolute deviation = .

[0168] When the absolute value of the deviation of each quality characteristic is lower than the first threshold, the current process parameters remain unchanged. In this embodiment of the invention, the first threshold is 0.05, that is, when the deviation between the predicted value and the actual value is within five percent, it indicates that the prediction accuracy of the quality prediction model is high enough, the actual execution effect of the process is consistent with the expectation, and no adjustment is required.

[0169] When the absolute value of the deviation of any quality feature is between a first threshold and a second threshold (in this embodiment, the second threshold is 0.15), it indicates that the deviation is between 5% and 15%, requiring attention. At this time, the quality deviation is updated based on the type of quality feature with the largest deviation. The relationship between the actual and predicted compliance of the quality feature is determined: if the actual compliance is lower than the predicted compliance, it indicates that the actual quality has deteriorated compared to expectations, posing a quality risk. The deviation of this indicator is increased by an increment, increment = (absolute deviation - first threshold) × 2; if the actual compliance is higher than the predicted compliance, it indicates that the actual quality is better than expected, and the optimization effect of the current process parameters on this indicator exceeds the model prediction. The deviation of this indicator is decreased by a reduction, reduction = (absolute deviation - first threshold) × 1. When multiple quality features with the largest deviation exist simultaneously, they are adjusted independently according to the deviation direction of each indicator. If the adjustment directions conflict, the indicator requiring an increase in deviation is prioritized to prevent quality risks.

[0170] The first threshold is set based on the statistical distribution characteristics of the model's prediction error. After the quality prediction model is established, the prediction error between the predicted and measured values ​​of each quality indicator can be calculated using the validation set samples, and an error distribution model can be constructed accordingly. At a given confidence level, such as 95%, the boundary of the confidence interval for the predicted value is the first threshold. Its physical meaning is: when the actual detected value falls within this confidence interval, it indicates that the deviation between the model's predicted value and the actual value does not exceed the model's inherent prediction error range, which is a normal statistical fluctuation and should not trigger adjustment.

[0171] Meanwhile, the second threshold is set based on the risk boundary near the quality standard threshold. Based on the inverse principle of process capability index in statistics, the second threshold is set when the compliance rate shifts towards the quality standard threshold to a point where process capability is insufficient to guarantee that 99.7% of batches are within the quality standard threshold. Specifically, using the quality standard threshold as a benchmark, the compliance rate shift corresponding to the three process standard deviations is used as the second threshold. The physical meaning of this shift is: if the actual value continues to shift towards the quality standard threshold, a large number of batches will approach or even exceed the quality standard threshold, significantly increasing the risk of quality non-compliance. The two thresholds divide the deviation range into three levels: normal fluctuation range, early warning adjustment range, and abnormal protection range, corresponding to three response strategies: maintaining current parameters, triggering re-optimization, and forcibly restoring safe parameters, respectively.

[0172] Furthermore, when the absolute value of the deviation of any quality characteristic exceeds the second threshold (greater than 15%), the process parameter is determined to be abnormal, an early warning signal is generated, and the operators are notified via an audible and visual alarm and the central control room display interface. At this time, the temperature and time of each process are restored to the baseline safety parameters. Simultaneously, the manual intervention process is initiated, notifying the process engineer to investigate the cause of the abnormality on-site. Possible causes include equipment failure, instrument inaccuracy, and drastic fluctuations in raw material quality.

[0173] The baseline safety parameters are a set of conservative process parameters determined based on historical experience, set according to the following principles: the deodorization temperature is selected below the inflection point temperature that significantly accelerates the isomerization reaction of trans fatty acids, while being above the minimum temperature required to ensure the full volatilization of free fatty acids; the deodorization time is selected near the mass transfer equilibrium point where the gas-liquid mass transfer has basically reached equilibrium and further extending the time has a diminishing effect on the deacidification effect; the decolorization temperature is selected within the optimal temperature range that can fully utilize the activity of the adsorbent without triggering the oxidation side reaction of heat-sensitive pigments; the decolorization time is selected near the adsorption saturation point where the adsorption process is close to equilibrium and further extending the time has limited effect on the decolorization effect. The specific values ​​of each parameter are adjusted accordingly based on the type of oil and the characteristics of the equipment during application. Their common feature is that they can ensure the quality of the finished product meets the standards under any raw material quality conditions, serving as a safety net in case of process abnormalities.

[0174] As a specific implementation method, the deodorization temperature can be set to 235 degrees Celsius, the deodorization time to 120 minutes, the decolorization temperature to 100 degrees Celsius, and the decolorization time to 40 minutes. Although this set of parameters is not the most energy-efficient, it can ensure that the quality of the finished product meets the standard requirements under any circumstances.

[0175] This invention establishes a one-to-one correspondence between three weight adjustment coefficients and three quality deviations, enabling the weights of each optimization objective to adaptively adjust according to the actual fluctuations in raw material quality. The greater the quality deviation, the higher the corresponding objective weight, achieving dynamic matching between the optimization direction and the raw material quality status. By reducing the weight of the total energy consumption minimization objective as each quality deviation increases, the potential for substandard finished product quality due to excessive pursuit of energy consumption minimization when quality risk is high is effectively avoided. By employing a linear decreasing inertial weight strategy and introducing a quality-oriented particle re-initialization mechanism in the later stages of iteration, the shortcomings of the standard particle swarm optimization algorithm, which is prone to getting trapped in local optima, are effectively improved. When the global optimal fitness value does not improve for a long time and the quality conflict degree remains high, inferior particles are redistributed towards areas with lower conflict degrees, maintaining population diversity while guiding the search towards regions with better quality.

[0176] Meanwhile, by constructing a three-level deviation response mechanism, the deviation between the predicted and actual values ​​is handled in a graded manner: when the deviation is small, the parameters remain unchanged to maintain production stability; when the deviation is medium, a complete re-optimization is triggered to achieve adaptive correction of process parameters; and when the deviation is large, the parameters are automatically restored to the baseline safety parameters and manual intervention is initiated. This not only ensures the continuity and stability of the production process, but also provides multi-level safeguards when the quality deviates, forming a complete closed-loop control link from prediction to execution to feedback adjustment.

[0177] S500: Adjusts the execution parameters of each process according to the optimal combination of process parameters.

[0178] In this embodiment of the invention, the optimal combination of process parameters obtained by the improved particle swarm optimization algorithm is sent to the setpoint register of the distributed control system (DCS) or programmable logic controller (PLC) via industrial Ethernet or fieldbus communication protocol. The optimal combination of process parameters includes at least the optimal values ​​for four dimensions: deodorization temperature, deodorization time, decolorization temperature, and decolorization time. Based on the received setpoints, the DCS or PLC compares the setpoints with the current values ​​collected in real time by the field sensors of each process through its internal control logic operation module, calculates the deviation, and outputs the corresponding control signal according to the proportional-integral-derivative (PID) control algorithm or fuzzy PID control algorithm.

[0179] Furthermore, after signal conditioning and power amplification, the control signal drives the actuators corresponding to each process. Specifically, the deodorization temperature is controlled by adjusting the opening of the pneumatic regulating valve on the inlet pipe of the heating medium in the deodorization tower. This regulating valve receives a 4 to 20 mA standard current signal output from the DCS or PLC. The valve opening increases or decreases with the increase of the control signal, thereby changing the flow rate of the heating medium and gradually bringing the oil temperature in the deodorization tower closer to and stabilizing at the optimal set value. The deodorization time is controlled by adjusting the feed flow rate of the oil in the deodorization tower or the speed of the discharge pump. The discharge pump is equipped with a frequency converter, and the DCS or PLC adjusts the pump speed by changing the frequency setpoint of the frequency converter, thereby controlling the residence time of the oil in the deodorization tower. The decolorization temperature is controlled by adjusting the flow rate of the heating medium in the jacket or coil of the decolorization reaction tank, and its actuator is the same as that for the deodorization temperature control. The decolorization time is controlled by adjusting the opening of the discharge valve of the decolorization reaction tank or the operating frequency of the discharge pump.

[0180] In summary, the embodiments of the present invention have at least the following technical effects:

[0181] This invention achieves precise quantification and graded recovery of process waste heat by calculating heat exchanger efficiency and temperature gradient matching coefficients in real time during the refining process, significantly reducing the amount of external heat source replenishment. By constructing quality deviation and quality feature conflict indices, the optimization process can simultaneously consider the inherent constraints between raw material batch fluctuations and the three quality indicators. By using quality deviation as the basis for weight allocation, the algorithm automatically adjusts the optimization focus under different raw material quality conditions, ensuring that the finished product quality remains under control. By introducing a quality-oriented population reinitialization mechanism in the particle swarm optimization process, the algorithm effectively avoids getting trapped in local optima, ensuring the global optimality of the solution. These methods together achieve the technical effect of synergistic optimization of energy consumption, heat energy utilization, and quality assurance in the oil refining process, effectively reducing total energy consumption while ensuring that all quality indicators of the finished oil meet the standards.

[0182] In summary, this invention solves the technical problem that it is difficult to achieve unified optimization of total energy consumption, cascade utilization of heat energy and finished product quality in existing oil refining processes, resulting in high energy consumption or difficulty in effectively controlling quality fluctuations.

[0183] Example 2, as Figure 3 As shown, based on the same inventive concept as the intelligent control method for oil processing parameters for energy consumption optimization provided in Embodiment 1, this embodiment of the invention also provides an intelligent control system for oil processing parameters for energy consumption optimization, comprising:

[0184] The data acquisition module 11 is used to collect real-time process parameters and raw material quality indicators for each step in the oil refining process. The real-time process parameters include at least the temperature of each step, the inlet and outlet temperatures of the heat exchanger, and the amount of direct steam used. The raw material quality indicators include at least the acid value of crude oil, the peroxide value, and the trans fatty acid content.

[0185] The efficiency calculation and cascade matching module 12 is used to calculate the real-time efficiency of the heat exchanger based on the inlet and outlet temperatures of the heat exchanger, and to calculate the temperature cascade matching coefficient between processes based on the temperature of each process and the real-time efficiency of the heat exchanger.

[0186] The quality evaluation module 13 is used to determine the quality deviation based on the raw material quality indicators and to calculate the quality characteristic conflict degree under the current process parameters. The quality characteristic conflict degree characterizes the degree of mutual constraint between crude oil acid value, peroxide value and trans fatty acid content.

[0187] The multi-objective optimization module 14 is used to solve the optimal combination of process parameters with multiple objectives, namely minimizing total energy consumption, maximizing temperature gradient matching coefficient and minimizing quality characteristic conflict degree, through an improved optimization algorithm. The weight allocation of each objective is dynamically adjusted according to the quality deviation.

[0188] The execution control module 15 is used to control the execution parameters of each process according to the optimal combination of process parameters.

[0189] In one embodiment, the data acquisition module 11 is specifically used for:

[0190] The deodorization temperature, decolorization temperature, deacidification temperature, and degumming temperature are collected by thermocouple temperature sensors installed in each process.

[0191] The inlet and outlet temperatures of the heat exchanger are collected by resistance temperature sensors installed at the inlet and outlet of the hot side and the inlet and outlet of the cold side of the heat exchanger.

[0192] The direct steam consumption is collected by a vortex flow meter installed on the steam pipeline;

[0193] The acid value, peroxide value, and trans fatty acid content of crude oil were collected using an online near-infrared spectrometer installed on the crude oil feed pipeline.

[0194] The steam energy consumption of the deodorization process is calculated based on the direct steam consumption, steam pressure, and steam enthalpy of the deodorization process. The power energy consumption of each process is calculated based on the heating power and processing time of each process.

[0195] Calculate the cooling energy consumption based on the cooling water flow rate, inlet and outlet temperature difference, and specific heat capacity of the cooling system.

[0196] The total energy consumption is obtained by weighting and summing the steam energy consumption, electricity energy consumption, and cooling energy consumption. The weighting coefficients for the weighted summation are determined based on the standard coal conversion coefficients for various energy sources.

[0197] In one embodiment, the efficiency calculation and cascade matching module 12 is specifically used for:

[0198] Obtain the hot-side inlet temperature, hot-side outlet temperature, cold-side inlet temperature, and cold-side outlet temperature of the heat exchanger, as well as the hot-side flow rate and the specific heat capacity of the hot-side oil.

[0199] The actual heat transfer of the heat exchanger is calculated based on the temperature difference between the hot side inlet temperature and the hot side outlet temperature, the hot side flow rate, and the specific heat capacity of the hot side oil.

[0200] The theoretical maximum heat transfer of the heat exchanger is calculated based on the temperature difference between the hot side inlet temperature and the cold side inlet temperature, the hot side flow rate, and the specific heat capacity of the hot side oil.

[0201] The real-time efficiency of the heat exchanger is determined based on the degree to which the actual heat transfer of the heat exchanger is maintained relative to the theoretical maximum heat transfer of the heat exchanger.

[0202] Specifically, when the real-time efficiency of the heat exchanger is lower than the preset efficiency threshold, a heat exchanger cleaning reminder signal is generated. The preset efficiency threshold is adaptively set according to the heat exchanger type and running time.

[0203] Calculate the temperature gradient matching coefficient between processes based on the temperature of each process and the real-time efficiency of the heat exchanger, including:

[0204] The outlet temperature of the upstream process is extracted as the waste heat source temperature, and the inlet demand temperature of the downstream process is extracted as the heat consumption temperature.

[0205] Calculate the recoverable heat from the upstream process based on the waste heat source temperature, the oil flow rate from the upstream process, and the specific heat capacity of the oil.

[0206] Calculate the heat demand of the downstream process based on the heat application temperature, the oil flow rate of the downstream process, and the specific heat capacity of the oil.

[0207] Calculate the actual heat transfer capacity of the heat exchanger based on its real-time efficiency, recoverable heat, and required heat.

[0208] The temperature gradient matching coefficient is determined based on the degree to which the actual transferable heat meets the required heat.

[0209] In one embodiment, the quality evaluation module 13 is specifically used for:

[0210] The benchmark acid value, benchmark peroxide value, and benchmark trans fatty acid content of the same raw material type are extracted from the historical raw material database. The benchmark index is the median of the corresponding index of the historical raw material.

[0211] Calculate the deviation of the current crude oil acid value from the benchmark acid value, the deviation of the current crude oil peroxide value from the benchmark peroxide value, and the deviation of the current crude oil trans fatty acid content from the benchmark trans fatty acid content, and use these as the acid value deviation, peroxide value deviation, and trans fatty acid deviation, respectively.

[0212] Obtain the preset values ​​of process parameters for each process step. The preset values ​​of process parameters include at least the deodorization temperature, deodorization time, decolorization temperature, and decolorization time.

[0213] Based on the preset values ​​of process parameters and raw material quality indicators, the predicted oleic acid value, predicted peroxide value and predicted trans fatty acid content after the deodorization process are predicted by a pre-trained quality prediction model.

[0214] The quality standard thresholds for refined oil products are obtained, including the upper limit of acid value, the upper limit of peroxide value, and the upper limit of trans fatty acid content.

[0215] The degree of acid price compliance is determined by the relative deviation between the predicted oleic acid price and the upper limit of the acid price.

[0216] The peroxide value compliance rate is determined based on the relative deviation between the predicted peroxide value and the upper limit of the peroxide value.

[0217] The trans fatty acid compliance rate is determined based on the relative deviation between the predicted trans fatty acid content and the upper limit of trans fatty acid content.

[0218] The deviation of quality characteristics is determined comprehensively based on the degree of deviation of acid value compliance, peroxide value compliance, and trans fatty acid compliance from the standard values.

[0219] The standard deviations among acid value compliance, peroxide value compliance, and trans fatty acid compliance are calculated as the dispersion of quality characteristics.

[0220] The quality feature conflict degree is obtained by weighted summing of the quality feature deviation and the quality feature dispersion.

[0221] In one embodiment, the multi-objective optimization module 14 is specifically used for:

[0222] Obtain the deviation of acid value, peroxide value, and trans fatty acid;

[0223] The first weighting adjustment coefficient is determined based on the acid value deviation, and the first weighting adjustment coefficient is positively correlated with the acid value deviation.

[0224] The second weighting adjustment coefficient is determined based on the peroxide value deviation, and the second weighting adjustment coefficient is positively correlated with the peroxide value deviation.

[0225] The third weighting adjustment coefficient is determined based on the deviation of trans fatty acids, and the third weighting adjustment coefficient is positively correlated with the deviation of trans fatty acids.

[0226] Obtain the basic weights for the target of minimizing total energy consumption, the basic weights for the target of maximizing temperature gradient matching coefficient, and the basic weights for the target of minimizing quality characteristic conflict.

[0227] The adjusted weights of the total energy consumption minimization target are obtained by dividing the base weights of the total energy consumption minimization target by the arithmetic mean of the first weight adjustment coefficient, the second weight adjustment coefficient, and the third weight adjustment coefficient.

[0228] Multiply the base weight of the temperature gradient matching coefficient maximization target by the first weight adjustment factor to obtain the adjusted temperature gradient matching coefficient maximization target weight;

[0229] The adjusted weights for minimizing quality feature conflict are obtained by multiplying the base weight of the quality feature conflict minimization objective by the average of the second and third weight adjustment coefficients.

[0230] The optimal combination of process parameters is solved using an improved optimization algorithm, including:

[0231] Obtain the process parameter constraint ranges for each process, including the deodorization temperature range, deodorization time range, decolorization temperature range, and decolorization time range;

[0232] An initial particle swarm is generated within the constraints of process parameters. The position vector of each particle includes the deodorization temperature, deodorization time, decolorization temperature, and decolorization time.

[0233] The initial solution set generation strategy is determined based on the quality deviation.

[0234] A multi-objective fitness function is constructed based on the adjusted weights of each optimization objective. In each iteration, the fitness value of each particle is calculated, and the individual optimal position and global optimal position of each particle are updated.

[0235] Based on the velocity update formula and position update formula of the particle swarm optimization algorithm, the velocity and position of each particle are iteratively updated until the number of iterations reaches the preset upper limit or the fitness value converges.

[0236] In the iteration process, when the improvement of the global optimal fitness value is lower than the preset convergence threshold for several consecutive iterations, the quality feature conflict degree corresponding to the current global optimal solution is calculated. If the quality feature conflict degree is higher than the preset conflict degree threshold, the particles with the worst fitness value in the population are re-initialized with quality guidance.

[0237] The deodorization temperature, deodorization time, decolorization temperature, and decolorization temperature corresponding to the globally optimal position are taken as the optimal combination of process parameters.

[0238] Based on the temperature and time of each process in the optimal combination of process parameters, the acid value, peroxide value and trans fatty acid compliance of the finished oil are predicted by the quality prediction model.

[0239] During the execution of each process according to the optimal combination of process parameters, finished oil samples are collected at preset intervals to test the actual acid value, actual peroxide value and actual trans fatty acid content, and to calculate the compliance rate of actual acid value, actual peroxide value and actual trans fatty acid.

[0240] Calculate the absolute value of the deviation between the predicted and actual compliance rates for each quality characteristic;

[0241] When the absolute value of the deviation of each quality characteristic is lower than the first threshold, the current process parameters remain unchanged;

[0242] When the absolute value of the deviation of any quality feature is between the first threshold and the second threshold, the quality deviation is updated according to the type of quality feature with the largest deviation, the weight allocation of each optimization objective is recalculated, and the full re-optimization process is triggered.

[0243] When the absolute value of the deviation of any quality characteristic exceeds the second threshold, the process parameter is determined to be abnormal, an early warning signal is generated, the temperature and time of each process are restored to the baseline safe parameters, and the manual intervention process is initiated.

[0244] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0245] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0246] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. A method for intelligent control of oil processing parameters for energy consumption optimization, characterized in that, include: Real-time process parameters and raw material quality indicators are collected for each step in the oil refining process. The real-time process parameters include at least the temperature of each step, the inlet and outlet temperatures of the heat exchanger, and the direct steam consumption. The raw material quality indicators include at least the crude oil acid value, peroxide value, and trans fatty acid content. The real-time efficiency of the heat exchanger is calculated based on the inlet and outlet temperatures of the heat exchanger, and the temperature gradient matching coefficient between processes is calculated based on the temperature of each process and the real-time efficiency of the heat exchanger. The quality deviation is determined based on the raw material quality indicators, and the quality characteristic conflict degree under the current process parameters is calculated. The quality characteristic conflict degree characterizes the degree of mutual constraint between crude oil acid value, peroxide value and trans fatty acid content. With the goals of minimizing total energy consumption, maximizing temperature gradient matching coefficient, and minimizing quality characteristic conflict, an improved optimization algorithm is used to solve for the optimal combination of process parameters, wherein the weight allocation of each goal is dynamically adjusted according to the quality deviation. The execution parameters of each process are adjusted according to the optimal combination of process parameters.

2. The intelligent control method for oil processing parameters for energy consumption optimization as described in claim 1, characterized in that, Real-time process parameters and raw material quality indicators for each step in the oil refining process are collected, including: The deodorization temperature, decolorization temperature, deacidification temperature, and degumming temperature are collected by thermocouple temperature sensors installed in each process. The inlet and outlet temperatures of the heat exchanger are collected by resistance temperature sensors installed at the inlet and outlet of the hot side and the inlet and outlet of the cold side of the heat exchanger. The direct steam consumption is collected by a vortex flow meter installed on the steam pipeline; The acid value, peroxide value, and trans fatty acid content of crude oil were collected using an online near-infrared spectrometer installed on the crude oil feed pipeline. The steam energy consumption of the deodorization process is calculated based on the direct steam consumption, steam pressure, and steam enthalpy of the deodorization process. The power energy consumption of each process is calculated based on the heating power and processing time of each process. Calculate the cooling energy consumption based on the cooling water flow rate, inlet and outlet temperature difference, and specific heat capacity of the cooling system. The total energy consumption is obtained by weighting and summing the steam energy consumption, electricity energy consumption, and cooling energy consumption. The weighting coefficients for the weighted summation are determined based on the standard coal conversion coefficients for various energy sources.

3. The intelligent control method for oil processing parameters for energy consumption optimization as described in claim 1, characterized in that, The real-time efficiency of the heat exchanger is calculated based on the inlet and outlet temperatures, including: Obtain the hot-side inlet temperature, hot-side outlet temperature, cold-side inlet temperature, and cold-side outlet temperature of the heat exchanger, as well as the hot-side flow rate and the specific heat capacity of the hot-side oil. The actual heat transfer of the heat exchanger is calculated based on the temperature difference between the hot side inlet temperature and the hot side outlet temperature, the hot side flow rate, and the specific heat capacity of the hot side oil. The theoretical maximum heat transfer of the heat exchanger is calculated based on the temperature difference between the hot side inlet temperature and the cold side inlet temperature, the hot side flow rate, and the specific heat capacity of the hot side oil. The real-time efficiency of the heat exchanger is determined based on the degree to which the actual heat transfer of the heat exchanger is maintained relative to the theoretical maximum heat transfer of the heat exchanger. Specifically, when the real-time efficiency of the heat exchanger is lower than a preset efficiency threshold, a heat exchanger cleaning reminder signal is generated. The preset efficiency threshold is adaptively set according to the heat exchanger type and running time.

4. The intelligent control method for oil processing parameters for energy consumption optimization as described in claim 3, characterized in that, The temperature gradient matching coefficient between processes is calculated based on the temperature of each process and the real-time efficiency of the heat exchanger, including: The outlet temperature of the upstream process is extracted as the waste heat source temperature, and the inlet demand temperature of the downstream process is extracted as the heat consumption temperature. Calculate the recoverable heat from the upstream process based on the waste heat source temperature, the oil flow rate from the upstream process, and the specific heat capacity of the oil. Calculate the heat demand of the downstream process based on the heat application temperature, the oil flow rate of the downstream process, and the specific heat capacity of the oil. Calculate the actual heat transfer capacity of the heat exchanger based on its real-time efficiency, recoverable heat, and required heat. The temperature gradient matching coefficient is determined based on the degree to which the actual transferable heat meets the required heat.

5. The intelligent control method for oil processing parameters for energy consumption optimization as described in claim 1, characterized in that, The quality deviation is determined based on the aforementioned raw material quality indicators, including: The benchmark acid value, benchmark peroxide value, and benchmark trans fatty acid content of the same raw material type are extracted from the historical raw material database. The benchmark index is the median of the corresponding index of the historical raw material. Calculate the deviation of the current crude oil acid value from the benchmark acid value, the deviation of the current crude oil peroxide value from the benchmark peroxide value, and the deviation of the current crude oil trans fatty acid content from the benchmark trans fatty acid content, and use these as the acid value deviation, peroxide value deviation, and trans fatty acid deviation, respectively.

6. The intelligent control method for oil processing parameters for energy consumption optimization as described in claim 5, characterized in that, Calculate the degree of conflict of quality characteristics under the current process parameters, including: Obtain the preset values ​​of process parameters for each process step, wherein the preset values ​​of process parameters include at least the deodorization temperature, deodorization time, decolorization temperature, and decolorization time; Based on the preset values ​​of the process parameters and the raw material quality indicators, the predicted oleic acid value, predicted peroxide value and predicted trans fatty acid content after the deodorization process are predicted by a pre-trained quality prediction model. The quality standard thresholds for refined oil products are obtained, including the upper limit of acid value, the upper limit of peroxide value, and the upper limit of trans fatty acid content. The degree of acid price compliance is determined by the relative deviation between the predicted oleic acid price and the upper limit of the acid price. The peroxide value compliance rate is determined based on the relative deviation between the predicted peroxide value and the upper limit of the peroxide value. The trans fatty acid compliance rate is determined based on the relative deviation between the predicted trans fatty acid content and the upper limit of trans fatty acid content. The deviation of quality characteristics is determined comprehensively based on the degree of deviation of acid value compliance, peroxide value compliance, and trans fatty acid compliance from the standard values. The standard deviations among the acid value compliance, peroxide value compliance, and trans fatty acid compliance are calculated as the quality characteristic dispersion. The quality feature conflict degree is obtained by weighted summing of the deviation degree and the dispersion degree of the quality feature.

7. The intelligent control method for oil processing parameters for energy consumption optimization according to claim 1, characterized in that, The weight allocation of each optimization objective is dynamically adjusted based on the quality deviation, including: Obtain the deviation of acid value, peroxide value, and trans fatty acid; A first weighting adjustment coefficient is determined based on the acid value deviation, and the first weighting adjustment coefficient is positively correlated with the acid value deviation; A second weighting adjustment coefficient is determined based on the peroxide value deviation, and the second weighting adjustment coefficient is positively correlated with the peroxide value deviation. A third weighting adjustment coefficient is determined based on the trans fatty acid deviation, and the third weighting adjustment coefficient is positively correlated with the trans fatty acid deviation. Obtain the basic weights for the target of minimizing total energy consumption, the basic weights for the target of maximizing temperature gradient matching coefficient, and the basic weights for the target of minimizing quality characteristic conflict. The adjusted weights of the total energy consumption minimization target are obtained by dividing the base weights of the total energy consumption minimization target by the arithmetic mean of the first weight adjustment coefficient, the second weight adjustment coefficient, and the third weight adjustment coefficient. Multiply the base weight of the temperature gradient matching coefficient maximization target by the first weight adjustment coefficient to obtain the adjusted temperature gradient matching coefficient maximization target weight; The adjusted weights for minimizing quality feature conflict are obtained by multiplying the base weights of the quality feature conflict minimization target by the average of the second and third weight adjustment coefficients.

8. The intelligent control method for oil processing parameters for energy consumption optimization according to claim 7, characterized in that, The optimal combination of process parameters is solved using an improved optimization algorithm, including: Obtain the process parameter constraint range for each process step, including the deodorization temperature range, deodorization time range, decolorization temperature range, and decolorization time range; An initial particle swarm is generated within the constraints of the process parameters, and the position vector of each particle includes the deodorization temperature, deodorization time, decolorization temperature, and decolorization time. The initial solution set generation strategy is determined based on the quality deviation. A multi-objective fitness function is constructed based on the adjusted weights of each optimization objective. In each iteration, the fitness value of each particle is calculated, and the individual optimal position and global optimal position of each particle are updated. Based on the velocity update formula and position update formula of the particle swarm optimization algorithm, the velocity and position of each particle are iteratively updated until the number of iterations reaches the preset upper limit or the fitness value converges. In the iteration process, when the improvement of the global optimal fitness value is lower than the preset convergence threshold for several consecutive iterations, the quality feature conflict degree corresponding to the current global optimal solution is calculated. If the quality feature conflict degree is higher than the preset conflict degree threshold, the particles with the worst fitness value in the population are re-initialized with quality guidance. The deodorization temperature, deodorization time, decolorization temperature, and decolorization temperature corresponding to the globally optimal position are taken as the optimal combination of process parameters.

9. The intelligent control method for oil processing parameters for energy consumption optimization according to claim 8, characterized in that, Also includes: Based on the temperature and time of each process in the optimal process parameter combination, the acid value compliance, peroxide value compliance, and trans fatty acid compliance of the finished oil are predicted by the quality prediction model. During the execution of each process according to the optimal process parameter combination, finished oil samples are collected at preset intervals to detect the actual acid value, actual peroxide value and actual trans fatty acid content, and to calculate the compliance rate of actual acid value, actual peroxide value and actual trans fatty acid. Calculate the absolute value of the deviation between the predicted and actual compliance rates for each quality characteristic; When the absolute value of the deviation of each quality characteristic is lower than the first threshold, the current process parameters remain unchanged; When the absolute value of the deviation of any quality feature is between the first threshold and the second threshold, the quality deviation is updated according to the type of quality feature with the largest deviation, the weight allocation of each optimization objective is recalculated, and the full re-optimization process is triggered. When the absolute value of the deviation of any quality characteristic exceeds the second threshold, the process parameter is determined to be abnormal, an early warning signal is generated, the temperature and time of each process are restored to the baseline safe parameters, and the manual intervention process is initiated.

10. An intelligent control system for oil processing parameters for energy consumption optimization, characterized in that, The method for intelligent control of oil processing parameters for energy consumption optimization as described in any one of claims 1 to 9 includes: The data acquisition module is used to collect real-time process parameters and raw material quality indicators for each step in the oil refining process. The real-time process parameters include at least the temperature of each step, the inlet and outlet temperatures of the heat exchanger, and the amount of direct steam used. The raw material quality indicators include at least the crude oil acid value, peroxide value, and trans fatty acid content. The efficiency calculation and cascade matching module is used to calculate the real-time efficiency of the heat exchanger based on the inlet and outlet temperatures of the heat exchanger, and to calculate the temperature cascade matching coefficient between processes based on the temperature of each process and the real-time efficiency of the heat exchanger. The quality evaluation module is used to determine the quality deviation based on the raw material quality indicators and to calculate the quality characteristic conflict degree under the current process parameters. The quality characteristic conflict degree characterizes the degree of mutual constraint between crude oil acid value, peroxide value and trans fatty acid content. The multi-objective optimization module is used to solve for the optimal combination of process parameters with multiple objectives, namely minimizing total energy consumption, maximizing temperature gradient matching coefficient, and minimizing quality characteristic conflict degree, through an improved optimization algorithm. The weight allocation of each objective is dynamically adjusted according to the quality deviation. The execution control module is used to control the execution parameters of each process according to the optimal combination of process parameters.

Images