Intelligent warehouse caching scheduling method and system for production of oryzanol

Abstract

The present application relates to the technical field of warehouse scheduling, in particular to an intelligent warehouse caching scheduling method and system for phytosterol production; batch information of materials entering a warehouse system is acquired; in the material storage process, initial physical characteristic data of the materials is collected through online sensing equipment; a quality attenuation model is established for each batch of materials according to the initial physical characteristic data, and material attenuation data is obtained based on the quality attenuation model; production material consumption demand in a future period is periodically acquired, and analysis is carried out based on the material attenuation data and the production material consumption data, so as to balance potential value loss, task quality matching loss and warehouse operation cost loss of the materials, and generate a warehouse scheduling strategy in which the comprehensive loss in the period is minimum and the material batch calling sequence is planned in time sequence. The present application generates and issues an automatic instruction through the warehouse scheduling strategy, so as to guide the material delivery.

Description

Technical Field

[0001] This invention relates to the field of warehouse scheduling technology, specifically to an intelligent warehouse cache scheduling method and system for oryzanol production. Background Technology

[0002] In the supply chain and logistics sector, warehouse management is a core element that determines a company's operational efficiency and cost control. Its main goal is to achieve refined management of materials from warehousing and storage to outbound through effective data processing and resource scheduling, in order to respond to production needs, reduce inventory holding costs, and maximize asset value.

[0003] However, for many industries, such as pharmaceuticals, food processing, and fine chemicals, raw materials or products are time-varying, meaning their quality or value changes over time. In these scenarios, traditional inventory management strategies, such as "first-in, first-out" (FIFO) or simple inventory models based on fixed replenishment points, expose their inherent limitations. These strategies primarily manage materials based on the arrival time or quantity of material batches, ignoring the heterogeneity in quality and dynamic decay characteristics of different batches. This leads to blind spots in management decisions and fails to optimize the value of inventory assets.

[0004] Therefore, in the fields of inventory management and logistics scheduling technology, there is an urgent need for a more intelligent data processing method that integrates advanced predictive models and multi-objective optimization algorithms, using the dynamic quality data of materials as the core decision variable to optimize warehouse scheduling strategies.

[0005] To address this, a smart warehouse cache scheduling method and system for oryzanol production is proposed. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent warehouse cache scheduling method and system for oryzanol production. By balancing the potential value loss of materials, task quality matching loss, and warehouse operation cost loss, a warehouse scheduling strategy with the minimum comprehensive loss within the cycle and a time-ordered material batch call sequence is generated to guide material outbound.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A smart warehouse cache scheduling method for oryzanol production includes:

[0009] Obtain batch information of materials entering the warehousing system. The batch information includes a unique identifier, material source, material quantity, and material entry time. The material is a raw material for the production of oryzanol.

[0010] During the material storage process, initial physical characteristic data of the materials are collected through online sensing devices; based on the initial physical characteristic data, a quality decay model is established for each batch of materials to predict the quality degradation trajectory; and material decay data is identified based on the quality decay model.

[0011] The production material consumption requirements for future periods are periodically obtained, including the quantity of materials required for production tasks, material quality standards, and planned usage time.

[0012] Based on the analysis of material decay data and production material consumption data, the potential value loss of materials, task quality matching loss and warehouse operation cost loss are balanced to generate a warehouse scheduling strategy that minimizes the overall loss within the cycle and plans the material batch call sequence in time order. Based on the warehouse scheduling strategy, automated instructions are generated and issued to guide the material outbound.

[0013] The online sensing device collects initial physical characteristic data of materials by using a hyperspectral imaging device to scan and detect material samples at detection points during the material warehousing process, thereby obtaining multidimensional sensing data of the materials.

[0014] The multidimensional sensing data includes spatial dimension information and spectral dimension information. The spatial dimension information is used to characterize the physical morphology of the material sample, and the spectral dimension information is used to characterize the chemical composition characteristics of the material sample.

[0015] Establishing a quality decay model for each batch of materials specifically means: pre-constructing a hybrid basic model that integrates a physical mechanism module and a data-driven module; after the initial physical characteristic data of a new batch of materials is collected, using the initial physical characteristic data to perform personalized calibration of the parameters in the hybrid basic model to generate a quality decay model;

[0016] The physical mechanism module uses chemical kinetic equations to generate baseline degradation prediction data based on environmental data. The data-driven module uses a deep neural network model to receive initial physical characteristic data of materials, baseline degradation prediction data and environmental data as input, and outputs corrected data of the baseline degradation prediction data.

[0017] Material decay data is obtained based on the quality decay model. Specifically, this includes: calling the established quality decay model, inputting real-time collected and predicted periodic environmental data in the warehouse, performing simulation calculations, and generating material decay data; the material decay data includes the predicted center point values ​​of key material quality indicators.

[0018] The key quality indicators of the material include acid value data, peroxide data, oryzanol content data, and moisture content data.

[0019] The warehouse scheduling strategy is implemented using a heuristic optimization algorithm, which searches for a solution by evaluating and iterating through a population of candidate warehouse scheduling strategies.

[0020] An initial population consisting of multiple candidate warehouse scheduling strategies is generated. The fitness of the candidate warehouse scheduling strategies in the initial population is evaluated to obtain a fitness evaluation value. The fitness evaluation value is calculated through a preset comprehensive objective function.

[0021] Based on the fitness evaluation value, the current population is operated on by selection, crossover and mutation operators to generate a new offspring population; the population is iteratively identified until the preset convergence condition is met, and finally the candidate warehouse scheduling strategy with the best fitness evaluation value in the iteration process is output.

[0022] The comprehensive objective function is calculated by combining potential value loss, task quality matching loss and warehouse operation cost loss to obtain a fitness evaluation value.

[0023] The potential value loss is calculated by quantifying the quality degradation of each batch of materials involved in the candidate warehousing scheduling strategy within the planning period based on the quality degradation model, and then summing them up to obtain the potential value loss.

[0024] Task quality matching loss is calculated based on the deviation between the predicted quality index of the material batch and the material quality standard of the production task. The predicted quality index is calculated based on material decay data.

[0025] The aforementioned warehousing operation cost loss is derived by estimating the energy consumption required to complete material handling; the energy consumption is estimated based on the storage location of each batch of materials in the warehousing system, the target outbound location, and the performance parameters of the executing equipment.

[0026] The combined calculation is specifically a weighted summation. Before the weighted summation, the three loss terms are normalized in terms of dimensions. Specifically, the normalization process involves mapping each loss term to a dimensionless standardized scoring interval.

[0027] A smart warehouse cache scheduling system for oryzanol production includes:

[0028] The inbound inspection module acquires batch information of materials entering the warehousing system. The batch information includes a unique identifier, material source, material quantity, and material inbound time. The material is a raw material for the production of oryzanol.

[0029] The degradation identification module collects initial physical characteristic data of materials through online sensing devices during the material storage process; based on the initial physical characteristic data, it establishes a quality degradation model for each batch of materials to predict the quality deterioration trajectory; and identifies material degradation data based on the quality degradation model.

[0030] The task requirement module periodically acquires the production material consumption requirements for future periods. The production material consumption requirements include the quantity of materials required for the production task, the material quality standards, and the planned usage time.

[0031] The warehouse scheduling module analyzes material decay data and production material consumption data to balance the potential value loss of materials, task quality matching loss, and warehouse operation cost loss. It generates a warehouse scheduling strategy that minimizes the overall loss within the cycle and plans the material batch call sequence in chronological order. Based on the warehouse scheduling strategy, it generates and issues automated instructions to guide the material outbound process.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0033] 1. This invention utilizes hyperspectral imaging technology to non-invasively and rapidly acquire spatial and spectral information of material samples. The spatial information reflects the physical properties of the material, such as uniformity and impurity distribution, while the spectral information accurately characterizes the intrinsic properties of key chemical components, such as acid value and oryzanol content. This high-dimensional, information-rich initial feature data provides high-quality input for the subsequent establishment of highly personalized and accurate mass decay models, ensuring the accuracy and reliability of the prediction model from the source.

[0034] 2. This invention provides a degradation trend benchmark for the model that conforms to basic scientific laws through the physical mechanism module, ensuring the stability and interpretability of the prediction; then, with the help of the powerful nonlinear fitting capability of the data-driven module, it learns from the high-dimensional initial feature data and makes up for the personalized influences that the physical model cannot identify, outputting a fine correction to the benchmark prediction; thus solving the inherent defects of single physical mechanism models or pure data-driven models when predicting complex industrial objects, thereby achieving high-precision and robust prediction of material quality decay trajectory.

[0035] 3. This invention employs a heuristic optimization algorithm that, by simulating the iterative processes of natural selection and genetics, performs an intelligent search within a vast solution space. This allows it to find a near-optimal scheduling strategy with minimal overall cost within an acceptable timeframe. Through a series of operations including evaluation, selection, crossover, and mutation, the algorithm continuously optimizes the candidate strategy population, ultimately converging to a high-quality solution. This not only ensures the quality of the decision-making process but also guarantees that the entire intelligent scheduling system can meet the real-world demands of industrial production for rapid response and dynamic adjustment. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of a smart warehouse cache scheduling method for oryzanol production according to the present invention.

[0037] Figure 2 This is a schematic diagram illustrating the logic for obtaining the warehouse scheduling strategy of the present invention;

[0038] Figure 3 This is a schematic diagram of the structure of an intelligent warehousing cache scheduling system for oryzanol production according to the present invention. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0040] Oryzanol is an important nutritional supplement and pharmaceutical raw material, primarily extracted from rice bran oil. In the production process of oryzanol, its core raw material is usually crude rice bran oil or pre-processed rice bran oil, and the quality of the rice bran oil has a decisive impact on the yield and quality of the final product. However, these oily raw materials rich in unsaturated fatty acids are highly susceptible to hydrolysis and oxidation reactions during storage due to environmental factors such as temperature, humidity, oxygen, and light, leading to continuous deterioration in quality. This is mainly manifested in increased acid value and peroxide value, as well as a decrease in the content of the core active ingredient, oryzanol.

[0041] To address this, a smart warehouse cache scheduling method and system for oryzanol production is proposed.

[0042] Example 1:

[0043] This invention proposes an intelligent warehouse cache scheduling method for oryzanol production, the process of which is as follows: Figure 1 As shown, it includes:

[0044] Obtain batch information of materials entering the warehousing system. The batch information includes a unique identifier, material source, material quantity, and material entry time. The material is rice bran oil, a raw material for the production of oryzanol.

[0045] During the material storage process, initial physical characteristic data of the materials are collected through online sensing devices; based on the initial physical characteristic data, a quality decay model is established for each batch of materials to predict the quality degradation trajectory; and material decay data is identified based on the quality decay model.

[0046] The production material consumption requirements for future periods are periodically obtained, including the quantity of materials required for production tasks, material quality standards, and planned usage time.

[0047] Based on the analysis of material decay data and production material consumption data, the potential value loss of materials, task quality matching loss and warehouse operation cost loss are balanced to generate a warehouse scheduling strategy that minimizes the overall loss within the cycle and plans the material batch call sequence in time order. Based on the warehouse scheduling strategy, automated instructions are generated and issued to guide the material outbound.

[0048] The logic for obtaining the warehouse scheduling strategy is as follows: Figure 2 As shown.

[0049] Preferably, a unique identifier is assigned to each newly received batch of materials, and its associated data is recorded, including the material source, material quantity, and specific material receipt time.

[0050] Preferably, online sensing equipment is deployed along the route that raw materials must take to enter the warehouse. In this embodiment, hyperspectral imaging equipment is preferred. This equipment performs non-invasive scanning of the flowing oil to obtain high-dimensional initial physical feature data.

[0051] This data comprises two dimensions: spatial dimension information and spectral dimension information. Spatial dimension information consists of a set of image data characterizing the physical morphology of the oilseed, such as uniformity and the presence of minute suspended impurities. Spectral dimension information consists of a high-dimensional data matrix (or "hyperspectral data cube"), where each pixel corresponds to a complete spectral curve, recording the absorbance / reflectance of the oilseed in hundreds of consecutive narrow wavelength bands. This forms a chemical fingerprint, comprehensively characterizing the current chemical composition of the material, including acid value data, peroxide data, oryzanol content data, and moisture content data.

[0052] This invention uses hyperspectral imaging technology to acquire spatial and spectral information of material samples. The spatial information reflects the physical properties of the material, such as uniformity and impurity distribution, while the spectral information accurately characterizes the intrinsic properties of key chemical components, such as acid value and oryzanol content. This provides high-quality input for establishing highly personalized and accurate mass decay models, ensuring the accuracy and reliability of the prediction models from the source.

[0053] Establishing a quality decay model for each batch of materials specifically means: pre-constructing a hybrid basic model that integrates a physical mechanism module and a data-driven module; after the initial physical characteristic data of a new batch of materials is collected, using the initial physical characteristic data to perform personalized calibration of the parameters in the hybrid basic model to generate a quality decay model;

[0054] Specifically, the mass decay model also includes a mapping process from spectral data to chemical indicators:

[0055] Randomly selected rice bran oil samples were collected using the hyperspectral imaging system, and their acid value, peroxide value, oryzanol content, and moisture content were accurately measured using standard experimental methods to obtain the true values.

[0056] Before spectral preprocessing, the spectral data is calibrated and denoised using spatial dimension information. This step includes:

[0057] Anomaly region identification and removal: Analyze the spatial dimension of the image and identify and mark non-representative regions using image segmentation or object detection algorithms, such as bubbles caused by uneven flow, glare reflected from pipe walls, or tiny suspended impurities. The spectral information of these regions is considered noise or interference.

[0058] Effective region spectral extraction: The pixels in the above-mentioned abnormal regions are masked, and only the uniform and interference-free main body of the material is retained as the region of interest; the average spectrum of all effective pixels in the region of interest is calculated to generate spectral data that can more accurately and stably represent the overall chemical composition characteristics of the batch of materials.

[0059] The collected spectral data are preprocessed, for example, by using standard normal variable transformation (SNV) to eliminate the influence of particle size and surface scattering; the partial least squares regression (PLSR) algorithm is used to establish a spectral prediction module for each key quality indicator, with the preprocessed spectral data as the independent variable and the true value of chemical measurement as the dependent variable; the model performance is evaluated through cross-validation to ensure that the root mean square error and coefficient of determination of the model meet the preset accuracy requirements.

[0060] The physical mechanism module incorporates chemical kinetic equations and the Arrhenius equation to describe the oxidation and hydrolysis reactions of oils, thereby describing the effects of environmental factors such as temperature on the hydrolysis and oxidation rates of rice bran oil. This provides a baseline prediction of quality deterioration that conforms to basic scientific principles, and yields baseline deterioration prediction data.

[0061] Specifically, the data inputs of the physical mechanism module include initial physical characteristic data, baseline degradation prediction data, and environmental data, including temperature, humidity, and oxygen concentration.

[0062] Data-driven modules typically employ a deep neural network model (such as CNN or Transformer) pre-trained on a large amount of historical data; they receive initial physical feature data and environmental data as input, and their powerful nonlinear fitting capabilities enable them to learn complex influencing factors that physical models cannot describe, outputting corrected data for benchmark predictions.

[0063] Specifically, the data input to the data-driven module includes initial physical feature data, baseline degradation prediction data, and environmental data. In this embodiment, the data-driven module employs a convolutional neural network (CNN) containing multiple convolutional layers and multiple fully connected layers. It takes a spectral feature map and a one-dimensional vector containing five environmental parameters as input; the output is a correction vector with four elements, corresponding to the correction values ​​of four key quality indicators. This correction value is added to the output value of the physical mechanism module to obtain the final center point prediction value.

[0064] Once the initial physical characteristic data of the new batch of materials is collected, the hybrid basic model is quickly and individually calibrated using techniques such as transfer learning, resulting in a "tailor-made" mass decay model for the new batch.

[0065] The personalized calibration includes: collecting initial hyperspectral data when a new batch of materials enters the warehouse; inputting the initial hyperspectral data into a pre-trained hybrid basic model and comparing it with the measured chemical indicators at the time of entry; using the residual between the two, fine-tuning the weights of the fully connected layers in the model through a backpropagation algorithm to generate a personalized quality degradation model for that batch.

[0066] This invention provides a degradation trend benchmark for the model that conforms to basic scientific laws through a physical mechanism module, ensuring the stability and interpretability of the prediction. Then, by leveraging the powerful nonlinear fitting capability of the data-driven module, it learns from high-dimensional initial feature data and compensates for personalized influences that the physical model cannot identify, outputting a fine correction to the benchmark prediction. This solves the inherent defects of single physical mechanism models or pure data-driven models when predicting complex industrial objects, thereby achieving high-precision and robust prediction of material quality decay trajectories.

[0067] Material degradation data is obtained based on the quality degradation model. Specifically, this includes: calling the established quality degradation model, inputting real-time collected and predicted periodic environmental data in the warehouse, performing simulation calculations, and generating material degradation data; the material degradation data includes the center point prediction value of key quality indicators of the material; the center point prediction value is obtained based on the baseline deterioration prediction data and the correction data.

[0068] Furthermore, in addition to the predicted center point values ​​of key quality indicators, the material attenuation data also includes a quantitative assessment of the uncertainty of the predicted center point values. Specifically, this can be achieved by using techniques such as Bayesian neural networks or Monte Carlo Dropout, so that the model provides a corresponding confidence interval or probability distribution while outputting the predicted values. This uncertainty information will be used by the downstream warehouse scheduling module to optimize the reliability of warehouse scheduling.

[0069] For example, when making scheduling decisions, the system may avoid assigning batches of materials with high forecast uncertainty to production tasks with extremely stringent quality requirements, or introduce a risk penalty term related to uncertainty when calculating task quality matching losses, thereby making the scheduling strategy more robust and risk-avoiding.

[0070] The key quality indicators of the material include acid value data, peroxide data, oryzanol content data, and moisture content data.

[0071] In daily operations, the quality degradation model for each batch in the warehouse is periodically invoked. Real-time collected and predicted environmental data for the current cycle are input, and the simulation calculates the predicted center point value for that batch over a future period, yielding material degradation data. This material degradation data accurately depicts the changing trajectories of key quality indicators such as acid value, peroxide value, oryzanol content, and moisture content over time.

[0072] This invention further clarifies the core output of the quality decay model, transforming the abstract prediction of material deterioration into specific, quantifiable, and directly related key quality indicators for the production of oryzanol. By specifying that the material decay data must include the predicted center point values ​​of four indicators—acid value, peroxide value, oryzanol content, and moisture content—it ensures that the decision-making basis of the scheduling system is aligned with the actual needs of the production process and quality control standards.

[0073] By communicating with the factory's manufacturing execution system or enterprise resource planning system on a regular basis via API, the system automatically obtains the production material consumption requirements for future cycles and parses them into structured data. This data clarifies the planned usage time, required material quantity, and detailed quality standards for each production task, ultimately yielding the required material quantity, material quality standards, and planned usage time for each production task.

[0074] The warehouse scheduling strategy is implemented using a heuristic optimization algorithm, which searches for a solution by evaluating and iterating through a population of candidate warehouse scheduling strategies.

[0075] An initial population consisting of multiple candidate warehouse scheduling strategies is generated. The fitness of the candidate warehouse scheduling strategies in the initial population is evaluated to obtain a fitness evaluation value. The fitness evaluation value is calculated through a preset comprehensive objective function.

[0076] Based on the fitness evaluation value, the current population is operated on by selection, crossover and mutation operators to generate a new offspring population; the population is iteratively identified until the preset convergence condition is met, and finally the candidate warehouse scheduling strategy with the best fitness evaluation value in the iteration process is output.

[0077] This invention employs a heuristic optimization algorithm that intelligently searches a vast solution space by simulating the iterative processes of natural selection and genetics. It can find a near-optimal scheduling strategy with minimal overall cost within an acceptable timeframe. Through a series of operations including evaluation, selection, crossover, and mutation, the algorithm continuously optimizes the candidate strategy population, ultimately converging to a high-quality solution. This not only guarantees the quality of decision-making but also ensures that the entire intelligent scheduling system meets the real-world demands of industrial production for rapid response and dynamic adjustment.

[0078] The comprehensive objective function is calculated by combining potential value loss, task quality matching loss and warehouse operation cost loss to obtain a fitness evaluation value.

[0079] The potential value loss is calculated by quantifying the quality degradation of each batch of materials involved in the candidate warehousing scheduling strategy within the planning period based on the quality degradation model, and then summing them up to obtain the potential value loss.

[0080] Task quality matching loss is calculated based on the deviation between the predicted quality index of the material batch and the material quality standard of the production task. The predicted quality index is calculated based on material decay data.

[0081] The aforementioned warehousing operation cost loss is derived by estimating the energy consumption required to complete material handling; the energy consumption is estimated based on the storage location of each batch of materials in the warehousing system, the target outbound location, and the performance parameters of the executing equipment.

[0082] Regarding potential value loss:

[0083] For each batch of materials to be shipped out under the strategy, a quality degradation model is invoked to predict key quality indicators (such as acid value and oryzanol content) at the planned future usage time. By comparing the predicted quality with the current quality, and combining this with preset value assessment rules, the quality degradation caused by storage is converted into a specific value loss. The total value loss is obtained by summing the value losses of all batches in the strategy.

[0084] If the material is not planned to be used within the current cycle, the potential value loss is calculated using the key quality indicators at the beginning and end of the cycle.

[0085] The valuation rules include: establishing a benchmark value linked to core indicators, with the content of the core active ingredient, oryzanol, as the primary factor determining the material's basic value. Further, the rule can be set as follows: the benchmark value of a material is directly proportional to its oryzanol content. For negative indicators such as acid value, peroxide value, and moisture content, these are set as value penalties. Whenever the acid value exceeds the ideal benchmark value, a corresponding penalty is deducted from the material's value for each unit increase. The penalty rules for peroxide value and moisture content are similar. The potential value loss of a material can be measured by its market price; different qualities of material correspond to different market prices.

[0086] In summary, the instantaneous value of a batch of materials at any given moment is calculated as follows: the basic value of the material minus the penalty for negative indicators.

[0087] Regarding task quality matching loss:

[0088] The predicted quality of the material batches assigned to a production task in the scheduling strategy at the planned usage time is compared with the material quality standard explicitly required by the production task, and the degree of deviation between the two is calculated. Whether it is the cost waste caused by excessively high material quality or the production risk caused by substandard quality, it will be quantified as matching loss. This total loss is the sum of the matching losses of all tasks within the strategy.

[0089] Specifically, the material quality standard required for the production task is processed into a standard quality vector; the predicted quality of the material at the planned usage time is processed into a predicted quality vector; and the matching loss is measured based on the distance between the standard quality vector and the predicted quality.

[0090] Furthermore, task quality matching loss can also be measured by cost. As shown in the potential value loss, there are differences in market acquisition costs between materials with different key quality indicators. The standard material cost is determined based on the material quality standards required for the production task, and the predicted material cost is determined based on the predicted quality of the material at the planned usage time. The task quality matching loss is determined based on the standard material cost and the predicted material cost.

[0091] Regarding warehouse operation cost losses:

[0092] Based on the current storage location of each batch of materials in the warehouse, the target outbound location planned by the scheduling strategy, and the performance parameters (such as energy consumption per unit distance) of the automated equipment performing the handling task, the energy consumption required to complete this physical displacement is estimated. The estimated energy consumption of all handling tasks is accumulated to form the total loss of warehouse operation costs; furthermore, energy consumption can be converted into energy cost.

[0093] This invention constructs a comprehensive and quantitative method for evaluating scheduling strategies based on overall losses. It decomposes the merits of scheduling strategies into three interrelated yet distinct core dimensions: potential value loss, task quality matching loss, and warehouse operation cost loss. This transforms the fuzzy "optimal" objective into a structured and computable comprehensive objective function.

[0094] Since the three losses mentioned above may have different dimensions, they are first normalized, mapping each loss to a standard, dimensionless scoring interval. Then, the system assigns different weight coefficients to the three standardized losses based on preset business priorities (such as cost control, quality priority, etc.) and performs a weighted summation. The final single value is the comprehensive fitness evaluation value of the candidate warehouse scheduling strategy, used for iterative selection in the heuristic optimization algorithm.

[0095] This invention maps various losses to a unified, standardized scoring range through "dimensional normalization," eliminating interference from differences in units and scales and making them additive. Based on this, a "weighted summation" method is used for combination, allowing for dynamic adjustment of the weight coefficients of different loss items according to market changes, production priorities, or cost control strategies, thereby flexibly adjusting the optimization objective.

[0096] Furthermore, the warehouse scheduling strategy can be generated using a multi-objective optimization algorithm, rather than weighted summation of multiple loss targets into a single objective.

[0097] The multi-objective optimization algorithm treats potential value loss, task quality matching loss, and warehouse operation cost loss as three independent optimization objectives simultaneously, and finally outputs a Pareto optimal frontier. The Pareto optimal frontier consists of a set of non-dominant solutions, each of which represents a different loss trade-off strategy; for example, a solution with the lowest operating cost but slightly higher value loss, or a solution with the best quality matching but slightly higher operating cost.

[0098] This design provides warehouse managers with multiple optimized and selectable scheduling options, enabling them to make flexible decisions based on temporary business priorities and significantly improving management flexibility.

[0099] This invention also proposes an intelligent warehousing cache scheduling system for oryzanol production, with the following structure: Figure 3 As shown, it includes:

[0100] The inbound inspection module acquires batch information of materials entering the warehousing system. The batch information includes a unique identifier, material source, material quantity, and material inbound time. The material is a raw material for the production of oryzanol.

[0101] The degradation identification module collects initial physical characteristic data of materials through online sensing devices during the material storage process; based on the initial physical characteristic data, it establishes a quality degradation model for each batch of materials to predict the quality deterioration trajectory; and identifies material degradation data based on the quality degradation model.

[0102] The task requirement module periodically acquires the production material consumption requirements for future periods. The production material consumption requirements include the quantity of materials required for the production task, the material quality standards, and the planned usage time.

[0103] The warehouse scheduling module analyzes material decay data and production material consumption data to balance the potential value loss of materials, task quality matching loss, and warehouse operation cost loss. It generates a warehouse scheduling strategy that minimizes the overall loss within the cycle and plans the material batch call sequence in chronological order. Based on the warehouse scheduling strategy, it generates and issues automated instructions to guide the material outbound process.

[0104] Example 2:

[0105] This invention proposes an intelligent warehouse cache scheduling method for oryzanol production, comprising:

[0106] Obtain batch information of materials entering the warehousing system. The batch information includes a unique identifier, material source, material quantity, and material entry time. The material is rice bran oil, a raw material for the production of oryzanol.

[0107] During the material storage process, initial physical characteristic data of the materials are collected through online sensing devices; based on the initial physical characteristic data, a quality decay model is established for each batch of materials to predict the quality degradation trajectory; and material decay data is identified based on the quality decay model.

[0108] In this embodiment, a pushbroom near-infrared hyperspectral imager with a spectral range of 900-1700nm and a spectral resolution of 5nm is used. The detection point is set on a section of quartz glass flow cell in the inlet pipeline. The oil flows through at a rate of 0.5L / min, and two 150W halogen lamps are mounted at a 45-degree angle above the flow cell as the light source.

[0109] The acquired raw hyperspectral image data were sequentially corrected using black and white plates to eliminate the effects of dark current and light source inhomogeneity. Subsequently, Savitzky-Golay smoothing (window size 9, polynomial order 2) and standard normal variable transformation (SNV) were used for spectral preprocessing to eliminate noise caused by sample surface scattering and optical path variation.

[0110] The production material consumption requirements for future periods are periodically obtained, including the quantity of materials required for production tasks, material quality standards, and planned usage time.

[0111] Based on the analysis of material decay data and production material consumption data, the potential value loss of materials, task quality matching loss and warehouse operation cost loss are balanced to generate a warehouse scheduling strategy that minimizes the overall loss within the cycle and plans the material batch call sequence in time order. Based on the warehouse scheduling strategy, automated instructions are generated and issued to guide the material outbound.

[0112] The warehouse scheduling strategy is implemented using a heuristic optimization algorithm, which searches for a solution by evaluating and iterating through a population of candidate warehouse scheduling strategies.

[0113] An initial population consisting of multiple candidate warehouse scheduling strategies is generated. The fitness of the candidate warehouse scheduling strategies in the initial population is evaluated to obtain a fitness evaluation value. The fitness evaluation value is calculated through a preset comprehensive objective function.

[0114] Based on the fitness evaluation value, the current population is operated on by selection, crossover and mutation operators to generate a new offspring population; the population is iteratively identified until the preset convergence condition is met, and finally the candidate warehouse scheduling strategy with the best fitness evaluation value in the iteration process is output.

[0115] In this embodiment, a genetic algorithm is used to solve the above-mentioned warehouse scheduling strategy.

[0116] Encoding scheme: Each individual (chromosome) is an integer permutation, the length of which is equal to the total number of outbound tasks within the planning period. The gene value at the i-th position on the chromosome represents the unique identifier of the material batch used by the i-th outbound task.

[0117] Specific operators: The selection operator uses roulette wheel selection; the crossover operator uses sequential crossover; and the mutation operator uses exchange mutation, which randomly selects two genes on the chromosome and swaps them.

[0118] Parameter settings: Initial population size is set to 100, maximum number of iterations is 500, crossover probability is 0.8, and mutation probability is 0.1. The algorithm converges when the optimal fitness value changes by less than 0.01% over 50 consecutive generations.

[0119] The potential value loss is calculated by quantifying the quality degradation of each batch of materials involved in the candidate warehousing scheduling strategy within the planning period based on the quality degradation model, and then summing them up to obtain the potential value loss.

[0120] The rule can be set as follows: the base value of a material is proportional to its oryzanol content.

[0121] Based on historical production and sales data regarding raw material prices with varying oryzanol content, a benchmark price ladder is established for rice bran oil with different oryzanol content ranges. Simultaneously, negative indicators such as acid value, peroxide value, and moisture content are designated as penalty items. The rule can be set as follows: whenever a negative indicator exceeds an ideal benchmark value, the material value is deducted for each unit change. For example, "Compared to the ideal benchmark value, for every 1 mg / g increase in acid value, the value is reduced by 50 yuan / ton."

[0122] Task quality matching loss is calculated based on the deviation between the predicted quality index of the material batch and the material quality standard of the production task. The predicted quality index is calculated based on material decay data.

[0123] The aforementioned warehousing operation cost loss is derived by estimating the energy consumption required to complete material handling; the energy consumption is estimated based on the storage location of each batch of materials in the warehousing system, the target outbound location, and the performance parameters of the executing equipment.

[0124] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart warehouse cache scheduling method for oryzanol production, characterized in that, include: Obtain batch information of materials entering the warehousing system. The batch information includes a unique identifier, material source, material quantity, and material entry time. The material is rice bran oil, a raw material for the production of oryzanol. During the material storage process, initial physical characteristic data of the material are collected through online sensing devices. Specifically, hyperspectral imaging equipment is used to obtain spatial and spectral information of the material sample. The spatial information reflects the uniformity and impurity distribution of the material, while the spectral information characterizes the intrinsic properties of acid value, peroxide value, oryzanol content, and moisture content. Based on the initial physical characteristic data, a quality decay model is established for each material batch to predict the quality degradation trajectory; material decay data is then identified based on the quality decay model. The quality decay model for predicting the quality degradation trajectory for each batch of materials specifically refers to: pre-constructing a hybrid basic model that integrates a physical mechanism module and a data-driven module; after the initial physical characteristic data of a new batch of materials is collected, using the initial physical characteristic data to perform personalized calibration on the parameters in the hybrid basic model to generate a quality decay model; The physical mechanism module uses chemical kinetic equations to generate baseline degradation prediction data based on environmental data, including temperature, humidity, and oxygen concentration. The data-driven module employs a convolutional neural network with multiple convolutional and fully connected layers, receiving initial physical characteristic data of the material, baseline degradation prediction data, and environmental data as input, and outputting corrected data for the baseline degradation prediction data. Specifically, the output of the data-driven module is a correction vector containing four elements, corresponding to the correction values ​​of four key quality indicators. This correction value is added to the output value of the physical mechanism module to obtain the final center point prediction value. The key quality indicators include acid value data, peroxide data, oryzanol content data, and moisture content data. The production material consumption requirements for future periods are periodically obtained, including the quantity of materials required for production tasks, material quality standards, and planned usage time. Based on the analysis of material decay data and production material consumption requirements, the potential value loss of materials, task quality matching loss and warehouse operation cost loss are balanced to generate a warehouse scheduling strategy that minimizes the overall loss within the cycle and plans the material batch call sequence in time order. Based on the warehouse scheduling strategy, automated instructions are generated and issued to guide the material outbound. The warehouse scheduling strategy is implemented through a heuristic optimization algorithm, which searches for a solution by evaluating and iterating through a population of candidate warehouse scheduling strategies. An initial population consisting of multiple candidate warehouse scheduling strategies is generated. The fitness of the candidate warehouse scheduling strategies in the initial population is evaluated to obtain a fitness evaluation value. The fitness evaluation value is calculated by a preset comprehensive objective function. The comprehensive objective function is calculated by combining potential value loss, task quality matching loss and warehouse operation cost loss to obtain the fitness evaluation value.

2. The intelligent warehousing cache scheduling method for oryzanol production according to claim 1, characterized in that: The online sensing device collects initial physical characteristic data of materials by using a hyperspectral imaging device to scan and detect material samples at detection points during the material warehousing process, thereby obtaining multidimensional sensing data of the materials. The multidimensional sensing data includes spatial dimension information and spectral dimension information. The spatial dimension information is used to characterize the physical morphology of the material sample, and the spectral dimension information is used to characterize the chemical composition characteristics of the material sample.

3. The intelligent warehousing cache scheduling method for oryzanol production according to claim 1, characterized in that: Material decay data is obtained based on the quality decay model. Specifically, this includes: calling the established quality decay model, inputting real-time collected and predicted periodic environmental data in the warehouse, performing simulation calculations, and generating material decay data; the material decay data includes the predicted center point values ​​of key quality indicators of the material.

4. The intelligent warehousing cache scheduling method for oryzanol production according to claim 1, characterized in that: Based on the fitness evaluation value, the current population is operated on by selection, crossover and mutation operators to generate a new offspring population; the new offspring population is iteratively identified until the preset convergence condition is met, and finally the candidate warehouse scheduling strategy with the best fitness evaluation value in the iteration process is output.

5. The intelligent warehouse cache scheduling method for oryzanol production according to claim 4, characterized in that: The potential value loss is calculated by quantifying the quality degradation of each batch of materials involved in the candidate warehousing scheduling strategy within the planning period based on the quality degradation model, and then summing them up to obtain the potential value loss. Task quality matching loss is calculated based on the deviation between the predicted quality index of the material batch and the material quality standard of the production task. The predicted quality index is calculated based on material decay data. The aforementioned warehouse operation cost loss is derived by estimating the energy consumption required to complete material handling. Energy consumption is estimated based on the storage location of each batch of materials in the warehousing system, the target outbound location, and the performance parameters of the executing equipment.

6. The intelligent warehouse cache scheduling method for oryzanol production according to claim 5, characterized in that: The combined calculation is specifically a weighted summation. Before the weighted summation, the three loss terms are normalized in terms of dimensions. Specifically, the normalization process involves mapping each loss term to a dimensionless standardized scoring interval.

7. A smart warehouse cache scheduling system for oryzanol production, used to implement the smart warehouse cache scheduling method for oryzanol production as described in any one of claims 1-6, characterized in that, include: The inbound inspection module obtains batch information of materials entering the warehousing system. The batch information includes a unique identifier, material source, material quantity, and material inbound time. The material is a raw material for the production of oryzanol; The attenuation identification module collects the initial physical characteristic data of the material through online sensing devices during the material storage process; Based on the initial physical characteristic data, a quality decay model is established for each material batch to predict the quality degradation trajectory; material decay data is then identified based on the quality decay model. The task requirement module periodically acquires the production material consumption requirements for future periods. The production material consumption requirements include the quantity of materials required for the production task, the material quality standards, and the planned usage time. The warehouse scheduling module analyzes material decay data and production material consumption requirements to balance potential value loss of materials, task quality matching loss, and warehouse operation cost loss. It generates a warehouse scheduling strategy that minimizes overall loss within the cycle and plans the material batch call sequence in chronological order. Based on the warehouse scheduling strategy, it generates and issues automated instructions to guide material outbound.

Images