Optimization method of composite antioxidant and preparation method of mixed base insulating oil of composite antioxidant

Abstract

The invention discloses a composite antioxidant optimization method and a preparation method of mixed base insulating oil thereof, and relates to the technical field of insulating oil formula optimization. According to the optimization method, historical characteristic data of an insulating oil formula are collected, a database is established through cleaning and standardization, an XGBoost algorithm model is constructed and trained, and new formula component prediction is achieved. According to the preparation method, a new formula is obtained based on the optimization method, specific base oil is selected, and the high-oxidation-stability insulating oil is obtained through pretreatment, composite antioxidant preparation, mixed dissolution under nitrogen protection and vacuum filtration. The method converts a traditional trial and error experiment into data-driven research and development, quickly finds a global optimal solution, shortens the research and development period, reduces the cost, effectively solves the problem of oxidation stability of the insulating oil, improves the operation reliability of equipment, and prolongs the service life of the equipment.

Description

Technical Field

[0001] This invention relates to the field of insulating oil formulation optimization technology, and in particular to a method for optimizing composite antioxidants and a method for preparing mixed-based insulating oils. Background Technology

[0002] Insulating oil, as a crucial insulating medium in oil-immersed transformers, directly impacts the operational reliability and service life of the equipment due to its oxidation stability. Currently, mineral insulating oils are widely used due to their excellent electrical properties and low cost; however, their poor biodegradability and low flash point are becoming increasingly prominent issues. While natural ester insulating oils (such as soybean oil) offer advantages in terms of environmental friendliness and renewability, the presence of unsaturated double bonds in their molecular structure makes them prone to oxidation under high temperatures and in the presence of oxygen. This leads to increased acid value, precipitation, and compromised insulation performance. Therefore, it is necessary to add composite antioxidants to natural ester insulating oils.

[0003] Traditional optimization of composite antioxidant formulations in insulating oils relies heavily on the experience and trial-and-error methods of researchers, resulting in long development cycles, high costs, and difficulty in finding globally optimal solutions. While machine learning technology has shown great potential in materials science in recent years, its application in insulating oil research is still relatively limited, particularly lacking systematic and engineerable solutions.

[0004] Therefore, there is a need for an optimization method for composite antioxidants and a method for preparing mixed-based insulating oils. Summary of the Invention

[0005] To address the problems of existing technologies where formulation optimization mainly relies on the experience and trial-and-error methods of experimenters, resulting in long development cycles, high costs, and difficulty in finding the global optimum, this invention provides a method for optimizing composite antioxidants and a method for preparing mixed-based insulating oils. This method transforms traditional trial-and-error experiments into data-driven development, providing a new paradigm for insulating oil formulation optimization, finding the global optimum more quickly, shortening the development cycle, reducing development costs, and achieving rapid optimization and performance prediction of insulating oil formulations. The specific technical solution is as follows: A method for optimizing composite antioxidants, comprising: S1: Collect historical characteristic data of insulating oil formulation; S2: Clean and standardize the collected feature data, and establish an insulating oil formulation-performance relationship database; S3: Construct an XGBoost algorithm model with recipe parameters as input features and performance indicators as output labels; S4: The dataset in the insulating oil formulation-performance relationship database is divided into training and testing sets. Cross-validation is used to train and test the XGBoost algorithm model in order to optimize the parameters of the XGBoost algorithm model. S5: Evaluate the predictive power of the trained XGBoost algorithm model using an independent validation set, and apply the XGBoost algorithm model to predict the compound antioxidant components of the new formulation.

[0006] Furthermore, in step S1, the historical characteristic data includes the base oil type, the mixing ratio of the base oil, the initial acid value of the base oil, the moisture content of the base oil, the type and ratio of antioxidants in the composite antioxidant, the amount of composite antioxidant added, the acid value of the insulating oil, and process parameters.

[0007] Furthermore, the process parameters include pretreatment temperature, stirring speed, and time.

[0008] Furthermore, the performance indicators include the acid value and precipitate content after aging.

[0009] Furthermore, in step S5, applying the XGBoost algorithm model to predict the compound antioxidant components of the new formulation includes the following steps: S51: Set target performance indicators based on actual application requirements, including acid value range and precipitate content range; S52: Based on the XGBoost algorithm model, a search is performed in the full parameter space to find the formulation region that meets the target; the formulation region includes the proportion range of base oil, the moisture content of base oil, the initial acid value of base oil, and the ratio range of corresponding paired composite antioxidants.

[0010] S53: Select potential formulations from the formulation region for experimental verification, and feed the verification results back to the insulating oil formulation-performance relationship database for continuous optimization of the XGBoost algorithm model.

[0011] A method for preparing a mixed-base insulating oil based on composite antioxidants includes the following steps: A new formulation of hybrid-based insulating oil was obtained by applying the above-described composite antioxidant optimization method; Soybean oil and mineral oil were selected as base oils based on the base oil ratio range, base oil moisture content, and initial acid value of the base oil in the new formula. Place the base oil in a clean, heatable reactor and slowly heat it to 60-80°C. Maintain the temperature at 60-80°C and stir at 200-400 rpm for 20-40 minutes to reduce the viscosity of the base oil, remove dissolved gases, and create favorable conditions for the uniform dispersion of the antioxidants. In a dry environment, using an analytical balance, weigh the main antioxidant BHT and the auxiliary antioxidant PG according to the compound antioxidant ratio range of the new formulation; place the weighed powder in a mixer and mix thoroughly for 10-20 minutes at a speed of 20-40 rpm to obtain the compound antioxidant premix. Under the protection of continuous high-purity nitrogen, the composite antioxidant premix is ​​slowly and in batches added to the base oil that has been preheated and cooled to 20-35°C; mechanical stirring is started under nitrogen protection, the speed is controlled at 500-800 rpm, and stirring is continued for 30-60 minutes to ensure that the antioxidant molecules are fully dissolved and uniformly dispersed in the oil phase to form a stable homogeneous system. The oil mixed with composite antioxidants was transferred to a vacuum filtration device and filtered at 50-70℃ and a vacuum degree of -0.08 to -0.10 MPa to obtain a high oxidation stability insulating oil.

[0012] Furthermore, the initial acid value of the base oil is less than 0.05 mg KOH / g.

[0013] Furthermore, the water content of the base oil is less than 50 ppm.

[0014] Furthermore, the ratio of the primary antioxidant BHT to the secondary antioxidant PG is BHT:PG = 1:5-5:1.

[0015] Furthermore, the base oil ratio range is: mineral oil and soybean oil mixed in a ratio of 95:5 to 5:95.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Transforming traditional trial-and-error experiments into data-driven R&D provides a new paradigm for optimizing insulating oil formulations, enabling faster identification of the globally optimal solution, shortening the R&D cycle, and reducing R&D costs. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0018] Figure 1 This is a flowchart illustrating a method for optimizing a composite antioxidant. Figure 2 This is a schematic diagram of a process for preparing a hybrid insulating oil based on composite antioxidants. Detailed Implementation

[0019] 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, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be understood that, when used in this application, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0021] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should also be further understood that the term “and / or” as used in this application refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations.

[0023] Example 1 like Figure 1 The diagram shows a flowchart of a method for optimizing composite antioxidants, including: S1: Collect historical characteristic data of insulating oil formulation.

[0024] The data collection described in this step includes historical experimental data, literature data, and newly added experimental data.

[0025] Furthermore, in step S1, the historical characteristic data includes the base oil type, the mixing ratio of the base oil, the initial acid value of the base oil, the moisture content of the base oil, the type and ratio of antioxidants in the composite antioxidant, the amount of composite antioxidant added, the acid value of the insulating oil, and process parameters.

[0026] Furthermore, the process parameters include pretreatment temperature, stirring speed, and time.

[0027] A total of 485 sets of historical characteristic data were collected, of which 320 sets originated from the applicant's R&D experimental records over the past 5 years, and 165 sets were compiled from publicly available academic literature (such as IEEE Dielectrics and Electrical Insulation, Transactions on Electrical Insulation, etc.) and industry standard databases. The newly added experimental data consists of 50 sets of confirmatory experimental data designed based on preliminary model predictions. The dataset covers the mainstream formulation space with mineral oil content ranging from 5% to 95% and total added compound antioxidant content ranging from 0.05% to 1.5%.

[0028] S2: Clean and standardize the collected feature data, and establish an insulating oil formulation-performance relationship database.

[0029] By systematically collecting experimental data, literature data, and newly added experimental data, a comprehensive database of insulating oil formulation-performance relationships is constructed. This database not only contains basic formulation information but also detailed process parameters and performance indicators, providing high-quality training data for machine learning.

[0030] The collected data undergoes rigorous cleaning and standardization to ensure accuracy and consistency. In particular, a unified standardized process has been established for data from different sources, including unit conversion, outlier removal, and handling of missing values.

[0031] Specifically, the following preprocessing steps are performed on the collected data: Outlier removal: For each continuous feature (such as acid value, amount added), Tukey's fences method (IQR rule) is used to treat values ​​less than or greater than a certain threshold as outliers and remove them.

[0032] Missing value handling: For numerical features with a missing rate of less than 5%, the K-nearest neighbor algorithm (K=5) was used for imputation; for records with a missing rate of more than 5% or missing non-numerical features, the record was directly removed. A total of 12 records were removed.

[0033] Standardization: All numerical input features are standardized using Z-score, so that their mean is 0 and their standard deviation is 1.

[0034] S3: Construct an XGBoost algorithm model with recipe parameters as input features and performance indicators as output labels.

[0035] Furthermore, the performance indicators include the acid value and precipitate content after aging.

[0036] Furthermore, the XGBoost model uses a Bayesian optimization method to determine the optimal combination of hyperparameters.

[0037] Furthermore, the XGBoost regression algorithm is used for model training. This algorithm has the following advantages: it can automatically handle nonlinear relationships between features; it provides feature importance ranking, which is helpful for mechanism analysis; and it supports incremental learning, which facilitates continuous model optimization.

[0038] S4: The dataset in the insulating oil formulation-performance relationship database is divided into training and testing sets. Cross-validation is used to train and test the XGBoost algorithm model in order to optimize the parameters of the XGBoost algorithm model.

[0039] Furthermore, the dataset was divided into training and test sets in a 7:3 ratio. 5-fold cross-validation was combined with Bayesian optimization based on the TPE algorithm (the optimization objective is to minimize the root mean square error (RMSE) of the validation fold) to search for the optimal combination of hyperparameters, including learning rate, maximum tree depth, and subsampling ratio.

[0040] S5: Evaluate the predictive power of the trained XGBoost algorithm model using an independent validation set, and apply the XGBoost algorithm model to predict the compound antioxidant components of the new formulation.

[0041] Furthermore, in step S5, applying the XGBoost algorithm model to predict the compound antioxidant components of the new formulation includes the following steps: S51: Set target performance indicators according to actual application requirements, including acid value range and precipitate content range; such as acid value ≤0.08mgKOH / g, precipitate ≤0.01%.

[0042] S52: Based on the XGBoost algorithm model, a search is performed in the full parameter space to find the formulation region that meets the target; the formulation region includes the proportion range of base oil, the moisture content of base oil, the initial acid value of base oil, and the corresponding pairing range of compound antioxidant ratio and addition amount.

[0043] S53: Select potential formulations from the formulation region for experimental verification, and feed the verification results back to the insulating oil formulation-performance relationship database for continuous optimization of the XGBoost algorithm model.

[0044] The trained XGBoost model is used as the objective function, and a constrained non-dominated sorting genetic algorithm (NSGA-II) is used for multi-objective inverse optimization search.

[0045] Optimization objectives: Minimize the predicted acid value and minimize the predicted precipitate content.

[0046] Decision variables: base oil mixing ratio, initial acid value, moisture content, BHT addition, PG addition, pretreatment temperature, stirring speed, and stirring time.

[0047] Constraints: Predicted acid value ≤ 0.08 mgKOH / g, predicted precipitate value ≤ 0.01%, and all variables within the preset physical range.

[0048] Output: After running the algorithm, a set of Pareto optimal solutions (approximately 15-30 non-dominated solutions) is obtained. This solution set is the "formulation region". Technicians can select specific formulations from this set based on cost and process preferences, such as choosing the formulation with the lowest predicted acid value or the lowest total cost of antioxidants.

[0049] Example 2 Preparation process: Starting from the pretreatment of base oil, through the formulation and addition of composite antioxidants, the modified insulating oil is finally obtained through post-treatment.

[0050] like Figure 2 The diagram shows a process flow chart for preparing a hybrid insulating oil based on composite antioxidants, including the following steps: T1: A new formulation of hybrid-based insulating oil is obtained by applying the above-described composite antioxidant optimization method.

[0051] From the Pareto solution set output by the NSGA-II algorithm above, a specific formulation that balances performance and cost is selected for preparation in this embodiment.

[0052] T2: Soybean oil and mineral oil are selected as base oils based on the base oil ratio range, base oil moisture content, and initial acid value of the base oil in the new formula.

[0053] Furthermore, the base oil ratio range is: mineral oil and soybean oil mixed in a ratio of 95:5 to 5:95.

[0054] Furthermore, the initial acid value of the base oil is less than 0.05 mg KOH / g.

[0055] Furthermore, the water content of the base oil is less than 50 ppm.

[0056] Base oil: No. 25 mineral insulating oil (compliant with GB 2536) and refined grade 1 soybean oil (compliant with GB / T 1535) are mixed at a mass ratio of 85:15.

[0057] Base oil specifications: initial acid value after mixing ≤0.03 mgKOH / g, moisture content ≤25 ppm.

[0058] T3: Place the base oil in a clean, heatable reactor and slowly heat it to 60-80°C. Maintain the temperature at 60-80°C and stir at 200-400 rpm for 20-40 minutes to reduce the viscosity of the base oil and remove dissolved gases, creating favorable conditions for the uniform dispersion of the antioxidants.

[0059] Process parameters: Pretreatment temperature 70℃, pretreatment stirring speed 300 rpm, pretreatment time 30 minutes; main mixing stirring speed 650 rpm, main mixing time 45 minutes; vacuum filtration temperature 60℃, vacuum degree -0.09 MPa.

[0060] Pretreatment temperature 70℃: Tests showed that at this temperature, the viscosity of the mixed base oil (85:15) dropped to about 125 cSt, which is 1 / 5 of the room temperature viscosity. This effectively promotes the escape of dissolved gases (mainly air), and its volatility reaches more than 95% within 30 minutes. Moreover, the FTIR spectrum of the oil sample showed no new oxidation characteristic peaks.

[0061] T4: In a dry environment, using an analytical balance with an accuracy of 0.1 mg, weigh the main antioxidant BHT and the auxiliary antioxidant PG according to the compound antioxidant ratio range of the new formulation; place the weighed powder in a mixer and mix thoroughly for 10-20 minutes at a speed of 20-40 rpm to obtain a highly uniform compound antioxidant premix.

[0062] Furthermore, the ratio of the primary antioxidant BHT (purity ≥99%) to the secondary antioxidant PG (purity ≥98%) is BHT:PG = 1:5-5:1.

[0063] Furthermore, the amount of the composite antioxidant added is 0.1%-1.0%.

[0064] Furthermore, the ratio and amount of the composite antioxidant were predicted using the XGBoost algorithm model.

[0065] Compound antioxidant: The main antioxidant is BHT (purity ≥99.5%), and the auxiliary antioxidant is PG (purity ≥98%), formulated at a mass ratio of 3:2.

[0066] Dosage: The total amount of compound antioxidant added is 0.55% of the mass of the mixed base oil.

[0067] T5: Under the protection of continuous high-purity nitrogen, slowly and in batches add 0.1%-1.0% of the composite antioxidant premix by weight of the base oil to the preheated and cooled base oil to 20-35℃; start mechanical stirring under nitrogen protection, control the speed at 500-800 rpm, and stir continuously for 30-60 minutes to ensure that the antioxidant molecules are fully dissolved and uniformly dispersed in the oil phase to form a stable homogeneous system.

[0068] The main mixing speed is 650 rpm: Based on the 50L frame stirred reactor used in this embodiment, calculations and preliminary experiments show that the stirring Reynolds number Re > 10 at this speed. 4 The flow field enters the fully turbulent region, and the volumetric mass transfer coefficient Kla reaches 0.015 s⁻¹, which can ensure that the antioxidant powder is uniformly dispersed throughout the can within 20 minutes (multiple sampling and detection, concentration variation coefficient CV < 5%).

[0069] The premixed compound antioxidant was divided into four equal portions. One portion was added at minutes 0, 5, 10, and 15 under continuous nitrogen purging and stirring. After each portion was added, a brief (approximately 10-15 seconds) localized turbidity was observed, which then rapidly disappeared under strong turbulence.

[0070] T6: The oil mixed with composite antioxidants is transferred to a vacuum filtration device and filtered at 50-70℃ and a vacuum degree of -0.08 to -0.10MPa to obtain a high oxidation stability insulating oil.

[0071] After vacuum filtration, take a small amount of oil sample into a colorimetric tube. Visually, it should be clear and transparent, without fluorescence or suspended matter. Use an online laser particle counter (in accordance with DL / T 432) to detect the particles. The number of particles larger than 5 μm should be less than 5000 per 100 mL (better than NAS grade 9).

[0072] Example 3 Evaluation process: The prepared insulating oil sample was placed in an accelerated thermal oxidation experiment. The acid value and the content of the final precipitate were analyzed by sampling at multiple time points to scientifically verify its oxidation stability.

[0073] Based on the above technical solution, the following prediction experiment is set up: Experimental design principle: Based on the importance analysis of model features, key factors (such as antioxidant type, ratio, and process parameters) are determined, and a BHT and PG composite system is adopted.

[0074] Experimental steps: Base oil pretreatment: Soybean oil and mineral oil are mixed, preheated and stirred.

[0075] Preparation of compound antioxidant: Accurately weigh BHT and PG, and mix them evenly.

[0076] Modified oil preparation: Add antioxidants under nitrogen protection and stir to dissolve.

[0077] Post-processing: Vacuum filtration yields the finished product.

[0078] Expected results: After 164 hours of accelerated aging, the acid value is controlled at 0.04-0.06 mgKOH / g, which is more than 40% higher than that of traditional formulas; the model prediction accuracy is R²≥0.95.

[0079] Using the aforementioned model, which showed performance of 0.927 for acid value prediction and 0.891 for precipitate prediction on the test set, the formulation "BHT:PG = 2:1, total addition 0.5%, base oil 70:30 mineral oil / soybean oil" was predicted to have an acid value of 0.052 mgKOH / g and a precipitate prediction of 0.006% after 164 hours of aging.

[0080] Example 4 To evaluate the model's generalization ability, multiple sets of experiments were designed: Experimental matrix: including groups such as 100% soybean oil, 70:30 mixed oil, and 50:50 mixed oil, with different antioxidant ratios and process parameters.

[0081] Experimental procedure summary: Same as Examples 2 and 3, with parameters adjusted according to model recommendations. A 164-hour accelerated thermal oxidation test was conducted, with sampling at multiple time points.

[0082] Expected findings: The model prediction error is less than 8%, confirming that the method in Example 1 is applicable to diversified formulations; at the same time, the proportion effect shows that there is greater room for optimization of mixed oils.

[0083] Example 5 Comparative Design: Benchmark Validation Experiment To highlight the advantages of the invention, a comparative example is designed: Comparative Example 1 (without antioxidant): The expected acid value increased to 0.40-0.50 mg KOH / g, which was used as the degradation baseline.

[0084] Comparative Example 2 (Single Antioxidant): PG or BHT added alone, with an expected acid value of 0.10-0.18 mgKOH / g, and performance lower than the composite system.

[0085] Comparative Example 3 (Non-optimal ratio): Using the model's non-recommended ratio, the expected acid value is 0.10-0.12 mgKOH / g, and the performance decreases by 30%.

[0086] Oxidation stability evaluation experiment The above embodiments and comparative samples were systematically evaluated according to the evaluation method provided in this application: (1) Catalyst preparation: The standard low carbon steel coil and copper coil are carefully polished with metallographic sandpaper until the surface is as bright as a mirror. They are then ultrasonically cleaned with acetone and anhydrous ethanol for 10 minutes each to thoroughly remove surface contaminants. They are then dried in a forced-air drying oven at 105°C for 1 hour and cooled to room temperature for later use. (2) Accelerated thermal oxidation experiment: 50.0 mL of the example sample and each comparative sample were measured into clean and dry oxidation tubes, and a treated low-carbon steel coil and a copper coil were placed in each tube. The oxidation tubes were precisely installed in a constant temperature oil bath stabilized at 100.0℃±0.2℃. The gas circuit system was connected, and the dry air flow rate was precisely adjusted to 1.02 L / h. Timing was started, and a standard accelerated aging experiment was conducted for 164 hours. (3) System sampling and testing: At 0h, 72h, 124h and 164h after the start of aging, approximately 5g of oil sample was accurately extracted from each oxidation tube using a dedicated clean glass syringe, and the acid value was immediately determined according to GB / T264 standard. After the 164-hour experiment, the precipitate content was determined according to the method specified in Appendix of GB / T12580-1990.

[0087] Results Analysis The verification results of this application significantly demonstrate its technical advantages. The composite insulating oil formulation developed under the guidance of a machine learning model effectively suppressed key performance indicators, particularly the increase in acid value and the formation of precipitates, after simulated accelerated oxidation experiments under long-term aging conditions. Its performance was significantly superior to the control system without added antioxidants or using only a single antioxidant. This not only verifies the potential synergistic effect of the composite antioxidants (BHT and PG) but also confirms that the prediction model based on the XGBoost algorithm can effectively capture the complex nonlinear relationship between formulation parameters and oxidation stability, providing a reliable basis for precise optimization. Further multi-scale verification experiments show that the model has good generalization ability and can provide effective optimization directions for systems with different base oil ratios, thus highlighting the potential improvement in efficiency and accuracy of the method of this invention compared to traditional trial-and-error methods.

[0088] In summary, this application successfully establishes a new paradigm that deeply integrates artificial intelligence with the research and development of insulation materials. By constructing a data-driven closed-loop system of "experiment-prediction-verification-optimization," it transforms traditional experience-dependent R&D into precise and efficient intelligent design. The method presented in this application not only provides an innovative solution to the key technical challenge of the oxidation stability of natural ester insulating oils, significantly improving the operational reliability and lifespan of power equipment, but its core machine learning model and system framework also possess broad adaptability, opening up new paths for formulation development and performance prediction in the entire field of insulation materials and even new materials, demonstrating significant engineering application value and promising prospects for widespread application.

[0089] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Transforming traditional trial-and-error experiments into data-driven R&D provides a new paradigm for optimizing insulating oil formulations, enabling faster identification of the globally optimal solution, shortening the R&D cycle, and reducing R&D costs.

[0090] 2. A new data-driven R&D model By systematically collecting experimental data, a comprehensive database of insulating oil formulations is constructed. This database not only contains basic formulation information but also detailed process parameters and performance indicators, providing high-quality training data for machine learning.

[0091] 3. Development of intelligent prediction models The XGBoost algorithm is used to construct a prediction model. This algorithm can effectively handle the complex nonlinear relationships between high-dimensional features and provide feature importance ranking, which helps to understand the influence mechanism of various factors on oxidation stability.

[0092] 4. Continuously optimized learning system A complete closed-loop system of "experiment-prediction-verification-optimization" was designed. New experimental data can continuously enrich the database, enabling the prediction model to continuously improve with the accumulation of data, forming a virtuous cycle.

[0093] This application discloses an optimization method for composite antioxidants and a preparation method for mixed-base insulating oil, relating to the field of insulating oil formulation optimization technology. The optimization method collects historical characteristic data of insulating oil formulations, establishes a database through cleaning and standardization, constructs and trains an XGBoost algorithm model to predict the components of new formulations. The preparation method, based on this optimization method, obtains new formulations, selects specific base oils, and performs pretreatment, composite antioxidant formulation, mixing and dissolution under nitrogen protection, and vacuum filtration to obtain insulating oil with high oxidation stability. This invention transforms traditional trial-and-error experiments into data-driven R&D, quickly finding the globally optimal solution, shortening the R&D cycle, reducing costs, effectively solving the problem of oxidation stability of insulating oil, and improving equipment operational reliability and lifespan.

[0094] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0095] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0096] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0097] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of this application.

Claims

1. A method for optimizing composite antioxidants, characterized in that, include: S1: Collect historical characteristic data of insulating oil formulation; S2: Clean and standardize the collected feature data, and establish an insulating oil formulation-performance relationship database; S3: Construct an XGBoost algorithm model with recipe parameters as input features and performance indicators as output labels; S4: The dataset in the insulating oil formulation-performance relationship database is divided into training and testing sets. Cross-validation is used to train and test the XGBoost algorithm model in order to optimize the parameters of the XGBoost algorithm model. S5: Evaluate the predictive power of the trained XGBoost algorithm model using an independent validation set, and apply the XGBoost algorithm model to predict the compound antioxidant components of the new formulation.

2. The method for optimizing composite antioxidants according to claim 1, characterized in that, In step S1, the historical characteristic data includes the base oil type, the mixing ratio of the base oil, the initial acid value of the base oil, the moisture content of the base oil, the type and ratio of antioxidants in the composite antioxidant, the amount of composite antioxidant added, the acid value of the insulating oil, and process parameters.

3. The method for optimizing composite antioxidants according to claim 2, characterized in that, The process parameters include pretreatment temperature, stirring speed, and time.

4. The method for optimizing composite antioxidants according to claim 1, characterized in that, The performance indicators include the acid value and precipitate content after aging.

5. The method for optimizing composite antioxidants according to claim 4, characterized in that, In step S5, applying the XGBoost algorithm model to predict the compound antioxidant components of the new formulation includes the following steps: S51: Set target performance indicators based on actual application requirements, including acid value range and precipitate content range; S52: Based on the XGBoost algorithm model, a search is performed in the full parameter space to find the formulation region that meets the target; the formulation region includes the proportion range of base oil, the moisture content of base oil, the initial acid value of base oil, and the corresponding range of the ratio of paired composite antioxidants. S53: Select potential formulations from the formulation region for experimental verification, and feed the verification results back to the insulating oil formulation-performance relationship database for continuous optimization of the XGBoost algorithm model.

6. A method for preparing a mixed-base insulating oil based on composite antioxidants, characterized in that, Includes the following steps: A new formulation of a mixed-based insulating oil is obtained by applying the composite antioxidant optimization method according to any one of claims 1 to 5; Soybean oil and mineral oil were selected as base oils based on the base oil ratio range, base oil moisture content, and initial acid value of the base oil in the new formula. Place the base oil in a clean, heatable reactor and slowly heat it to 60-80°C. Maintain the temperature at 60-80°C and stir at 200-400 rpm for 20-40 minutes to reduce the viscosity of the base oil, remove dissolved gases, and create favorable conditions for the uniform dispersion of the antioxidants. In a dry environment, using an analytical balance, weigh the main antioxidant BHT and the auxiliary antioxidant PG according to the compound antioxidant ratio range of the new formulation; place the weighed powder in a mixer and mix thoroughly for 10-20 minutes at a speed of 20-40 rpm to obtain the compound antioxidant premix. Under continuous high-purity nitrogen protection, the composite antioxidant premix is ​​slowly and in batches added to the base oil that has been preheated and cooled to 20-35°C; mechanical stirring is started under nitrogen protection, the speed is controlled at 500-800 rpm, and stirring is continued for 30-60 minutes to ensure that the antioxidant molecules are fully dissolved and uniformly dispersed in the oil phase to form a stable homogeneous system. The oil mixed with composite antioxidants was transferred to a vacuum filtration device and filtered at 50-70℃ and a vacuum degree of -0.08 to -0.10 MPa to obtain a high oxidation stability insulating oil.

7. The method for preparing a mixed-based insulating oil based on a composite antioxidant according to claim 6, characterized in that, The base oil has an initial acid value of less than 0.05 mg KOH / g.

8. The method for preparing a mixed-based insulating oil based on a composite antioxidant according to claim 6, characterized in that, The base oil has a water content of less than 50 ppm.

9. The method for preparing a mixed-based insulating oil based on a composite antioxidant according to claim 6, characterized in that, The ratio of the primary antioxidant BHT to the secondary antioxidant PG is BHT:PG = 1:5-5:

1.

10. The method for preparing a mixed-based insulating oil based on a composite antioxidant according to claim 6, characterized in that, The base oil ratio range is: mineral oil and soybean oil mixed in a ratio of 95:5 to 5:95.

Images