Self-adaptive optimization control system and method for rapeseed oil squeezing

Abstract

The invention discloses a self-adaptive optimization control system and method for rapeseed oil squeezing, and the method comprises the steps: obtaining rapeseed oil squeezing historical data, constructing a rapeseed oil squeezing reference library and a rapeseed oil squeezing prediction model, matching the rapeseed oil squeezing reference library to obtain an initial operation technology of an oil press, and carrying out the rapeseed oil squeezing to obtain an actual oil output index; obtaining a first predicted oil output index, obtaining an oil press bearing wear rate and an oil press state index according to the oil press operation current and the oil press operation state, correcting the first predicted oil output index to obtain a second predicted oil output index, and calculating an oil output index deviation; and selecting a process optimization strategy according to the oil outlet index deviation and the oil press bearing wear rate, and obtaining the optimized operation process of the oil press according to the process optimization strategy. The method not only can improve the efficiency and accuracy of rapeseed oil squeezing optimization control, but also has good interpretability, and can be directly applied to a rapeseed oil squeezing self-adaptive optimization control system.

Description

Technical Field

[0001] The present invention relates to the field of optimized control of oil presses, and particularly to an adaptive optimization control system and method for rapeseed oil pressing. Background Art

[0002] The efficiency and stability of rapeseed oil during the pressing process directly affect the oil quality, production cost, and equipment life. In the intelligent development of the food industry, adaptive optimization control technology can help optimize core indicators such as oil yield and energy consumption by real-time sensing the raw material characteristics and equipment status and dynamically adjusting process parameters. It can also extend the equipment operation cycle through predictive maintenance, which is of great significance for realizing green intelligent manufacturing.

[0003] Comparing with the many deficiencies of traditional rapeseed oil pressing technology: First of all, traditional pressing processes are mostly based on fixed experience and are difficult to adapt to the characteristics of rapeseed from different origins and varieties and the diverse market demands for oil quality. At the same time, the equipment status monitoring lags behind, mainly relying on manual inspections or off-line detections, making it difficult to real-time evaluate key parameters such as bearing wear and energy consumption efficiency, which easily leads to equipment failures or over-maintenance. In addition, the weak multi-objective optimization ability makes it difficult to dynamically balance among oil quality, energy consumption, and equipment stability, and the comprehensive performance decreases significantly after long-term operation. Therefore, the present invention proposes an adaptive optimization control system and method for rapeseed oil pressing. By obtaining the historical data of rapeseed oil pressing to construct a rapeseed oil pressing reference library and a prediction model, and combining real-time rapeseed information with the state data of the oil press, it realizes the dynamic adjustment and optimization of the operation process of the oil press, can effectively overcome the deficiencies of the existing technology, realize the intelligentization of rapeseed oil pressing, improve the oil yield and oil quality, and can also adjust the operation parameters in real time according to the equipment status, providing a new path to solve the long-existing contradiction between efficiency and stability in the oil pressing industry, and having important application value for promoting the intelligent upgrading of the industry. Summary of the Invention

[0004] The purpose of the present invention is to provide an adaptive optimization control system and method for rapeseed oil pressing.

[0005] To achieve the above purpose, the present invention is implemented according to the following technical solutions:

[0006] The present invention includes the following steps:

[0007] Obtain the historical data of rapeseed oil pressing, and construct a rapeseed oil pressing reference library and a rapeseed oil pressing prediction model according to the historical data of rapeseed oil pressing;

[0008] Obtain rapeseed information, and match the rapeseed information with the rapeseed oil pressing reference library to obtain the initial operation process of the oil press;

[0009] According to the initial operation process of the oil press, rapeseeds to be pressed are used to extract oil to obtain the actual oil output index and the operating current of the oil press. Information on the rapeseeds to be pressed and the initial operation process of the oil press are input into the rapeseed oil pressing prediction model to obtain the first predicted oil output index;

[0010] The bearing wear rate of the oil press is obtained by analyzing the operating current of the oil press. The operating state of the oil press and the bearing wear rate of the oil press are input into the state function to obtain the state index of the oil press. The first predicted oil output index is corrected according to the state index of the oil press to obtain the second predicted oil output index;

[0011] The deviation of the oil output index is calculated, and a process optimization strategy is selected according to the deviation of the oil output index and the bearing wear rate of the oil press; the process optimization strategy includes a first process optimization strategy and a second process optimization strategy.

[0012] Further, the method for constructing the rapeseed oil pressing reference library and the rapeseed oil pressing prediction model includes:

[0013] Historical data on rapeseed oil pressing is obtained. Decision trees are used to extract features and classify features from the historical data on rapeseed oil pressing to obtain historical oil pressing features, and timestamps are set according to time information; the historical oil pressing features include historical rapeseed features, the historical operation process of the oil press, and historical oil output features;

[0014] The historical rapeseed features and the historical operation process of the oil press are associated according to the timestamp and form the rapeseed oil pressing reference library;

[0015] Random forests are used to divide the historical oil pressing features into a training set and a test set in a ratio of 8:2, and a rapeseed oil pressing prediction model is constructed; the rapeseed oil pressing prediction model includes an input layer, a hidden layer, and an output layer;

[0016] The hidden layer uses a fully connected neural network to capture the data dependence relationships among the historical rapeseed features, the historical operation process of the oil press, and the historical oil output features, and predicts the oil output features according to the rapeseed features and the operation process of the oil press; the output layer uses two parallel fully connected layers to output continuous oil output features and discrete oil output features respectively; the continuous oil output features specifically refer to the oil yield, oil output speed, oil temperature, impurity rate, and oil residue density; the discrete oil output features specifically refer to the oil body color grade;

[0017] The mean squared error is used as the loss function to measure the difference between the predicted value and the true value of the continuous oil output features, the cross-entropy loss function is used to measure the difference between the predicted probability distribution and the true label distribution of the discrete oil output features, the Adam optimizer is used to adjust the model learning rate, and the test set is used to evaluate the model performance and input it into the rapeseed oil pressing prediction model.

[0018] Further, the method for obtaining the initial operation process of the oil press includes:

[0019] Extract the characteristics F of the rapeseed to be pressed obtained from the rapeseed information characteristics, where F = {f 1 , f 2 , …, f n}, n is the number of characteristics. Calculate the comprehensive similarity between the characteristics of the rapeseed to be pressed and the historical rapeseed characteristics H j = {h j1 , h j2 , …, h jn} in the rapeseed oil pressing reference library. Extract the timestamp of the historical rapeseed characteristics with the highest similarity, and select the corresponding historical operating process of the oil press as the initial operating process of the oil press according to the timestamp. The calculation of the comprehensive similarity is expressed as:

[0020]

[0021] Among them, S(F, H j ) is the comprehensive similarity between the feature vectors F and H j , α is the control factor, c is the cross number constant, n(F ∩ H j ) is the number of intersections of the two feature vectors, n is the number of characteristics of the feature vector, t far is the farthest recorded timestamp in the rapeseed oil pressing reference library, t near is the nearest recorded timestamp in the rapeseed oil pressing reference library, t j is the recorded timestamp of the jth group of historical rapeseed characteristics, S cov (F, H j ) is the cosine similarity between the feature vectors F and H j , and S d (F, H j ) is the Euclidean distance between the feature vectors F and H j .

[0022] Furthermore, the method for obtaining the second predicted oil yield index includes:

[0023] Perform Fourier transform on the operating current of the oil press to obtain the current spectrum, extract the harmonic characteristics of the bearing fault characteristic frequency, and input the harmonic characteristics into the harmonic characteristic - wear rate mapping model to obtain the bearing wear rate of the oil press; the harmonic characteristics include current harmonic energy entropy, spectrum energy entropy, harmonic amplitude, harmonic coupling coefficient, harmonic distortion rate, and load speed parameter;

[0024] The steps to obtain the harmonic feature - wear rate mapping model are as follows: Simulate the progressive wear process of the oil press bearing through an accelerated wear test bench to obtain current signals at different wear stages, process the current signals at different wear stages to obtain harmonic features corresponding to the bearing fault characteristic frequencies, and construct a non - linear harmonic feature - wear rate mapping model based on the harmonic features and the actual wear rate; The non - linear harmonic feature - wear rate mapping model includes a gradient boosting tree, Bayesian optimization, and a mean squared error loss function; The gradient boosting tree learns the data relationship between harmonic features and wear rates, and predicts the wear rate of the oil press bearing according to the input harmonic features; Bayesian optimization optimizes the hyperparameters of the model; The mean squared error loss function evaluates the difference between the predicted wear rate value and the actual value;

[0025] Input the operating state of the oil press and the wear rate of the oil press bearing into the state function to obtain the oil press state index, and the expression is:

[0026]

[0027] where E(t) is the oil press state index at time t, t last is the time since the last maintenance, β is the maintenance attenuation coefficient, w 1 is the deterioration weight of the working duration, t act is the cumulative working time, t sta is the designed standard life, γ is the fatigue cumulative coefficient, w 2 is the deterioration weight of the working intensity, P cur is the current working pressure, P max is the rated maximum pressure, w 3 is the deterioration weight of the temperature, k is the temperature sensitivity coefficient, T act is the current temperature, T safe is the safe temperature, w 4 is the deterioration weight of the vibration, m is the number of vibration monitoring points, v i is the effective value of the vibration acceleration at the monitoring point i, v sta is the vibration threshold, w 5 is the deterioration weight of the wear, a cur is the current wear rate, a fai is the critical wear rate of failure, is the change trend of the wear rate within the time interval Δt

[0028] Modify the first predicted oil output index according to the oil press state index to obtain the second predicted oil output index.

[0029] Furthermore, the method for carrying out equipment operation early warning includes:

[0030] Calculate the oil output index deviation based on the second predicted oil output index and the actual oil output index. When the oil output index deviation and the wear rate of the oil press bearing are greater than the corresponding warning thresholds, implement the first process optimization strategy; the first process optimization strategy includes equipment operation warning and adjustment of the corresponding operation process.

[0031] The equipment operation warning specifically includes: a. Conduct process warning when any oil output index deviation is greater than the corresponding warning threshold; b. Conduct structural warning when the wear rate of the oil press bearing is greater than the corresponding warning threshold, and conduct wear analysis and maintain the oil press bearing.

[0032] Furthermore, the method for determining the pressing objective function includes:

[0033] Determine the oil output index weight δ according to the oil output target j , according to the oil output index weight δ j and the oil output index deviation Δx j Determine the pressing objective function Aim, and the expression is:

[0034]

[0035] where is the oil output index deviation weight, is the energy efficiency weight, is the dynamic wear weight, is the stability weight, Δx j =x j -x j,real , x j is the second predicted oil output index, x j,real is the actual oil output index, P total is the total power corresponding to the current oil press operation process, Q oil is the oil output per unit time corresponding to the current oil press operation process, P rated is the standard rated power, ξ is the overload penalty coefficient, t run is the predicted operation time of the oil press this time, a t is the wear rate of the oil press at time t, θ 1 、θ 2 、θ 3 are dynamic adjustment coefficients, is the variance of vibration acceleration, is the variance of the temperature gradient in the pressing chamber, MTBF is the historical sliding window statistical value of the mean time between failures, and the dynamic update strategy of the wear weight is is the wear weight updated at time t + 1, a cur is the current wear rate, a fai is the critical wear rate for failure, and the wear weight update frequency is 30 min / time.

[0036] Further, the method for optimizing the operation process of the oil press by dynamic game optimization output includes:

[0037] According to the pressing objective function, the number of players in the dynamic game is determined to be 4, and the corresponding player set is P = {p 1 , p 2 , p 3 , p 4}, where p 1 is the deviation of the oil output index, p 2 is the pressing energy efficiency, p 3 is the machine wear degree, p 4 is the machine stability. The strategy space S = S 1 ∪ S 2 ∪ S 3 ∪ S 4 corresponding to the dynamic game of the players is determined, where S 1 , S 2 , S 3 , S 4 are the strategy sets of the corresponding indexed players respectively. The players execute the corresponding determined payoff functions to obtain the payoff function set U = {u 1 , u 2 , u 3 , u 4}, where u 1 , u 2 , u 3 , u 4 are the payoff functions of the corresponding indexed players, u 1 = 1 / minR 1 , u 2 = 1 / minR 2 , u 3 = 1 / minR 3 , u 4 = 1 / minR 4 . A dynamic game G = {P; S; U} of the oil press operation process is formed;

[0038] Input the dynamic game tree parameter G = {P; S; U}, use MapReduce to split the game tree into K sub-game forests, construct the payoff matrix of the sub-games and calculate the number of stages y max , and perform pre-pruning on the game trees with low payoffs using the dynamic benefit threshold. The pruning condition is:

[0039]

[0040] where is the payoff value of the i-th subtree at the game tree node k, and ε ∈ [0.1, 0.3,] is the payoff value threshold coefficient. It is the maximum payoff value of all subtrees at the game tree node k;

[0041] Use the pre-trained multi-agent Q-network to predict the Q-value scores of each strategy, and select the strategy with the highest score as the sub-game equilibrium path; the multi-agent Q-network predicts the Q-value scores of each strategy according to the historical strategy distribution, the sparsity of the payoff matrix, and the node depth; the Q-value scores are used to guide the Monte Carlo tree search direction;

[0042] Process each sub-game block iteratively backward from the deepest stage to the initial stage of the sub-game equilibrium path, calculate the confidence of the strategy for each decision node respectively, mark the strategy with a confidence greater than 0.9 as the dominant strategy and lock the path, and use the Monte Carlo tree search to correct the strategy scores for non-dominant strategies;

[0043] Merge the local game paths of all sub-game blocks according to the game structure compression technology to generate an intermediate result of the global strategy combination; the game structure compression technology includes sub-game equivalence class merging and Huffman coding compression; the sub-game equivalence class merging identifies structurally similar sub-games through KL divergence and merges and calculates repeated strategy sequences; the Huffman coding compression encodes high-frequency strategy patterns to reduce the storage space occupancy of the strategy sequence;

[0044] Verify the payoff distribution and squeezing objective function across sub-games based on the Shapley value allocation model, output the global game strategy combination and the confidence evaluation report, output the optimized operation process of the oil press according to the global game strategy combination, and store the optimized operation process of the oil press and the corresponding rapeseed characteristics with timestamps in the rapeseed oil pressing reference library.

[0045] In a second aspect, an adaptive optimization control system for rapeseed oil pressing includes:

[0046] Initial operation process module: used to obtain the rapeseed oil pressing historical data, construct a rapeseed oil pressing reference library according to the rapeseed oil pressing historical data, and match the rapeseed oil pressing reference library according to the rapeseed information to be pressed to obtain the initial operation process of the oil press;

[0047] Prediction model module: construct a rapeseed oil pressing prediction model according to the rapeseed oil pressing historical data, and input the rapeseed information to be pressed and the initial operation process of the oil press into the rapeseed oil pressing prediction model to obtain the first predicted oil output index;

[0048] Correction module: used to analyze the operating current of the oil press to obtain the bearing wear rate of the oil press, input the operating state of the oil press and the bearing wear rate of the oil press into the state function to obtain the state index of the oil press, and correct the first predicted oil output index according to the state index of the oil press to obtain the second predicted oil output index;

[0049] The operation process optimization module: It is used to calculate the deviation of the oil output index, determine the weight of the oil output index, determine the pressing objective function according to the weight of the oil output index and the deviation of the oil output index, and perform dynamic game optimization on the initial operation process of the oil press according to the pressing objective function to output the optimized operation process of the oil press;

[0050] The management platform module: It is used to view, manage and store the historical data of rapeseed oil pressing, the deviation of the oil output index, the wear rate of the oil press bearing and the optimized operation process of the oil press, and adjust the operation process of the oil press according to the optimized operation process of the oil press.

[0051] The beneficial effects of the present invention are:

[0052] The present invention is an adaptive optimization control system and method for rapeseed oil pressing. Compared with the prior art, the present invention has the following technical effects:

[0053] By constructing steps of prediction model, data matching, index correction, process optimization and strategy determination, the present invention can improve the data preprocessing ability and enhance the model adaptability in the adaptive optimization control of rapeseed oil pressing, thereby improving the efficiency and accuracy of the adaptive optimization control of rapeseed oil pressing, optimizing the adaptive optimization control technology of rapeseed oil pressing, greatly saving resources, improving work efficiency, realizing the efficient optimization of the operation process of rapeseed oil pressing, providing more reliable technical support for the adaptive optimization control of rapeseed oil pressing, helping to extend the service life of equipment, reduce maintenance costs, ensure the stability and consistency of oil quality, promote the technological progress and industrial upgrading of the industry, and can meet the requirements of the adaptive optimization control system for different rapeseed oil pressings and the adaptive optimization control of rapeseed oil pressing for different users, having a certain universality. Brief Description of the Drawings

[0054] Figure 1 It is the step flow chart of an adaptive optimization control method for rapeseed oil pressing of the present invention. Detailed Embodiments

[0055] The present invention will be further described below through specific embodiments. The illustrative embodiments and explanations of the present invention are used to explain the present invention, but do not limit the present invention.

[0056] An adaptive optimization control system and method for rapeseed oil pressing of the present invention includes the following steps:

[0057] As Figure 1 shown, in this embodiment, it includes the following steps:

[0058] Obtain the historical data of rapeseed oil pressing, and construct a rapeseed oil pressing reference library and a rapeseed oil pressing prediction model according to the historical data of rapeseed oil pressing;

[0059] Obtain rapeseed information, and match the rapeseed oil pressing reference library according to the rapeseed information to obtain the initial operating process of the oil press;

[0060] Press the rapeseed to be pressed according to the initial operating process of the oil press to obtain the actual oil output index and the operating current of the oil press, and input the rapeseed information to be pressed and the initial operating process of the oil press into the rapeseed oil pressing prediction model to obtain the first predicted oil output index;

[0061] Analyze the operating current of the oil press to obtain the bearing wear rate of the oil press, input the operating state of the oil press and the bearing wear rate of the oil press into the state function to obtain the state index of the oil press, and correct the first predicted oil output index according to the state index of the oil press to obtain the second predicted oil output index;

[0062] Calculate the deviation of the oil output index, and select a process optimization strategy according to the deviation of the oil output index and the bearing wear rate of the oil press; the process optimization strategy includes a first process optimization strategy and a second process optimization strategy.

[0063] In this embodiment, the method for constructing the rapeseed oil pressing reference library and the rapeseed oil pressing prediction model includes:

[0064] Obtain the historical data of rapeseed oil pressing, use a decision tree to perform feature extraction and feature classification on the historical data of rapeseed oil pressing to obtain historical oil pressing features, and set a time stamp according to the time information; the historical oil pressing features include historical rapeseed features, the historical operating process of the oil press, and historical oil output features;

[0065] Associate the historical rapeseed features and the historical operating process of the oil press according to the time stamp and form a rapeseed oil pressing reference library;

[0066] Use a random forest to divide the historical oil pressing features into a training set and a test set according to an 8:2 ratio, and construct a rapeseed oil pressing prediction model; the rapeseed oil pressing prediction model includes an input layer, a hidden layer, and an output layer;

[0067] The hidden layer uses a fully connected neural network to capture the data dependence relationship of historical rapeseed features, the historical operating process of the oil press, and historical oil output features, and predicts the oil output features according to the rapeseed features and the operating process of the oil press; the output layer uses two parallel fully connected layers to output continuous oil output features and discrete oil output features respectively; the continuous oil output features specifically refer to the oil yield, oil output speed, oil temperature, impurity rate, and oil residue density; the discrete oil output features specifically refer to the oil body color grade;

[0068] The mean squared error is used as the loss function to measure the difference between the predicted value and the true value of the continuous oil output characteristics, the cross-entropy loss function is used to measure the difference between the predicted probability distribution and the true label distribution of the discrete oil output characteristics, the Adam optimizer is used to adjust the model learning rate, and the test set is used to evaluate the model performance and output the rapeseed oil pressing prediction model;

[0069] In the actual evaluation, the rapeseed characteristics specifically include oil content, moisture, particle size, variety and origin, and the operation process of the oil press specifically includes heating temperature, pressing force, feeding speed and rotation speed;

[0070] The random forest is used to divide the training set (800 pieces) and the test set (200 pieces). The input layer is set with 10 nodes, the hidden layer is set with 2 layers / 16 nodes per layer / ReLU activation function, the output layer is 5 nodes for continuous type / 3 nodes for discrete type, the mean squared error in the loss function is taken as 0.5, the cross-entropy is taken as 0.3, and the training result is that the prediction error of the oil yield is ±1.2%, and the prediction accuracy of the color grade is 90%. The rapeseed oil pressing prediction model is output.

[0071] In this embodiment, the method for obtaining the initial operation process of the oil press includes:

[0072] Extract the rapeseed characteristics to be pressed F = {f 1 , f 2 , …, f n}, where n is the number of characteristics, calculate the comprehensive similarity between the rapeseed characteristics to be pressed and the j-th group of historical rapeseed characteristics H j = {h j1 , h j2 , …, h jn} in the rapeseed oil pressing reference library, extract the timestamp of the historical rapeseed characteristics with the highest similarity, and select the corresponding historical operation process of the oil press as the initial operation process of the oil press according to the timestamp. The calculation of the comprehensive similarity is expressed as:

[0073]

[0074] Where S(F, H j ) is the comprehensive similarity between the feature vectors F and H j , α is the control factor, c is the cross number constant, n(F ∩ H j ) is the number of intersections of the two feature vectors, n is the number of features of the feature vector, t far is the farthest record timestamp in the rapeseed oil pressing reference library, t near is the nearest record timestamp in the rapeseed oil pressing reference library, t j is the record timestamp of the j-th group of historical rapeseed characteristics, and S cov (F, H j ) is the comprehensive similarity between the feature vectors F and H jCosine similarity, S d (F, H j ) is the Euclidean distance between feature vectors F and H j ;

[0075] In an actual evaluation, taking a rapeseed oil pressing factory processing rapeseeds produced in Hubei as an example, the rapeseed information obtained is oil content 20.5%, moisture 9.8%, particle diameter 2.1 mm, double-low rapeseed (variety library number 62), and production area coordinates (30°, 115°). The corresponding rapeseed feature vector is [20.5%, 9.8%, 2.1, 62, 30°, 115°]. Taking two sets of historical rapeseed data as an example: 1. Feature vectors [20.3%, 10.1%, 2.0, 59, 29°31’, 112°12’], [21%, 9.5%, 2.0, 46, 32°37’, 110°48’], 2. Timestamps 2023-06-15, 2021-05-20;

[0076] The farthest recorded timestamps and the farthest recorded timestamps in the rapeseed oil pressing reference library are 2020-01-01 and 2024-12 respectively. The control factor α is taken as 5, and the cross number constant c is taken as 1.2. The comprehensive similarity between the rapeseed features to be pressed and the historical rapeseed features in the rapeseed oil pressing reference library is calculated to be 0.9225 and 0.6383;

[0077] The highest comprehensive similarity between the rapeseed features to be pressed and the historical rapeseed features is 0.9225. The timestamp 2023-06-15 of this group of historical rapeseed features is extracted, and the corresponding operating process is obtained as the initial operating process: heating temperature 82°C, pressing force 30 MPa, feeding speed 60 kg / h, rotation speed 500 rpm;

[0078] Input the rapeseed features to be pressed and the initial operating process into the rapeseed oil pressing prediction model to obtain the first predicted oil output indicators: oil yield 40.2%, oil output speed 145 kg / h, oil temperature 61°C, impurity rate 0.6%, oil residue density 425 kg / m 3 、brightness 68, RGB(235, 215, 170).

[0079] In this embodiment, the method for obtaining the second predicted oil output indicators includes:

[0080] Perform Fourier transform on the operating current of the oil press to obtain the current frequency spectrum, extract the harmonic features of the bearing fault characteristic frequency, and input the harmonic features into the harmonic feature-wear rate mapping model to obtain the bearing wear rate of the oil press; The harmonic features include current harmonic energy entropy, spectrum energy entropy, harmonic amplitude, harmonic coupling coefficient, harmonic distortion rate, and load rotation speed parameters;

[0081] The steps for obtaining the harmonic feature - wear rate mapping model are as follows: Simulate the progressive wear process of the oil press bearing through an accelerated wear test bench to obtain current signals at different wear stages. Process the current signals at different wear stages to obtain harmonic features corresponding to the bearing fault characteristic frequencies. Construct a non - linear harmonic feature - wear rate mapping model based on the harmonic features and the actual wear rate. The non - linear harmonic feature - wear rate mapping model includes a gradient boosting tree, Bayesian optimization, and a mean squared error loss function. The gradient boosting tree learns the data relationship between harmonic features and wear rates, and predicts the wear rate of the oil press bearing according to the input harmonic features. Bayesian optimization optimizes the hyperparameters of the model. The mean squared error loss function evaluates the difference between the predicted wear rate value and the actual value.

[0082] Input the operating state of the oil press and the wear rate of the oil press bearing into the state function to obtain the oil press state index. The expression is:

[0083]

[0084] where E(t) is the oil press state index at time t, t last is the time since the last maintenance, β is the maintenance attenuation coefficient, w 1 is the deterioration weight of the working duration, t act is the cumulative working time, t sta is the designed standard life, γ is the fatigue cumulative coefficient, w 2 is the deterioration weight of the working intensity, P cur is the current working pressure, P max is the rated maximum pressure, w 3 is the deterioration weight of the temperature, k is the temperature sensitivity coefficient, T act is the current temperature, T safe is the safety temperature, w 4 is the deterioration weight of the vibration, m is the number of vibration monitoring points, v i is the effective value of the vibration acceleration at the monitoring point i, v sta is the vibration threshold, w 5 is the deterioration weight of the wear, a cur is the current wear rate, a fai is the failure - critical wear rate, is the change trend of the wear rate within the time interval Δt

[0085] Modify the first predicted oil output index according to the oil press state index to obtain the second predicted oil output index.

[0086] In the actual evaluation, according to the initial operation process of the oil press, the rapeseed to be pressed is used to obtain the state data of the oil press, and the harmonic characteristics are obtained by processing the running current signal of the oil press: the harmonic characteristics include the current harmonic energy entropy of 0.75, the spectral energy entropy of 0.68, the harmonic amplitudes of 5th / 7th: 3.2 A / 2.8 A, the harmonic coupling coefficient of 0.45, the harmonic distortion rate of 12%, and the load rotation speed parameter of 480 rpm. The harmonic characteristics are input into the harmonic characteristic-wear rate mapping model to obtain a predicted wear rate of 0.08;

[0087] Take the overhaul attenuation coefficient β as 0.001, the working duration deterioration weight w 1 as 0.2, the working intensity deterioration weight w 2 as 0.3, the temperature deterioration weight w 3 as 0.15, the vibration deterioration weight w 4 as 0.2, the wear deterioration weight w 5 as 0.15, the fatigue accumulation coefficient γ as 0.001, the temperature sensitivity coefficient k as 0.5,

[0088] When the time t last since the last overhaul is 300 h, the design standard life t sta is 8000 h, the cumulative working time t act is 5000 h, the current working pressure P cur is 22 MPa, the rated maximum pressure P max is 30 MPa, the number of vibration monitoring points m is 3, the vibration threshold v sta is 2.5 m / s 2 、the failure critical wear rate a fai is 0.5, the current wear rate a cur is 0.08, calculate the state index of the oil press as 0.1972;

[0089] According to the state index of the oil press, the first predicted oil output index is corrected to obtain the second predicted oil output index: the oil yield is 37.8%, the oil output speed is 134 kg / h, the oil temperature is 61 °C, the impurity rate is 0.6%, the oil residue density is 428 kg / m 3 、the brightness is 71, RGB(229, 210, 180).

[0090] In this embodiment, the method for carrying out equipment operation warning includes:

[0091] Calculate the deviation of the oil output index according to the second predicted oil output index and the actual oil output index. When the deviation of the oil output index and the bearing wear rate of the oil press are greater than the corresponding warning thresholds, implement the first process optimization strategy; the first process optimization strategy includes equipment operation warning and adjustment of the corresponding operation process;

[0092] The early warning of the equipment operation specifically includes: a. Conducting process early warning when any oil output index deviation is greater than the corresponding early warning threshold; b. Conducting structural early warning when the wear rate of the oil press bearing is greater than the corresponding early warning threshold, and conducting wear analysis and maintaining the oil press bearing.

[0093] In the actual evaluation, the wear analysis includes checking dynamic load impact, checking lubrication condition, checking vibration fatigue, checking rapeseed wear, and checking assembly error.

[0094] According to the initial operation process of the oil press, the rapeseed to be pressed is processed to obtain the actual oil output indexes: oil yield 39.5%, oil output speed 140 kg / h, oil temperature 60 °C, impurity rate 0.72%, oil residue density 430 kg / m 3 , brightness 64, RGB(2331, 204, 182); corresponding oil output index deviations are 1.7%, 6 kg / h, 1 °C, 0.12%, 2 kg / m 3 , where the impurity rate deviation is greater than the impurity rate deviation threshold of 0.1%, the brightness deviation is greater than the brightness deviation threshold of 5, the wear rate of 0.08 is less than the wear rate threshold. Implement the first process optimization strategy (conducting early warning of equipment operation, adjusting the feeding speed and heating temperature according to the impurity deviation and brightness deviation, and checking the filtering device). After implementing the first process optimization strategy, recalculate the oil output index deviation. Without early warning, enter the second process optimization strategy.

[0095] In this embodiment, the method for determining the pressing target function includes:

[0096] Determining the oil output index weight δ according to the oil output target j , according to the oil output index weight δ j and the oil output index deviation Δx j to determine the pressing target function Aim, and the expression is:

[0097]

[0098]

[0099] where is the oil output index deviation weight, is the energy efficiency weight, is the dynamic wear weight, is the stability weight, Δx j = x j - x j,real , x j is the second predicted oil output index, x j,real is the actual oil output index, P total is the total power corresponding to the current operation process of the oil press, Q oilis the oil output per unit time corresponding to the current operation process of the oil press, P rated is the standard rated power, ξ is the overload penalty coefficient, t run is the expected operation time of the oil press this time, a t is the wear rate of the oil press at time t, θ 1 , θ 2 , θ 3 is the dynamic adjustment coefficient, is the variance of vibration acceleration, is the variance of the temperature gradient in the pressing chamber, MTBF is the historical sliding window statistical value of the mean time between failures, and the dynamic update strategy of the wear weight is is the wear weight updated at time t + 1, a cur is the current wear rate, a fai is the critical wear rate of failure, and the wear weight update frequency is 30 min / time;

[0100] In the actual evaluation, the parameter constraint conditions corresponding to the pressing objective function are:

[0101] 0.8η 0 ≤ η oil ≤ 1.2η 0

[0102] T oil ∈ [T min , T max

[0103] v i ≤ 4.5 m / s 2

[0104] where η oil is the current oil output rate, η 0 is the standard oil output rate, taking 20%, T oil is the current oil output temperature, T min , T max are the rated minimum and maximum oil output temperatures, taking 50°C and 85°C, v i is the vibration acceleration of monitoring point i.

[0105] In this embodiment, the method for performing dynamic game optimization to output the optimized operation process of the oil press includes:

[0106] Determine that the number of players in the dynamic game is 4 according to the pressing objective function, and the corresponding player set is P = {p 1 , p 2 , p 3 , p 4}, p 1 is the deviation of the oil output index, p 2 is the pressing energy efficiency, p 3 ​is the machine wear degree, p 4 is the machine stability. Determine the strategy space S = S 1 ∪S 2 ∪S 3 ∪S 4 , S 1 , S 2 , S 3 , S 4 are the strategy sets of the corresponding indexed players respectively. The players execute the corresponding ones to obtain the payoff function set U = {u 1 , u 2 , u 3 , u 4}, where u 1 , u 2 , u 3 , u 4 are the payoff functions of the corresponding indexed players, u 1 = 1 / minR 1 , u 2 = 1 / minR 2 , u 3 = 1 / minR 3 , u 4 = 1 / minR 4 , constituting the dynamic game G = {P; S; U} of the oil press operation process;

[0107] Input the dynamic game tree parameter G = {P; S; U}, use MapReduce to split the game tree into K sub-game forests, construct the payoff matrix of the sub-games and calculate the number of stages y max , and perform pre-pruning on the game trees with low payoffs using the dynamic benefit threshold. The pruning condition is:

[0108]

[0109] where is the payoff value of the i-th subtree at the game tree node k, ε ∈ [0.1, 0.3,] is the payoff value threshold coefficient, is the maximum payoff value of all subtrees at the game tree node k;

[0110] Use the pre-trained multi-agent Q-network to predict the Q-value scores of each strategy, and select the strategy with the highest score as the sub-game equilibrium path; The multi-agent Q-network predicts the Q-value scores of each strategy according to the historical strategy distribution, the sparsity of the payoff matrix and the node depth; The Q-value scores are used to guide the Monte Carlo tree search direction;

[0111] Process each sub-game block by backward iteration from the deepest stage to the initial stage according to the sub-game equilibrium path. Calculate the confidence of the strategies for each decision node respectively. Mark the strategies with a confidence greater than 0.9 as dominant strategies and lock the path. Use Monte Carlo tree search to correct the strategy scores for non-dominant strategies;

[0112] Merge the local game paths of all sub-game blocks according to the game structure compression technology to generate an intermediate result of the global strategy combination; the game structure compression technology includes sub-game equivalence class merging and Huffman coding compression; the sub-game equivalence class merging identifies structurally similar sub-games through KL divergence and merges and calculates repeated strategy sequences; the Huffman coding compression encodes high-frequency strategy patterns to reduce the storage space occupied by the strategy sequences;

[0113] Verify the payoff distribution and squeezing objective function across sub-games based on the Shapley value allocation model, output the global game strategy combination and the confidence evaluation report, output the optimized operation process of the oil press according to the global game strategy combination, and store the optimized operation process of the oil press and the corresponding rapeseed characteristics with timestamps in the rapeseed oil pressing reference library;

[0114] In the actual evaluation, the initial strategies are a heating temperature of 82 °C, a pressing force of 30 MPa, a feeding speed of 60 kg / h, and a rotation speed of 500 rpm, and the calculated pressing objective function is 1.22; it is defined that in two consecutive iterations: the adjustment range of the heating temperature < 1 °C, the adjustment range of the pressing pressure < 0.5 MPa, and the adjustment range of the rotation speed < 10 rpm, then terminate the search in the direction of this parameter; it is defined that after 3 consecutive iterations, the improvement range of the objective function value < 0.01, terminate the current strategy iteration;

[0115] The strategy sets \(S\) of the players 1 、\(S\) 2 、\(S\) 3 、\(S\) 4 are \(\{\pm1\ MPa\ pressure\ adjustment,\ \pm2\ ^{\circ}C\ temperature\ adjustment\}\), \(\{5\ kg / h\ feeding\ speed\ reduction,\ \pm10\ rpm\ rotation\ speed\ adjustment\}\), \(\{lubrication\ compensation,\ load\ dynamic\ balance\}\), \(\{vibration\ suppression,\ temperature\ gradient\ control\}\) respectively; input the dynamic game tree parameters \(G = \{P; S; U\}\), the initial game tree contains 200 nodes, split into \(K = 4\) sub-game forests, each sub-game block contains 50 nodes, take the payoff value threshold coefficient as 0.2, the maximum payoff value of all sub-trees is 1.44, eliminate the sub-trees with a payoff value less than 0.288, and after pruning operation, the pruning rate exceeds 35%;

[0116] Taking the first iteration optimization as an example: Use the pre-trained multi-agent Q-network to predict the Q-value scores of each strategy, and select the strategy \(S\) corresponding to the highest score of 0.92 1={Adjusting pressure - 3MPa} serves as the sub - game equilibrium path to guide the Monte Carlo tree search direction, and the Monte Carlo tree search is used to correct the non - dominated strategy S 2 ={Adjusting the feed rate - 5kg / h} obtains a strategy score of 0.87 and is still judged as a positive non - dominated strategy. According to the game structure compression technology, the local game paths of all sub - game blocks are merged to generate an intermediate result of the global strategy combination. The corresponding strategy is a heating temperature of 80°C, a pressing force of 28MPa, a feed rate of 60kg / h, a rotational speed of 480rpm, load dynamic balance, and temperature gradient control. The pressing objective function is calculated as 1.12;

[0117] When iterating to the 15th time, the decrease amplitude of the objective function value is less than 0.01 for three consecutive iterations, and the strategy iteration is stopped. The corresponding strategy is a heating temperature of 77°C, a pressing force of 28MPa, a feed rate of 60kg / h, a rotational speed of 450rpm, and temperature gradient control. The pressing objective function is calculated as 0.65. The Shapley value allocation model is used to verify that the income distribution across sub - games meets the consistency requirement. The optimized operation process of the oil press (heating temperature of 77°C, pressing force of 28MPa, feed rate of 60kg / h, rotational speed of 450rpm) and the temperature gradient control operation corresponding to the state are output, and the optimized operation process of the oil press and the corresponding rapeseed characteristics are stored in the rapeseed oil pressing reference library with a timestamp of 2024 - 05 - 01.

[0118] Second, an adaptive optimization control system for rapeseed oil pressing, comprising:

[0119] Initial operation process module: used to obtain the rapeseed oil pressing historical data, construct a rapeseed oil pressing reference library based on the rapeseed oil pressing historical data, and match the rapeseed oil pressing reference library according to the rapeseed information to be pressed to obtain the initial operation process of the oil press;

[0120] Prediction model module: construct a rapeseed oil pressing prediction model according to the rapeseed oil pressing historical data, and input the rapeseed information to be pressed and the initial operation process of the oil press into the rapeseed oil pressing prediction model to obtain the first predicted oil output index;

[0121] Correction module: used to analyze the operating current of the oil press to obtain the bearing wear rate of the oil press, input the operating state of the oil press and the bearing wear rate of the oil press into the state function to obtain the state index of the oil press, and correct the first predicted oil output index according to the state index of the oil press to obtain the second predicted oil output index;

[0122] Operating process optimization module: used to calculate the deviation of the oil output index, determine the weight of the oil output index, determine the pressing objective function according to the weight of the oil output index and the deviation of the oil output index, and perform dynamic game optimization on the initial operation process of the oil press according to the pressing objective function to output the optimized operation process of the oil press;

[0123] Management platform module: It is used to view, manage and store the historical data of rapeseed oil pressing, the deviation of oil output index, the wear rate of the bearings of the oil press, and the optimized operation process of the oil press, and adjust the operation process of the oil press according to the optimized operation process of the oil press.

[0124] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An adaptive optimization control method for rapeseed oil pressing, characterized in that: The following steps are involved: S1, obtaining rapeseed oil pressing historical data, and constructing a rapeseed oil pressing reference library and a rapeseed oil pressing prediction model according to the rapeseed oil pressing historical data; S2, obtaining rapeseed information, and matching the rapeseed oil pressing reference library according to the rapeseed information to obtain the initial operation process of the oil press; S3, according to the initial operation process of the oil press, the rapeseed to be pressed is pressed to obtain an actual oil yield index and an operating current of the oil press, and the information of the rapeseed to be pressed and the initial operation process of the oil press are input into the rapeseed oil pressing prediction model to obtain a first predicted oil yield index; S4, analyzing the operating current of the oil press to obtain the oil press bearing wear rate, inputting the operating state of the oil press and the oil press bearing wear rate into a state function to obtain an oil press state index, and correcting the first predicted oil output index according to the oil press state index to obtain a second predicted oil output index; S5, calculating the oil output index deviation, and selecting a process optimization strategy according to the oil output index deviation and the oil press bearing wear rate; the process optimization strategy includes a first process optimization strategy and a second process optimization strategy; The first process optimization strategy is specifically, when the oil output index deviation and the oil press bearing wear rate are greater than the corresponding warning threshold, an equipment operation warning is issued, and the corresponding operation process is adjusted to recalculate the oil output index deviation until no warning occurs; The second process optimization strategy is specifically as follows: when there is no warning situation, the oil output index weight is determined according to the oil output target, the pressing objective function is determined according to the oil output index weight and the oil output index deviation, and the initial operation process of the oil press is dynamically optimized according to the pressing objective function to output the optimized operation process of the oil press.

2. The adaptive optimization control method for rapeseed oil pressing according to claim 1, characterized in that: The method for constructing a rapeseed oil pressing reference library and a rapeseed oil pressing prediction model comprises: Acquire rapeseed oil pressing historical data, use decision tree to extract and classify the rapeseed oil pressing historical data to obtain historical oil pressing features, and set a timestamp according to time information; the historical oil pressing features include historical rapeseed features, historical operation process of the oil press, and historical oil output features; According to the timestamp, historical rapeseed characteristics and historical operation processes of the oil press are associated to form a rapeseed oil pressing reference library; Random forest is used to divide the historical oil pressing features into a training set and a test set at a ratio of 8:2, and a rapeseed oil pressing prediction model is constructed; the rapeseed oil pressing prediction model includes an input layer, a hidden layer and an output layer; The hidden layer uses a fully connected neural network to capture the data dependency of historical rapeseed characteristics, historical oil press operation processes and historical oil production characteristics, and predicts the oil production characteristics according to the rapeseed characteristics and the oil press operation processes; the output layer uses two parallel fully connected layers to output continuous oil production characteristics and discrete oil production characteristics respectively; the continuous oil production characteristics specifically indicate the oil rate, oil production speed, oil temperature, impurity rate and oil residue density; the discrete oil production characteristics specifically refer to the color grade of the oil body; The mean square error is used as the loss function to measure the difference between the predicted value and the true value of the continuous oil yield feature. The cross entropy loss function is used to measure the difference between the predicted probability distribution of discrete oil yield features and the true label distribution. The Adam optimizer is used to adjust the model learning rate. The test set is used to evaluate the model performance and input into the rapeseed oil pressing prediction model.

3. The adaptive optimization control method for rapeseed oil pressing according to claim 1, characterized in that: The method for obtaining the initial operation process of the oil press comprises: The rapeseed information feature is extracted to obtain the rapeseed features F = {f1, f2, ..., f n }, n is the number of features, calculate the rapeseed features to be pressed and the historical rapeseed features H of the jth group of rapeseed in the rapeseed oil pressing reference library j ={h j1 ,h j2 ,…,h jn }, extract the timestamp of the historical rapeseed features with the highest similarity, and select the corresponding historical operation process of the oil press as the initial operation process of the oil press according to the timestamp. The comprehensive similarity calculation is expressed as: Where Sim(F,H j ) are the eigenvectors F and H j The comprehensive similarity of , α is the control factor, c is the crossover constant, n(F∩H j ) is the number of intersections of two eigenvectors, n is the number of eigenvector features, t far is the oldest record timestamp in the rapeseed oil pressing reference library, t near is the timestamp of the latest record in the rapeseed oil pressing reference library, t j is the record timestamp of the jth group of historical rapeseed characteristics, Sim cov (F,H j ) are the eigenvectors F and H j The cosine similarity of d (F,H j ) are the eigenvectors F and H j The Euclidean distance of .

4. The adaptive optimization control method for rapeseed oil pressing according to claim 1, characterized in that: The method for obtaining the second predicted oil production index comprises: Perform Fourier transform on the running current of the oil press to obtain the current spectrum, extract the harmonic characteristics of the characteristic frequency of the bearing fault, and input the harmonic characteristics into the harmonic characteristic-wear rate mapping model to obtain the wear rate of the oil press bearing; the harmonic characteristics include current harmonic energy entropy, spectrum energy entropy, harmonic amplitude, harmonic coupling coefficient, harmonic distortion rate and load speed parameters; The steps of obtaining the harmonic feature-wear rate mapping model are as follows: simulating the progressive wear process of the oil press bearing by an accelerated wear test bench to obtain current signals at different wear stages, processing the current signals at different wear stages to obtain harmonic features corresponding to the characteristic frequency of the bearing fault, and constructing a nonlinear harmonic feature-wear rate mapping model according to the harmonic features and the actual wear rate; the nonlinear harmonic feature-wear rate mapping model includes a gradient boosting tree, Bayesian optimization and a mean square error loss function; the gradient boosting tree learns the data relationship between the harmonic features and the wear rate, and predicts the wear rate of the oil press bearing according to the input harmonic features; the Bayesian optimization optimizes the hyperparameters of the model; the mean square error loss function evaluates the difference between the predicted value and the actual value of the wear rate; The oil press operation status and oil press bearing wear rate are input into the state function to obtain the oil press state index, which is expressed as: Where E(t) is the oil press state index at time t, last is the time from the last maintenance, β is the maintenance attenuation coefficient, w1 is the working time degradation weight, t act is the accumulated working time, t sta is the design standard life, γ is the fatigue accumulation coefficient, w2 is the working strength degradation weight, P cur is the current work pressure, P max is the rated maximum pressure, w3 is the temperature degradation weight, k is the temperature sensitivity coefficient, T act is the current temperature, T safe is the safety temperature, w4 is the vibration degradation weight, m ​​is the number of vibration monitoring points, v i is the effective value of vibration acceleration at monitoring point i, v sta is the vibration threshold, w5 is the wear degradation weight, a cur is the current wear rate, a fai is the critical wear rate for failure, is the wear rate variation trend within the time interval Δt The first predicted oil yield index is corrected according to the oil press state index to obtain the second predicted oil yield index.

5. The adaptive optimization control method for rapeseed oil pressing according to claim 1, characterized in that: The method for performing equipment operation early warning comprises: Calculate the oil output index deviation according to the second predicted oil output index and the actual oil output index, and implement the first process optimization strategy when the oil output index deviation and the oil press bearing wear rate are greater than the corresponding warning threshold; the first process optimization strategy includes equipment operation warning and adjustment of the corresponding operation process; The equipment operation warning specifically includes: a. when any oil output index deviation is greater than the corresponding warning threshold, a process warning is performed; b. when the oil press bearing wear rate is greater than the corresponding warning threshold, a structural warning is performed, and wear analysis and maintenance of the oil press bearing are performed.

6. The adaptive optimization control method for rapeseed oil pressing according to claim 1, characterized in that: The method for determining the squeezing objective function comprises: Determine the oil production index weight δ according to the oil production target j , according to the oil production index weight δ j and the oil output index deviation Δx j Determine the squeeze objective function Aim, the expression is: R3=t run (a t -a t-1 ) in is the oil production index deviation weight, is the energy efficiency weight, is the dynamic wear weight, is the stability weight, R1 is the deviation objective function, R2 is the energy efficiency objective function, R3 is the wear objective function, R4 is the stability objective function, Δx j =x j -x j,real , x j is the second oil production prediction index, x j,real is the actual oil output index, P total is the total power corresponding to the current oil press operation process, Q oil is the oil output per unit time corresponding to the current oil press operation process, P rated is the standard rated power, ξ is the overload penalty coefficient, t run The estimated running time of the oil press is a t is the wear rate of the oil press at time t, θ1, θ2, θ3 are dynamic adjustment coefficients, is the vibration acceleration variance, is the variance of the temperature gradient of the pressing chamber, MTBF is the historical sliding window statistical value of the mean time between failures, and the dynamic update strategy of the wear weight is is the wear weight updated at time t+1, a cur is the current wear rate, a fai is the critical wear rate for failure, and the update frequency of wear weight is 30min / time.

7. The adaptive optimization control method for rapeseed oil pressing according to claim 1, characterized in that: The method for optimizing the operation process of the oil press by performing dynamic game optimization output includes: According to the pressing objective function, the number of players in the dynamic game is determined to be 4, and the corresponding player set is P = {p1, p2, p3, p4}, p1 is the oil output index deviation, p2 is the pressing energy efficiency, p3 is the machine wear, and p4 is the machine stability. The strategy space S = S1∪S2∪S3∪S4 corresponding to the dynamic game of the players is determined. S1, S2, S3, and S4 are the strategy sets of the corresponding index players respectively. The players execute the corresponding determined corresponding profit functions to obtain the profit function set U = {u1, u2, u3, u4}, where u1, u2, u3, and u4 are the profit functions of the corresponding index players, u1 = 1 / minR1, u2 = 1 / minR2, u3 = 1 / minR3, and u4 = 1 / minR4, which constitutes the dynamic game of the oil press operation process G = {P; S; U}; Input the dynamic game tree parameters G = {P; S; U}, use MapReduce to split the game tree into K sub-game forests, build the sub-game payoff matrix and calculate the number of stages y max , a dynamic benefit threshold is used to pre-prune the low-return game tree, and the pruning conditions are: in is the profit value of the ith subtree at the game tree node k, ε∈[0.1,0.3,] is the profit value threshold coefficient, is the maximum profit value of all subtrees at node k of the game tree; A pre-trained multi-agent Q-network is used to predict the Q-value score of each strategy, and the highest scoring strategy is selected as the sub-game equilibrium path; the multi-agent Q-network predicts the Q-value score of each strategy based on the historical strategy distribution, the sparsity of the payoff matrix and the node depth; the Q-value score is used to guide the Monte Carlo tree search direction; Iterate each sub-game block from the deepest stage to the initial stage according to the sub-game equilibrium path, calculate the confidence of the strategy of each decision node, mark the strategy with confidence greater than 0.9 as the dominant strategy and lock the path, and use Monte Carlo tree search for non-dominated strategies to correct the strategy score; The local game paths of all sub-game blocks are merged according to the game structure compression technology to generate the intermediate result of the global strategy combination; the game structure compression technology includes sub-game equivalence class merging and Huffman coding compression; the sub-game equivalence class merging identifies sub-games with similar structures through KL divergence and merges and calculates repeated strategy sequences; the Huffman coding compression encodes high-frequency strategy patterns to reduce the storage space occupied by the strategy sequence; Based on the Shapley value allocation model, the profit distribution and pressing objective function of the cross-subgame are verified, and the global game strategy combination and confidence evaluation report are output. According to the global game strategy combination, the optimized operation process of the oil press is output, and the optimized operation process of the oil press and the corresponding rapeseed characteristics are timestamped and stored in the rapeseed oil pressing reference library.

8. An adaptive optimization control system for rapeseed oil pressing, used to execute the method according to any one of claims 1 to 7, characterized in that: include: Initial operation process module: used to obtain rapeseed oil pressing historical data, build a rapeseed oil pressing reference library according to the rapeseed oil pressing historical data, and obtain the initial operation process of the oil press according to the information of rapeseed to be pressed and matching the rapeseed oil pressing reference library; Prediction model module: constructing a rapeseed oil pressing prediction model according to the rapeseed oil pressing historical data, inputting the rapeseed information to be pressed and the initial operation process of the oil press into the rapeseed oil pressing prediction model to obtain a first predicted oil yield index; Correction module: used for analyzing the running current of the oil press to obtain the bearing wear rate of the oil press, inputting the running state of the oil press and the bearing wear rate of the oil press into the state function to obtain the state index of the oil press, and correcting the first predicted oil output index according to the state index of the oil press to obtain the second predicted oil output index; Operation process optimization module: used to calculate the oil output index deviation, determine the oil output index weight, determine the pressing objective function according to the oil output index weight and the oil output index deviation, and perform dynamic game optimization on the initial operation process of the oil press according to the pressing objective function to output the optimized operation process of the oil press; Management platform module: used to view, manage and store the rapeseed oil pressing historical data, the oil output index deviation, the oil press bearing wear rate and the oil press optimized operation process, and adjust the oil press operation process according to the oil press optimized operation process.

Images