A method and system for intelligent distribution of powder additives

Abstract

The present invention provides a method and system for intelligent distribution of powder additives, which relates to the technical field of data processing. The method includes: obtaining the historical consumption speed of powder additives, the current inventory of consumption points and the distribution time of consumption points; combining the long short-term memory network model and the Bayesian algorithm to predict the consumption speed of consumption points, and outputting the predicted consumption speed sequence; integrating the predicted consumption speed sequence to obtain the available time of the current inventory of consumption points; obtaining the expected available time, and determining the expected consumption point distribution volume; when the distribution time of the consumption point is greater than the available time, increasing the initial distribution speed of the consumption point until the adjusted consumption point distribution time is less than the available time, and distributing the expected consumption point distribution volume to the corresponding consumption point according to the increased initial distribution speed of the consumption point, otherwise, distributing the expected consumption point distribution volume to the corresponding consumption point according to the initial distribution speed of the consumption point. Improve distribution efficiency and accuracy, and reduce the supply chain cost of powder additives.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an intelligent distribution method and system for powder additives. Background Art

[0002] Powder additives are a type of material widely used in various industrial fields. They exist in powder form and are mainly used to improve or enhance the performance of materials in industrial processes. These additives can include fillers, lubricants, adhesives, dispersants, plasticizers, defoamers, catalysts, etc. According to different application scenarios, powder additives can be added to the production process of various chemicals, plastics, rubber, coatings, ceramics, building materials and other products to achieve functions such as improving fluidity, increasing strength, enhancing adhesion, adjusting viscosity, and promoting chemical reactions.

[0003] Intelligent distribution of powder additives can ensure the continuity of the production process, prevent supply interruptions, and avoid inventory backlogs. Through accurate forecasting and dynamic adjustment of distribution plans, it can effectively respond to demand fluctuations, reduce logistics costs, improve production efficiency, and reduce environmental and safety risks, thereby comprehensively optimizing supply chain management.

[0004] However, the existing distribution of powder additives is often based on rough manual estimates to determine the demand at the consumption point, which is highly subjective and prone to misjudgment, leading to supply interruptions or inventory backlogs at the consumption point, thereby greatly increasing the supply chain cost of powder additives. Summary of the invention

[0005] In order to solve the technical problem that the distribution of powder additives in the prior art is often based on rough manual estimation to determine the demand at the consumption point, which is highly subjective and prone to misjudgment, resulting in supply interruption or inventory backlog at the consumption point, thereby greatly increasing the supply chain cost of the powder additives, the present invention provides a method and system for intelligent distribution of powder additives.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] First aspect

[0008] An embodiment of the present invention provides a method for intelligently distributing a powder additive, comprising:

[0009] S1: Obtain the historical consumption speed of powder additives, the current inventory of consumption points and the delivery time of consumption points, where the delivery speed of the consumption point delivery time is the initial delivery speed of the consumption point;

[0010] S2: Based on the historical consumption speed, the long short-term memory network model and the Bayesian algorithm are combined to make a single-point consumption speed prediction for the consumption point, and output the predicted consumption speed sequence;

[0011] S3: Integrate the predicted consumption speed sequence at each time step to obtain the available time of the inventory at the current consumption point;

[0012] S4: Get expected available time;

[0013] S5: Determine the expected consumption point delivery volume based on the consumption point delivery time and the expected available time;

[0014] S6: When the delivery time at the consumption point is longer than the available time, increase the initial delivery speed at the consumption point to adjust the delivery time at the consumption point until the adjusted delivery time at the consumption point is shorter than the available time, otherwise, proceed to step S8;

[0015] S7: Deliver the expected consumption point delivery volume to the corresponding consumption point according to the increased consumption point initial delivery speed;

[0016] S8: Deliver the expected consumption point delivery quantity to the corresponding consumption point according to the initial delivery speed of the consumption point.

[0017] Second aspect

[0018] An embodiment of the present invention provides an intelligent distribution system for powder additives, comprising:

[0019] processor;

[0020] A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the intelligent distribution method for powder additives as described in the first aspect is implemented.

[0021] The third aspect

[0022] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the intelligent distribution method of powder additives as described in the first aspect is implemented.

[0023] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0024] In the present invention, according to the historical consumption rate of powder additives, single-point prediction is performed through long short-term memory network model and Bayesian algorithm, the complex time series dependency in the consumption rate is captured, and the error accumulation is reduced through Bayesian correction, so as to improve the accuracy and stability of the prediction, and improve the accuracy of the predicted consumption rate sequence, and then the predicted consumption rate sequence is integrated at each corresponding time step based on the objective data to obtain the available time of the current consumption point inventory, and then the expected consumption point delivery volume is determined with the expected available time as the target, and the available time is compared with the consumption point delivery time to avoid the problem of powder additive supply chain interruption caused by the delivery time exceeding the available time. In addition, the expected consumption point delivery volume is obtained based on the accurate predicted consumption rate sequence, which can make full use of the discovered historical laws to determine a reasonable delivery volume, avoid inventory backlogs, improve the timeliness and accuracy of powder additive delivery, reduce the cost of the powder additive supply chain, and enhance the competitiveness of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0026] Figure 1 A schematic diagram of a process of an intelligent distribution method of powder additives provided by an embodiment of the present invention;

[0027] Figure 2 A schematic diagram of the structure of a powder additive intelligent distribution system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0029] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0030] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0031] Reference Manual Attached Figure 1, showing a schematic flow chart of a method for intelligent distribution of powder additives provided in an embodiment of the present invention.

[0032] The embodiment of the present invention provides a method for intelligent distribution of powder additives, which can be implemented by an intelligent distribution device for powder additives, and the intelligent distribution device for powder additives can be a terminal or a server. The processing flow of the method for intelligent distribution of powder additives can include the following steps:

[0033] S1: Obtain the historical consumption rate of powder additives, the current inventory of consumption points and the delivery time of consumption points, where the delivery speed of the consumption point delivery time is the initial delivery speed of the consumption point.

[0034] Among them, historical consumption speed, current consumption point inventory, consumption point delivery time and consumption point initial delivery speed are all data of the target consumption point to be analyzed. By collecting the historical consumption speed, current inventory and consumption point delivery time of powder additives, a data foundation is established to accurately predict the consumption speed and reasonably formulate a delivery plan in the subsequent steps, thereby ensuring the continuity and effectiveness of supply during the production or sales process.

[0035] S2: Based on the historical consumption speed, the long short-term memory network model and the Bayesian algorithm are combined to make a single-point consumption speed prediction for the consumption point, and output the predicted consumption speed sequence.

[0036] Among them, the long short-term memory network model is a special recursive neural network that can be used to process and predict time series data. It can effectively capture long-term dependencies and short-term fluctuations in the data. Through memory cells and gating mechanisms, it avoids the gradient vanishing problem in traditional recursive neural networks. The Bayesian algorithm is a probabilistic reasoning method based on the Bayesian theorem. By combining prior knowledge and observed data to update the prediction model, it can be used to correct the prediction value of the model in time series prediction, reduce prediction errors, and provide quantification of prediction uncertainty. Predicting the consumption speed sequence refers to using historical consumption speed data to generate a series of consumption speed estimates for future time steps through a prediction model (such as LSTM), which represents the possible consumption speed of each time step in a certain period of time in the future. By combining the temporal dependency capture capability of the long short-term memory network model and the error correction characteristics of the Bayesian algorithm, step S2 can more accurately predict the consumption speed sequence, which can not only reflect the long-term trend in historical data, but also dynamically adjust the prediction results to improve the overall accuracy and reliability of the prediction, thereby optimizing subsequent distribution decisions.

[0037] In a possible implementation, S2 specifically includes:

[0038] S201: Obtain historical consumption speed sequences at different sampling frequencies.

[0039] The sampling frequency may be every minute, every hour, every day or every week. The regularity information reflected by the data at different sampling frequencies is different. For example, the sampling frequency information per hour cannot reflect the regularity information of every day or every week. Obtaining information at different sampling frequencies can capture historical consumption information more comprehensively.

[0040] S202: Searching for the best alignment path for fusing each historical consumption speed sequence with the minimum alignment cost to obtain a historical consumption speed fusion sequence.

[0041] Among them, the minimum alignment cost is an indicator used to measure and optimize the similarity between two time series. When aligning time series, we hope to find a way to make the elements of the two sequences match as much as possible and the cost of the matching process is minimized. This cost is the sum of the distances (such as Euclidean distances) between the corresponding elements of the two sequences. When looking for the minimum alignment cost, we try to minimize the sum of these distances by selecting an optimal alignment path. The best alignment path refers to the path of a set of matching points selected when aligning two time series so that the total cost of alignment is minimized. The best alignment path reflects how to match the elements in the two sequences so that the difference between them is minimized. Through the minimum alignment cost method, different historical consumption speed sequences are optimized and fused to find the best alignment path between them, thereby generating a more accurate and consistent historical consumption speed fusion sequence. This process can effectively integrate multiple time series data, capture their respective time characteristics and trends, reduce noise and deviations in the data, and provide a more reliable basis for subsequent predictions.

[0042] In a possible implementation, S202 specifically includes:

[0043] S2021: Establish a distance matrix between the first historical consumption speed sequence and the second historical consumption speed sequence:

[0044]

[0045] Among them, the elements in the distance matrix D Represents the first historical consumption speed sequence The i-th element of the second historical consumption speed sequence is the j-th element of The Euclidean distance of Indicates calculating the Euclidean distance.

[0046] S2022: Based on the distance matrix, establish a cumulative distance matrix for describing the minimum alignment cost:

[0047]

[0048] Among them, the elements in the cumulative distance matrix C Indicates from arrive The minimum cumulative distance, min means taking the minimum value.

[0049] It should be noted that in the cumulative distance matrix, The boundary conditions of the alignment process are defined, and the starting point of the alignment path is set. The top means that the current alignment is to keep the second historical consumption speed sequence unchanged and take one more step on the first historical consumption speed sequence. The left represents the opposite process, that is, the first historical consumption speed sequence remains unchanged, and the second historical consumption speed sequence takes one more step. The top left means moving two sequences at the same time. This recursive formula constructs the overall best alignment path, that is, the minimum alignment cost, by selecting the path with the smallest cost among the three possible directions.

[0050] S2023: Using the minimum cumulative distance as the backtracking constraint, reverse backtracking is performed starting from the end point element of the cumulative distance matrix, and the best backtracking point, i.e., the best predecessor node, of each step is recorded. The backtracking constraint of each step is specifically as follows:

[0051]

[0052] in, represents the best backtracking point, It means to take the function that takes the minimum value i and j The value of .

[0053] S2024: The best backtracking point is used as a new end point element, i.e., a new backtracking starting point, and reverse backtracking is performed again until the starting point element of the cumulative distance matrix is ​​backtracked.

[0054] S2025: Reversely combine the best backtracking points in the backtracking order to obtain the best alignment path.

[0055] For example: Suppose there are two time series X=[1,2,3] and Y=[2,2,4], and their distance matrix is ​​calculated as: , the cumulative distance matrix is , the backtracking process is: from Start backtracking, and the optimal alignment path is: .

[0056] S2026: Align the first historical consumption speed sequence and the second historical consumption speed sequence according to the optimal alignment path to obtain a historical consumption speed fusion sequence.

[0057] Specifically, by constructing a distance matrix between two historical consumption speed series and building a cumulative distance matrix based on the matrix, the optimal path with the minimum alignment cost is found. On this basis, the optimal alignment method is determined by backtracking. In this way, multiple time series can be accurately aligned so that they are matched and fused to the greatest extent in the time dimension, thereby generating a more consistent and representative historical consumption speed fusion sequence. This fusion sequence not only retains the key features of each sequence, but also reduces the errors caused by time differences, providing a more reliable data basis for subsequent predictions.

[0058] In a possible implementation manner, after S202, the method further includes:

[0059] S202A: Combine the singular value decomposition algorithm and the empirical mode decomposition algorithm to extract the lookback window representing the periodic law of the historical consumption velocity fusion sequence.

[0060] Among them, the singular value decomposition algorithm is a linear algebraic decomposition technique that decomposes a matrix into the product of three matrices, namely the left singular matrix, the singular value matrix and the right singular matrix. It can be used for data dimensionality reduction and feature extraction to help identify the most important patterns or trends in the data. The empirical mode decomposition algorithm is a signal processing technology that is used to decompose complex time series into several intrinsic mode functions (IMFs). These IMFs represent different frequency components of the signal. They are suitable for processing nonlinear and non-stationary data and help analyze the local characteristics of time series. The lookback window is a time range used to determine the length of historical data that needs to be referenced when making predictions or analyses. The choice of the lookback window affects the accuracy and applicability of the prediction. By combining the singular value decomposition algorithm and the empirical mode decomposition algorithm to extract the lookback window that represents the periodic law of the historical consumption speed fusion sequence, this process can accurately capture the periodicity and trend characteristics in the data, ensure that the selected lookback window can better reflect the key patterns of historical data, thereby improving the accuracy and stability of the prediction and optimizing subsequent supply chain management decisions.

[0061] In a possible implementation manner, S202A specifically includes:

[0062] S202A1: Sequence extension of the historical consumption speed fusion sequence based on the singular value decomposition algorithm:

[0063]

[0064] in, X represents the embedding matrix of each time step of the fusion sequence including the historical consumption speed, Indicates the number of historical consumption speed fusion sequences i The value of the time step, , TIndicates the length of the historical consumption speed fusion sequence, i Represents the column index, n represents the number of rows in the embedding matrix representing the window length, i.e., the row index, U represents the embedding matrix Left singular vectors, Represents the singular values ​​of the embedding matrix The diagonal matrix of V The right singular vectors of the embedding matrix are The orthogonal matrix of k Indicates the number of principal components whose singular value contribution is greater than the preset singular value contribution. represents the approximate matrix reconstructed based on the number of principal components, Represent the corresponding left singular vector, diagonal matrix and right singular vector reconstructed based on the number of principal components, respectively. It represents the historical consumption speed fusion sequence after extension. Express The right singular matrix obtained by linear extrapolation is, Indicates transpose.

[0065] S202A2: Generate the complementary sequence of the extended historical consumption speed fusion sequence by difference method:

[0066]

[0067] in, Complementary sequence The p values, They represent the first p The value and p -1 value.

[0068] S202A3: Perform empirical mode decomposition on the complementary pair consisting of the extended historical consumption speed fusion sequence and the corresponding complementary sequence to extract the intrinsic mode function:

[0069]

[0070] in, Respectively represent the signal, that is, the complementary pair t The envelope of the moment and t The moment envelope, express t The mean value of the signal at each moment, Indicates the historical consumption speed fusion sequence or the corresponding complementary sequence after the extension t The value of the moment, express t Refine the signal at all times, represents the first extracted intrinsic mode function, Indicates the extracted t The moment k The intrinsic mode functions, Represents the residual signal obtained after extraction, which is a monotonic function.

[0071] Specifically, after extracting the first intrinsic mode function, when extracting the second intrinsic mode function, it is subtracted from the original sequence, i.e., the sequence before extraction, to obtain a residual signal. The above process is repeated to obtain the final residual signal. The residual signal is a monotonic function, which means that the residual function has a small fluctuation and further extraction is meaningless.

[0072] S202A4: Remove the extended portion of the intrinsic mode function corresponding to the historical consumption speed fusion sequence.

[0073] S202A5: Calculate the autocorrelation local maximum of the intrinsic mode function value and determine the intrinsic mode function period:

[0074]

[0075] in, Indicates time lag, express The eigenmode function value at time , M express The total length of Indicates the time lag The similarity between the eigenmode function of and itself, represents the lag time corresponding to the first local maximum, represents the period of the intrinsic mode function.

[0076] in, represents the lag time corresponding to the first local maximum, which is the time lag point at which the signal is highly similar to itself again for the first time. The time lag is specifically the amount by which the signal is offset in time. The autocorrelation function calculates the similarity of the signal at different lags.

[0077] S202A6: Output the time window covering the period of the intrinsic mode function as a backtracking window.

[0078] Specifically, this process first uses the singular value decomposition algorithm to extend the historical consumption speed fusion sequence to generate a richer data matrix. Then, the extended complementary sequence is generated by the difference method. Next, these sequences are used to extract the intrinsic mode functions (IMFs) through empirical mode decomposition. After extracting the IMFs, the extended part is removed and the local maximum of the autocorrelation function is calculated to determine the periodicity of the IMFs. Finally, based on this periodic information, the lookback window representing the historical consumption speed fusion sequence is determined and output. It can fully capture the complex patterns and periodic characteristics in the time series. The sequence extension of the singular value decomposition algorithm can reduce the impact of the endpoint effect, while the decomposition process of the empirical mode decomposition can more finely separate the modal functions of different frequencies, thereby avoiding mode aliasing and ensuring that the extracted periodic features are more accurate and stable. By generating extended sequences and complementary sequences and performing sophisticated modal decomposition and periodic analysis on them, we can accurately extract the key patterns representing historical consumption rates and optimize the selection of the lookback window. This method not only improves the depth of data analysis, but also enhances the reliability and stability of the prediction, thus providing a solid foundation for more accurate consumption rate prediction and supply chain management.

[0079] S203: extract data from the historical consumption speed fusion sequence according to the lookback window, and predict the extracted historical consumption speed fusion sequence in combination with the long short-term memory network model and the Bayesian algorithm, and output the predicted consumption speeds of multiple time steps, wherein each predicted consumption speed constitutes a predicted consumption speed sequence.

[0080] It should be noted that by obtaining the historical consumption rate series with different sampling frequencies, the regularity information of the data at different time scales can be captured, avoiding the information loss that may be caused by a single sampling frequency. Then, these multi-frequency information is fused through the minimum alignment cost method to generate a more representative historical consumption rate fusion sequence. Combining the long short-term memory network model and the Bayesian algorithm for prediction can make full use of these fused data and output more accurate multi-step prediction results. This process not only captures the complex patterns in the time series, but also reduces the cumulative error through Bayesian correction, thereby improving the accuracy and stability of the prediction, and ultimately optimizing the distribution and inventory management of powder additives.

[0081] In a possible implementation, S203 specifically includes:

[0082] S2031: Input the historical consumption speed fusion sequence in the lookback window into the pre-trained long short-term memory network model, and output the single-step prediction value:

[0083]

[0084] in, and They represent the forget gate The weight matrix and bias vector of represents the sigmoid activation function, represents matrix multiplication, t Indicates the current moment, Indicates the previous moment t -1 moment of hidden state, express t The moment input is the historical consumption speed fusion sequence, and Represents the input gate The weight matrix and bias vector of and Represent the candidate cell states at the current moment The weight matrix and bias vector of , tanh represents the hyperbolic tangent function, and Respectively represent the cell state at the current moment and the cell state at the previous moment, and Represents the output gate The weight matrix and bias vector of Indicates the current hidden state. The predicted value at the next moment is t +1 moment prediction value, and Represent the output layer weight matrix and output layer bias vector respectively.

[0085] Among them, the pre-trained long short-term memory network model is trained by using part of the historical data as input data and the other part of the historical data as output data. The hyperparameters of the long short-term memory network model are continuously adjusted during the entire training process until the model can accurately predict the output data.

[0086] S2032: Correct the single-step prediction value using the Bayesian algorithm:

[0087]

[0088] in, represents the corrected forecast value for the next moment, represents the variance of the predicted value, represents the variance of the prediction error at the current time step, express t The estimated true value at time, i represents the variable representing the time step, Indicates i The weight of the time step, Indicates i The actual consumption speed value of the time step, represents the Bayesian correction term, i.e., the degree of correction, and Respectively t The actual consumption speed value at each moment and t The predicted value at the moment, Indicates i The prediction error for each time step is N Indicates the size of the lookback window, i.e. the historical data reference window.

[0089] Among them, the historical data reference window is the maximum time step considered in the entire prediction process.

[0090] It should be noted that as the forecast progresses, the forecast error may accumulate. The correction can reduce the impact of this error accumulation at each step. If the predicted value deviates from the historical trend, the correction formula will be based on Adjust the predicted value to reduce the error. In the correction process, different weights are assigned according to the closeness to the current moment. The closer the data is to the current moment, the higher the weight is assigned to strengthen the impact of the most recent data, while reducing the interference of earlier data on the prediction, thereby increasing the sensitivity to the latest trend changes and improving the accuracy of the prediction. Especially in a rapidly changing environment, this method can better capture the real-time dynamics of the data.

[0091] S2033: Using the corrected single-step prediction value as the consumption speed value of the next time step of the historical consumption speed fusion sequence to update the historical consumption speed fusion sequence.

[0092] S2034: Re-input the updated historical consumption speed fusion sequence into the long short-term memory network model and output a new single-step prediction value.

[0093] S2035: Repeat steps S2031 to S2034 until a preset number of single-step prediction values ​​are obtained, and the single-step prediction values ​​are combined according to the prediction time to obtain a predicted consumption speed sequence.

[0094] It should be noted that by combining the time-series dependency capture capability of the long short-term memory network model (LSTM) and the error correction mechanism of the Bayesian algorithm, a predicted consumption speed sequence is gradually generated. Each prediction step is based on the previous time step, and Bayesian correction is used to reduce error accumulation, thereby ensuring the consistency and accuracy of the prediction results. This method not only improves the sensitivity to historical trends, but also can flexibly respond to data fluctuations in a rapidly changing environment. The final output of the predicted consumption speed sequence is more accurate and stable, which helps to optimize subsequent supply chain decisions.

[0095] S3: Integrate the predicted consumption speed sequence at each corresponding time step to obtain the available time of the inventory at the current consumption point.

[0096] It should be noted that by integrating the predicted consumption rate sequence at each time step, the length of time that the current inventory can support (i.e., the available time) can be accurately calculated. This process can help optimize inventory management and ensure that insufficient or excessive inventory is avoided while meeting demand, thereby improving the efficiency and stability of the supply chain.

[0097] In a possible implementation, S3 specifically includes:

[0098] S301: Integrate the predicted consumption speed sequence at each corresponding time step and output the predicted consumption.

[0099] S302: Calculate the integral time when the predicted consumption is equal to the inventory at the current consumption point, and output the integral time as the available time:

[0100]

[0101] in, Indicates the available time, i.e. the credit time. Indicates the inventory quantity at the consumption point, represents the discrete time step, Indicates i Time steps The consumption rate is the consumption per unit time. ,in, k Represents a time index.

[0102] It should be noted that by integrating the predicted consumption speed sequence, the time length that the current inventory can support (i.e., the available time) can be accurately calculated, thereby helping enterprises to dynamically adjust their supply chain strategies. This method can effectively prevent inventory shortages or excessive backlogs in an uncertain demand environment, and ensure the continuity and efficiency of production or sales.

[0103] S4: Obtain expected available time.

[0104] Among them, the expected availability time refers to the length of time that the desired inventory can be maintained when making a distribution plan. This length of time is usually set by the company based on production demand, supply chain response speed, market fluctuations and other factors to ensure that the inventory is neither overstocked nor affected by insufficient supply. By setting the expected availability time, a clear goal is provided for subsequent distribution decisions, making the distribution strategy more targeted and controllable, and avoiding blindness in inventory management.

[0105] S5: Determine the expected consumption point delivery volume based on the consumption point delivery time and the expected available time.

[0106] It should be noted that those skilled in the art can set the expected available time length according to actual needs, and the present invention is not limited thereto.

[0107] It should be noted that by comprehensively considering the delivery time and expected available time of the consumption point, the required delivery volume can be accurately calculated to ensure that while meeting consumption needs, inventory management is optimized to avoid excessive or insufficient inventory. This process improves the accuracy of the distribution plan and reduces uncertainty and operating costs in the supply chain.

[0108] In a possible implementation, the expected consumption point delivery volume is calculated as follows:

[0109]

[0110] in, represents the expected consumption point delivery quantity, Represents the predicted consumption speed sequence t Always predict consumption speed, Indicates the current moment, Indicates the expected availability time. Indicates the delivery time of the consumption point.

[0111] It should be noted that by calculating the expected consumption point delivery volume in this integral way, the changes in consumption speed in different time periods can be accurately considered, the difference between the expected available time and the actual delivery time can be comprehensively considered, and the consumption of the delivery process can be considered to ensure that demand can be accurately met in each time step, avoiding insufficient or excessive inventory due to demand fluctuations or delivery delays, thereby optimizing inventory management and delivery strategies.

[0112] S6: When the delivery time at the consumption point is longer than the available time, increase the initial delivery speed at the consumption point to adjust the delivery time at the consumption point until the adjusted delivery time at the consumption point is less than the available time, otherwise, proceed to step S8.

[0113] It should be noted that by dynamically adjusting the delivery speed of the consumption point, timely adjustments can be made when the delivery time is longer than the available time to avoid inventory shortages due to delivery delays. This can enable flexible response to changes in demand and supply chain fluctuations, improve the response speed and stability of the supply chain, and ensure that production or sales activities are not affected.

[0114] S7: Deliver the expected consumption point delivery volume to the corresponding consumption point according to the increased initial delivery speed of the consumption point.

[0115] S8: Deliver the expected consumption point delivery quantity to the corresponding consumption point according to the initial delivery speed of the consumption point.

[0116] It should be noted that under the adjusted and initial delivery speeds, the expected consumption point delivery volume is accurately delivered to the corresponding consumption points as planned. In this way, whether the delivery speed needs to be adjusted or not, it can ensure that the consumption points are replenished in time to prevent inventory shortages or surpluses, while maximizing delivery efficiency and flexibility to adapt to different actual needs and supply chain conditions.

[0117] In actual application, by combining historical data and advanced forecasting models, the consumption rate of powder additives can be accurately predicted, and the distribution strategy can be dynamically adjusted to ensure sufficient inventory and avoid supply interruptions. At the same time, the distribution speed and volume can be optimized to improve the flexibility and responsiveness of the supply chain. It can effectively prevent excessive or insufficient inventory, reduce operating costs, ensure the smooth progress of the production or sales process, and thus improve the efficiency and reliability of the overall supply chain.

[0118] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0119] In the present invention, according to the historical consumption rate of powder additives, single-point prediction is performed through long short-term memory network model and Bayesian algorithm, the complex temporal dependency in the consumption rate is captured, and the error accumulation is reduced through Bayesian correction, so as to improve the accuracy and stability of the prediction, and improve the accuracy of the predicted consumption rate sequence, and then the predicted consumption rate sequence is integrated at each corresponding time step based on the objective data to obtain the available time of the current consumption point inventory, and then the expected consumption point delivery volume is determined with the expected available time as the target, and the available time is compared with the consumption point delivery time to avoid the problem of powder additive supply chain interruption caused by the delivery time exceeding the available time. In addition, the expected consumption point delivery volume is obtained based on the accurate predicted consumption rate sequence, which can make full use of the discovered historical laws to determine a reasonable delivery volume, avoid inventory backlogs, improve the timeliness and accuracy of powder additive delivery, reduce the cost of the powder additive supply chain, and enhance the competitiveness of enterprises.

[0120] Reference Manual Attached Figure 2 , showing a schematic structural diagram of a powder additive intelligent distribution system provided by the present invention.

[0121] The present invention further provides a powder additive intelligent distribution system 20, which is applied to the above-mentioned powder additive intelligent distribution method, comprising:

[0122] Processor 201.

[0123] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, the intelligent distribution method of powder additives as described in the method embodiment is implemented.

[0124] The powder additive intelligent distribution system 20 provided by the present invention can execute the above-mentioned powder additive intelligent distribution method and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0125] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0126] In the present invention, according to the historical consumption rate of powder additives, single-point prediction is performed through long short-term memory network model and Bayesian algorithm, the complex time series dependency in the consumption rate is captured, and the error accumulation is reduced through Bayesian correction, so as to improve the accuracy and stability of the prediction, and improve the accuracy of the predicted consumption rate sequence, and then the predicted consumption rate sequence is integrated at each corresponding time step based on the objective data to obtain the available time of the current consumption point inventory, and then the expected consumption point delivery volume is determined with the expected available time as the target, and the available time is compared with the consumption point delivery time to avoid the problem of powder additive supply chain interruption caused by the delivery time exceeding the available time. In addition, the expected consumption point delivery volume is obtained based on the accurate predicted consumption rate sequence, which can make full use of the discovered historical laws to determine a reasonable delivery volume, avoid inventory backlogs, improve the timeliness and accuracy of powder additive delivery, reduce the cost of the powder additive supply chain, and enhance the competitiveness of enterprises.

[0127] It should be understood that the processor in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0128] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0129] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0130] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0131] In the present invention, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0132] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0133] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0134] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0135] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0136] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0137] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0138] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0139] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the intelligent distribution method of powder additives as described in the method embodiment is implemented.

[0140] A computer-readable storage medium provided by the present invention can implement the steps and effects of the intelligent distribution method of powder additives in the above method embodiment. To avoid repetition, the present invention will not go into details.

[0141] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0142] In the present invention, according to the historical consumption rate of powder additives, single-point prediction is performed through long short-term memory network model and Bayesian algorithm, the complex time series dependency in the consumption rate is captured, and the error accumulation is reduced through Bayesian correction, so as to improve the accuracy and stability of the prediction, and improve the accuracy of the predicted consumption rate sequence, and then the predicted consumption rate sequence is integrated at each corresponding time step based on the objective data to obtain the available time of the current consumption point inventory, and then the expected consumption point delivery volume is determined with the expected available time as the target, and the available time is compared with the consumption point delivery time to avoid the problem of powder additive supply chain interruption caused by the delivery time exceeding the available time. In addition, the expected consumption point delivery volume is obtained based on the accurate predicted consumption rate sequence, which can make full use of the discovered historical laws to determine a reasonable delivery volume, avoid inventory backlogs, improve the timeliness and accuracy of powder additive delivery, reduce the cost of the powder additive supply chain, and enhance the competitiveness of enterprises.

[0143] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0144] There are a few points to note:

[0145] (1) The drawings of the embodiments of the present invention only involve structures related to the embodiments of the present invention. Other structures may refer to conventional designs.

[0146] (2) For the sake of clarity, in the drawings used to describe the embodiments of the present invention, the thickness of layers or regions is exaggerated or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0147] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0148] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for intelligent distribution of powder additives, characterized in that: include: S1: Obtain the historical consumption speed of the powder additive, the current inventory of the consumption point and the delivery time of the consumption point, wherein the delivery speed of the consumption point delivery time is the initial delivery speed of the consumption point; S2: According to the historical consumption speed, a long short-term memory network model and a Bayesian algorithm are combined to perform a single-point consumption speed prediction for the consumption point, and output a predicted consumption speed sequence; S3: Integrate the predicted consumption speed sequence at each corresponding time step to obtain the available time of the inventory at the current consumption point; S4: Get expected available time; S5: Determine the expected consumption point delivery volume according to the consumption point delivery time and the expected available time; S6: when the delivery time of the consumption point is longer than the available time, increasing the initial delivery speed of the consumption point to adjust the delivery time of the consumption point until the adjusted delivery time of the consumption point is less than the available time, otherwise, proceeding to step S8; S7: delivering the expected consumption point delivery quantity to the corresponding consumption point according to the increased consumption point initial delivery speed; S8: delivering the expected consumption point delivery quantity to the corresponding consumption point according to the initial delivery speed of the consumption point; Wherein, the S2 specifically includes: S201: Obtain historical consumption speed sequences with different sampling frequencies; S202: searching for the best alignment path for fusing each historical consumption speed sequence with the minimum alignment cost to obtain a historical consumption speed fusion sequence; S203: extracting data from the historical consumption speed fusion sequence according to the backtracking window, and predicting the extracted historical consumption speed fusion sequence in combination with the long short-term memory network model and the Bayesian algorithm, and outputting predicted consumption speeds of multiple time steps, wherein each predicted consumption speed constitutes the predicted consumption speed sequence; Wherein, the S202 specifically includes: S2021: Establish a distance matrix between the first historical consumption speed sequence and the second historical consumption speed sequence: Among them, the elements in the distance matrix D Represents the first historical consumption speed sequence No. i The element of the second historical consumption rate sequence j Elements The Euclidean distance of Indicates calculation of Euclidean distance; S2022: Based on the distance matrix, establish a cumulative distance matrix for describing the minimum alignment cost: Among them, the elements in the cumulative distance matrix C Indicates from arrive The minimum cumulative distance, min means taking the minimum value; S2023: Using the minimum cumulative distance as a backtracking constraint, reverse backtracking is performed starting from the end point element of the cumulative distance matrix, and the best backtracking point, i.e., the best predecessor node, of each step is recorded. The backtracking constraint of each step is specifically: in, represents the best backtracking point, It means to take the function that takes the minimum value i and j The value of S2024: taking the best backtracking point as a new end point element, i.e., a new backtracking starting point, and re-backtracking in reverse until backtracking to the starting point element of the cumulative distance matrix; S2025: Reversely combine the best backtracking points in the backtracking order to obtain the best alignment path; S2026: Align the first historical consumption speed sequence and the second historical consumption speed sequence according to the optimal alignment path to obtain the historical consumption speed fusion sequence.

2. The intelligent distribution method of powder additives according to claim 1, characterized in that: After S202, the method further includes: S202A: extracting a lookback window representing the periodicity of the historical consumption speed fusion sequence by combining a singular value decomposition algorithm and an empirical mode decomposition algorithm.

3. The intelligent distribution method of powder additives according to claim 2, characterized in that: The S202A specifically includes: S202A1: Perform sequence extension on the historical consumption speed fusion sequence based on a singular value decomposition algorithm: in, X represents the embedding matrix of each time step of the fusion sequence including the historical consumption speed, Indicates the number of historical consumption speed fusion sequences i The value of the time step, , T Indicates the length of the historical consumption speed fusion sequence, i Represents the column index, n represents the number of rows in the embedding matrix representing the window length, i.e., the row index, U Represents the embedding matrix Left singular vectors, Represents the singular values ​​of the embedding matrix The diagonal matrix of V Represents the right singular vector of the embedding matrix, that is, The orthogonal matrix of k Indicates the number of principal components whose singular value contribution is greater than the preset singular value contribution. represents the approximate matrix reconstructed based on the number of principal components, Represent the corresponding left singular vector, diagonal matrix and right singular vector reconstructed based on the number of principal components, respectively. It represents the historical consumption speed fusion sequence after extension. Express The right singular matrix obtained by linear extrapolation is, represents transpose; S202A2: Generate the complementary sequence of the extended historical consumption speed fusion sequence by difference method: in, Complementary sequence The p values, They represent the first p The value and p -1 value; S202A3: Perform empirical mode decomposition on the complementary pair consisting of the extended historical consumption speed fusion sequence and the corresponding complementary sequence to extract the intrinsic mode function: in, Respectively represent the signal, that is, the complementary pair t The envelope of the moment and t The moment envelope, express t The mean value of the signal at each moment, Indicates the historical consumption speed fusion sequence or the corresponding complementary sequence after the extension t The value of the moment, express t Refine the signal at all times, represents the first extracted intrinsic mode function, Indicates the extracted t The moment k The intrinsic mode functions, It means that the residual signal obtained after extraction is a monotonic function; S202A4: removing the extended portion of the intrinsic mode function corresponding to the historical consumption speed fusion sequence; S202A5: Calculate the autocorrelation local maximum of the intrinsic mode function value and determine the intrinsic mode function period: in, Indicates time lag, express The eigenmode function value at time , M express The total length of Indicates the time lag The similarity between the eigenmode function of and itself, represents the lag time corresponding to the first local maximum, represents the period of the intrinsic mode function; S202A6: Output the time window covering the period of the intrinsic mode function as the backtracking window.

4. The intelligent distribution method of powder additives according to claim 1, characterized in that: The S203 specifically includes: S2031: Input the historical consumption speed fusion sequence in the lookback window into the pre-trained long short-term memory network model, and output the single-step prediction value: in, and They represent the forget gate The weight matrix and bias vector of represents the sigmoid activation function, represents matrix multiplication, t Indicates the current moment, Indicates the previous moment t -1 moment of hidden state, express t The moment input is the historical consumption speed fusion sequence, and Represents the input gate The weight matrix and bias vector of and Represent the candidate cell states at the current moment The weight matrix and bias vector of , tanh represents the hyperbolic tangent function, and Respectively represent the cell state at the current moment and the cell state at the previous moment, and Represents the output gate The weight matrix and bias vector of Indicates the current hidden state. The predicted value at the next moment is t +1 moment prediction value, and Represent the output layer weight matrix and output layer bias vector respectively; S2032: Correcting the single-step prediction value by using a Bayesian algorithm: in, represents the corrected forecast value for the next moment, represents the variance of the predicted value, represents the variance of the prediction error at the current time step, express t The estimated true value at time, i represents the variable representing the time step, Indicates i The weight of the time step, Indicates i The actual consumption speed value of the time step, represents the Bayesian correction term, i.e., the degree of correction, and Respectively t The actual consumption speed value at each moment and t The predicted value at the moment, Indicates i The prediction error for each time step is N Indicates the size of the lookback window, i.e. the historical data reference window; S2033: Using the corrected single-step prediction value as the consumption speed value of the next time step of the historical consumption speed fusion sequence to update the historical consumption speed fusion sequence; S2034: re-inputting the updated historical consumption speed fusion sequence into the long short-term memory network model, and outputting a new single-step prediction value; S2035: Repeat steps S2031 to S2034 until a preset number of single-step prediction values ​​are obtained, and the single-step prediction values ​​are combined according to the prediction time to obtain the predicted consumption speed sequence.

5. The intelligent distribution method of powder additives according to claim 1, characterized in that: The S3 specifically includes: S301: Integrate the predicted consumption speed sequence at each corresponding time step and output the predicted consumption; S302: Calculate the integral time when the predicted consumption is equal to the inventory at the current consumption point, and output the integral time as the available time: in, Indicates the available time, i.e. the credit time. Indicates the inventory quantity at the consumption point, represents the discrete time step, Indicates i Time steps The consumption rate is the consumption per unit time. ,in, k Represents a time index.

6. The intelligent distribution method of powder additives according to claim 1, characterized in that: The specific calculation method of the expected consumption point delivery volume is: in, represents the expected consumption point delivery quantity, Represents the predicted consumption speed sequence t Always predict consumption speed, Indicates the current moment, Indicates the expected availability time. Indicates the delivery time of the consumption point.

7. A powder additive intelligent distribution system, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the intelligent distribution method for powder additives according to any one of claims 1 to 6 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the intelligent distribution method of powder additives as described in any one of claims 1 to 6 is implemented.

Images