Inventory control method and device, nonvolatile storage medium and electronic equipment

Abstract

The invention discloses an inventory control method and device, a nonvolatile storage medium and electronic equipment. The method comprises the steps that based on historical material requirements in a power grid, a target model is adopted for processing, prediction errors of the target model are determined, the target model is obtained through training based on material requirements of different dimensions, and the historical material requirements are material requirements before the current moment; determining a target inventory according to the prediction error and a stockout rate threshold value; based on the current demand at the current moment, the target model is adopted for processing, and a next prediction demand after the current moment is obtained; and determining an inventory adjustment strategy based on the target inventory, the current inventory at the current moment and the next prediction demand. The technical problem that the inventory control accuracy is not ideal in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of inventory control and the field of power operation and maintenance. Specifically, the present invention relates to an inventory control method, device, non-volatile storage medium and electronic device. Background Art

[0002] In the power grid system, material supply is an important guarantee for the construction and operation of the power grid. The management of the power grid material supply chain is an important part of the whole process management of materials such as procurement management, warehousing management, material distribution, warehouse operation, and logistics management of power grid enterprises. It is an important cornerstone for supporting the construction and operation of the power grid and ensuring the safe, reliable, and efficient operation of the power grid.

[0003] The inventory quota is the reasonable inventory quantity determined to ensure the normal progress of power grid operation and maintenance activities, also known as the material reserve quota. The challenges of inventory quota management come from uncertain factors such as demand fluctuations, material shortages, supplier supply delays, and production cycle changes. In related technologies, inventory control for power equipment usually adopts the method of determining inventory based on empirical data. However, the scale of power users participating in the use process of equipment in the power grid is large, and the types and quantities of electrical equipment are numerous. The procurement volume of the power grid is relatively large compared with general enterprises, resulting in a large inventory backlog in the power grid and the problem of unsatisfactory prediction accuracy of inventory quotas.

[0004] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0005] Embodiments of the present invention provide an inventory control method, device, non-volatile storage medium and electronic device to at least solve the technical problem of unsatisfactory inventory control accuracy in related technologies.

[0006] According to one aspect of the embodiments of the present invention, an inventory control method is provided, including: processing historical material requirements in the power grid by using a target model to determine the prediction error of the target model, where the target model is trained based on material requirements in different dimensions, and the historical material requirements are the material requirements before the current moment; determining the target inventory according to the prediction error and the shortage rate threshold; processing the current demand at the current moment by using the target model to obtain the next predicted demand after the current moment; and determining an inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand.

[0007] Optionally, the target model includes a bidirectional neural network and an adversarial network. Based on the historical material requirements in the power grid, the target model is processed to determine the prediction error of the target model, including: processing the historical material requirements using the adversarial network to obtain training data; processing the training data using the bidirectional neural network to obtain a prediction result, where the prediction result includes the temporal correlation of the historical material requirements; and determining the prediction error based on the error between the historical predicted demand indicated by the prediction result and the historical material requirements.

[0008] Optionally, the adversarial network includes a generator and a discriminator. Processing the historical material requirements using the adversarial network to obtain training data includes: processing the historical material requirements to obtain preprocessed data; using the generator to process the preprocessed data to generate simulation data that conforms to the distribution characteristics of the material requirements in the power grid; using the discriminator to perform discriminant processing on the simulation data and the preprocessed data to determine the generation ability of the generator; and in the case where the generation ability meets a predetermined condition, using the generator with the generation ability to obtain training data based on the preprocessed data.

[0009] Optionally, processing the historical material requirements to obtain preprocessed data includes: processing the historical material requirements using a wavelet analysis method to obtain demand data in multiple dimensions, where the demand data in multiple dimensions at least includes: predetermined demand data, operation and maintenance demand data, disaster demand data, and supply chain demand data; and processing based on the demand data in multiple dimensions to obtain preprocessed data.

[0010] Optionally, determining the target inventory according to the prediction error and the out-of-stock rate threshold includes: determining the out-of-stock rate threshold and the power grid safety factor according to the material level of the power grid; determining the predetermined inventory based on the procurement lead time of the power grid and the inventory cycle demand; and determining the target inventory based on the out-of-stock rate threshold, the power grid safety factor, the predetermined inventory, and the prediction error.

[0011] Optionally, determining the inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand includes: determining the next inventory after the current moment; determining the inventory deviation amount based on the current inventory, the next inventory, and the next predicted demand; in the case where the inventory deviation amount is less than or equal to the target inventory, determining the order quantity based on the difference between the target inventory and the current inventory; and determining to execute a procurement action based on the order quantity.

[0012] Optionally, the method further includes: determining the change trend of inventory backlog based on the inventory cycle of the power grid; correcting the out-of-stock rate threshold based on the change trend of inventory backlog to obtain a corrected out-of-stock rate threshold; and performing inventory control processing based on the corrected out-of-stock rate threshold until the end of the inventory cycle, so that the inventory backlog of the power grid is less than or equal to a predetermined inventory backlog threshold.

[0013] According to another aspect of the embodiments of the present invention, an inventory control device is provided, including: a first prediction module, configured to process based on historical material requirements in the power grid by using a target model to determine the prediction error of the target model, where the target model is trained based on material requirements in different dimensions, and the historical material requirements are the material requirements before the current moment; a target inventory determination module, configured to determine a target inventory according to the prediction error and the out-of-stock rate threshold; a second prediction module, configured to process based on the current demand at the current moment by using the target model to obtain the next predicted demand after the current moment; and an inventory adjustment module, configured to determine an inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand.

[0014] According to another aspect of the embodiments of the present invention, a non-volatile storage medium is provided. The non-volatile storage medium stores multiple instructions, and the instructions are suitable for being loaded and executed by a processor to perform the inventory control method of any one of the above.

[0015] According to another aspect of the embodiments of the present invention, an electronic device is provided, including: one or more processors and a memory. The memory is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the inventory control method of any one of the above.

[0016] In the embodiments of the present invention, by processing based on historical material requirements in the power grid by using a target model to determine the prediction error of the target model, where the target model is trained based on material requirements in different dimensions, and the historical material requirements are the material requirements before the current moment; determining a target inventory according to the prediction error and the out-of-stock rate threshold; processing based on the current demand at the current moment by using the target model to obtain the next predicted demand after the current moment; and determining an inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand. The purpose of accurately predicting the target inventory and determining the inventory adjustment strategy is achieved, the technical effect of dynamically controlling the target inventory, reducing the out-of-stock rate and inventory backlog is realized, and thus the technical problem of unsatisfactory inventory control accuracy in the related art is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention, form a part of this application, and the illustrative embodiments and descriptions of the present invention are used to explain the present invention, and do not constitute an improper limitation to the present invention. In the drawings:

[0018] Figure 1 is a flowchart of an optional inventory control method provided according to the embodiments of the present invention;

[0019] Figure 2 It is a schematic diagram of an optional inventory control method provided according to an embodiment of the present invention;

[0020] Figure 3 It is a strategy flowchart of an optional inventory control method provided according to an embodiment of the present invention;

[0021] Figure 4 It is a schematic diagram of an optional inventory control device provided according to an embodiment of the present invention. Detailed implementation manners

[0022] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0024] For the convenience of description, some nouns or terms related to the embodiments of the present application are described below:

[0025] A bidirectional neural network generally refers to a bidirectional recurrent neural network (BRNN, Bidirectional Recurrent Neural Network), which is a deep learning model used to process sequence data. It considers both the forward and reverse directions of the input sequence, which helps the model better process the context and dependencies of the sequence data.

[0026] Generative Adversarial Networks (GANs) is a deep learning model used in unsupervised learning. This model consists of a Generator and a Discriminator. The Generator is used to learn the data distribution and generate fake data similar to the real data. The Discriminator is used to distinguish between the fake data generated by the Generator and the real training data. During the training process, the Generator and the Discriminator play against each other. The Generator tries to generate more realistic samples to test the Discriminator, while the Discriminator tries to more accurately distinguish between real data and generated data. The result of this game is that the Generator can learn the fine features of the data, thus generating high-quality simulation data.

[0027] Markov Decision Process (MDP) is a fundamental framework in dynamic programming and reinforcement learning for dealing with decision-making and planning problems in an uncertain environment. It is based on the Markov property, that is, the future state depends only on the current state and is independent of the past states.

[0028] A Lyapunov function is a function used to prove the stability of a dynamic system. In different application fields, the specific form of the Lyapunov function may vary, but its core idea is to prove the stability of the system by constructing a positive definite function.

[0029] According to an embodiment of the present invention, there is provided an embodiment of a method for inventory control. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0030] Figure 1 It is a flowchart of an optional inventory control method provided according to an embodiment of the present invention. As Figure 1 shown, the method includes the following steps:

[0031] Step S102, based on the historical material requirements in the power grid, process using the target model to determine the prediction error of the target model, where the target model is trained based on material requirements in different dimensions, and the historical material requirements are the material requirements before the current moment;

[0032] It can be understood that historical material requirements in the power grid are collected, and the historical material requirements include, but are not limited to, planned requirements for power construction, maintenance requirements brought about by normal consumption, maintenance requirements caused by natural disasters, and supply chain health maintenance requirements. Using the historical material requirements to train the target model can, according to the variability and uncertainty of power grid material requirements, through training and optimization, enable the target model to have a flexible inventory adjustment level, reduce the out-of-stock rate, and reduce inventory costs.

[0033] In an alternative embodiment, the target model includes a bidirectional neural network and an adversarial network. Based on the historical material requirements in the power grid, the target model is used for processing to determine the prediction error of the target model, including: processing the historical material requirements using the adversarial network to obtain training data; processing the training data using the bidirectional neural network to obtain a prediction result, where the prediction result includes the temporal correlation of the historical material requirements; and determining the prediction error based on the error between the historical predicted demand indicated by the prediction result and the historical material requirements.

[0034] It can be understood that the target model is a hybrid model containing a bidirectional neural network and an adversarial network. The above-mentioned adversarial network is used to help optimize the training of the bidirectional neural network and generate a training set. The above-mentioned bidirectional neural network is mainly used to capture the forward and backward dependencies of time series data, obtain a prediction result, and improve the prediction accuracy. The adversarial network and the bidirectional neural network are used in combination to improve the accuracy and efficiency of inventory quota management. The adversarial network is not only used to generate synthetic training data, but also to optimize the bidirectional neural network for demand prediction, which is beneficial to improving the prediction of random and uncertain demands, thereby improving the prediction accuracy and the robustness of the model. According to the error between the historical predicted demand and the historical material requirements, the prediction error is obtained, which is used for the calculation of safety inventory (i.e., target inventory). By predicting demand through a hybrid model of a bidirectional neural network and an adversarial network, the model parameters can be adjusted and the prediction method can be improved, thereby improving the prediction accuracy of the model.

[0035] Optionally, the processing method of the prediction error can be as follows. The standard deviation to be estimated comes from the cumulative prediction error during the lead time:

[0036]

[0037] where the sampling time i = 1, 2, 3…L, L represents the total cycle, ∈ t is the feature representation of the prediction error at time t, and the target model is assumed to follow a distribution random variable ξ t has an expectation of 0 and a standard deviation of After obtaining the predicted demand at each moment through demand forecasting, combined with the actual demand (i.e., the historical actual demand data of materials), the observed value of the prediction error at each moment is calculated to guide the setting of the safety inventory (i.e., the target inventory) level.

[0038] For the observed value of the prediction error at each specific moment, the following method can be used to calculate the prediction error.

[0039] e t = D t - F t

[0040] where D t represents the actual historical demand at time t, F t represents the historical predicted demand obtained at time t, and e t represents the value of the prediction error obtained at time t.

[0041] The target model includes two parts: a bidirectional neural network and an adversarial network. The adversarial network is used to train the simulated data generated by the generator to reduce the error between the simulated data and the real data and generate more accurate simulated data. Then, the generated simulated data is used to train and predict the target inventory by the bidirectional neural network. The bidirectional neural network considers the time correlation of material demand, making the prediction result more reasonable and accurate. The predicted target inventory is compared with the historical real target inventory to obtain the prediction error of the target inventory.

[0042] In an alternative embodiment, the adversarial network includes a generator and a discriminator. Based on the historical material demand, the adversarial network is used for processing to obtain training data, including: processing the historical material demand to obtain preprocessed data; using the generator to process the preprocessed data to generate simulated data that conforms to the distribution characteristics of material demand in the power grid; using the discriminator to perform discriminant processing on the simulated data and the preprocessed data to determine the generation ability of the generator; when the generation ability meets the predetermined conditions, using the generator with the generation ability to obtain training data based on the preprocessed data.

[0043] It can be understood that the above-mentioned adversarial network includes two parts: a generator and a discriminator. The historical material demand data is preprocessed to eliminate data that does not meet the data requirements of the above-mentioned adversarial network. Based on the preprocessed historical material demand data, the above-mentioned generator generates simulation data that conforms to the material demand distribution characteristics in the above-mentioned power grid. The above-mentioned discriminator is used to discriminate the simulation data generated by the above-mentioned generator, distinguish between the simulation data and the real data, and improve the ability of the generator to generate simulation data. The generation ability of the generator meets the requirements, that is, after the error between the simulation data and the real data reaches a certain threshold, the above-mentioned generator is used to generate a training data set based on the preprocessed historical material demand data. The adversarial network is optimized through the mutual confrontation between the generator and the discriminator, the data generation ability and prediction ability of the adversarial network are enhanced, and the accuracy of the simulation data is improved.

[0044] Optionally, the discriminator has two loss functions lossD and lossT as discrimination criteria. If the discriminator judges the real data as true and the simulated data as false, the loss function decreases. Otherwise, the loss function increases, thereby forcing the discriminator to adjust parameters to improve the discrimination ability. Therefore, the discriminator optimizes the discrimination ability by minimizing the loss functions lossD and lossT, thereby improving the quality of the simulated data generated by the generator. The loss functions lossD and lossT are functions used to measure the accuracy of the discriminator's judgment on real data and simulated data, respectively, where lossD measures the accuracy of the discriminator's judgment on real data, and lossT measures the accuracy of the discriminator's judgment on simulated data. By minimizing the above two loss functions lossD and lossT, the discriminator can continuously optimize its discrimination ability, thereby improving the quality of the simulated data generated by the generator.

[0045] In an optional embodiment, historical material demands are processed to obtain preprocessed data, including: using a wavelet analysis method to process historical material demands to obtain demand data in multiple dimensions, wherein the demand data in multiple dimensions include at least: reservation demand data, operation and maintenance demand data, disaster demand data, and supply chain demand data; processing demand data based on multiple dimensions to obtain preprocessed data.

[0046] It can be understood that the wavelet analysis method is used to separate the material demand dimension components, which are mainly divided into planned demand and random demand. Planned demand includes scheduled demand and supply chain demand, and random demand is disaster demand and operation and maintenance demand. Using these material demand dimensions, the historical material demand is calculated, and the data of the historical material demand data set is preprocessed.

[0047] Optionally, the material requirement D i The formula is:

[0048] D i= D i1 + D i2 + D i3 + D i4

[0049] wherein, D i1 is deterministically calculated for the planned requirements of power construction according to the planned schedule and the material lead time; i1 D i2 is for the maintenance requirements brought about by normal consumption, counting the service life of materials and predicting the maintenance requirements; i2 D i3 is for the maintenance requirements caused by natural disasters, predicting the maintenance caused by disasters by combining long-term weather forecasts and historical information; i4 is the demand for supply chain health maintenance, which is to improve the production level of upstream suppliers and release a certain number of orders during the off-season of production. This value is an empirical value α. When i1 + D i2 + D i3 ≥ α, D i4 = 0; when i1 + D i2 + D i3 < α, D i4 = α - ( i1 + D i2 + D i3 ).

[0050] Step S104, determine the target inventory according to the prediction error and the stockout rate threshold;

[0051] It can be understood that the real material requirements at the current moment and the moments before it are input into the bidirectional neural network to obtain the predicted requirements for the next moment. Then, the predicted requirements are compared with the real requirements to obtain the prediction error at the current moment. After accumulating the prediction errors of the lead time, the standard deviation to be estimated is obtained, and the prediction error observation values at each moment are calculated using the standard deviation to be estimated to guide the determination of the target inventory. The above-mentioned stockout rate is inversely related to the inventory service level and the inventory safety level. The larger the stockout rate, the worse the inventory service level and the lower the inventory safety level. Determine the target inventory according to the prediction error and the stockout rate threshold.

[0052] In an alternative embodiment, determining the target inventory according to the prediction error and the stockout rate threshold includes: determining the stockout rate threshold and the grid safety factor according to the material grade of the power grid; determining the predetermined inventory based on the procurement lead time of the power grid and the inventory cycle demand; determining the target inventory based on the stockout rate threshold, the grid safety factor, the predetermined inventory, and the prediction error.

[0053] It can be understood that the power grid material levels can be divided into six levels, including special level, level 1, level 2, level 3, level 4, and level 5. The higher the material level, the better the inventory service level, the more excellent the service level, the lower the out-of-stock rate, and the higher the power grid safety factor. The target inventory is determined based on the out-of-stock rate threshold of the inventory, the power grid safety factor, the predetermined inventory, and the prediction error.

[0054] Optionally, the gross demand can be determined by predicting the demand during the procurement lead time and the inventory cycle demand. The gross demand can be used to adjust the inventory demand and can be calculated in the following way:

[0055] Gross demand = Forecasted demand during lead time + zα × ss

[0056] Among them, ss is the standard safety inventory (i.e., the predetermined inventory) and can be expressed in the following way:

[0057]

[0058] Among them, the above-mentioned lead time is the procurement lead time. The inventory cycle demand for determining the demand on a daily basis, that is, the daily cycle demand, is statistically processed according to the variance and the mean to obtain the lead time mean, the lead time variance, the daily demand variance, and the daily demand mean.

[0059] Optionally, when initializing the inventory safety level, it is preferably to implement the secondary safety inventory strategy. After the business stabilizes, the daily inventory management strategy is implemented, that is, mainly based on the tertiary safety inventory strategy, and it is alternately adjusted to the secondary or the quaternary depending on the busyness or the storage capacity. The measurement of the predetermined inventory is generally on a monthly basis. It is recommended that the procurement time be in the first ten days of each month. If the procurement plan is executed in the middle ten days, it is recommended to increase the measured quantity by a predetermined first ratio, such as 10%. If the procurement plan is executed in the last ten days, it is recommended to increase the measured quantity by a predetermined second ratio, such as 20%. The above-mentioned predetermined first ratio is less than the predetermined second ratio.

[0060] Step S106, based on the current demand at the current moment, perform processing using the target model to obtain the next predicted demand after the current moment;

[0061] It can be understood that based on the real material demand (i.e., the current demand) at the current moment, it is input into the trained bidirectional neural network to obtain the predicted demand at the next moment (i.e., the next predicted demand).

[0062] In an alternative embodiment, based on the target inventory, the current inventory at the current moment, and the next predicted demand, an inventory adjustment strategy is determined, including: determining the next inventory after the current moment; based on the current inventory, the next inventory, and the next predicted demand, determining the inventory deviation amount; in the case where the inventory deviation amount is less than or equal to the target inventory, determining the order quantity based on the difference between the target inventory and the current inventory; and determining to execute the procurement action based on the order quantity.

[0063] It can be understood that the difference between the upcoming inventory (i.e., the next inventory), the current inventory and the next predicted demand is taken to obtain the inventory deviation amount, and it is determined whether a procurement action needs to be executed. If the inventory deviation amount is less than or equal to the target inventory, the target inventory is updated, and the difference between the updated target inventory and the sum of the current inventory and the upcoming inventory is taken to determine the procurement quantity and execute the procurement action.

[0064] Optionally, the inventory adjustment strategy includes two types: executing a procurement action and not executing a procurement action. If the difference between the upcoming inventory (i.e., the next inventory), the current inventory and the next predicted demand is less than or equal to the target inventory, the target inventory is updated, and the order quantity is determined based on the updated target inventory, the current inventory and the upcoming inventory, and the procurement action is executed; if the difference between the upcoming inventory (i.e., the next inventory), the current inventory and the next predicted demand is greater than the target inventory, no procurement action is executed.

[0065] Step S108, determine an inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand.

[0066] It can be understood that by comparing the difference between the upcoming inventory, the current inventory and the next predicted demand with the target inventory, the inventory adjustment strategy is determined, that is, whether a procurement action needs to be executed. During the execution of the procurement action, a Markov decision process is used to execute the inventory quota management strategy.

[0067] In an alternative embodiment, the method further includes: determining the inventory backlog change trend based on the inventory cycle of the power grid; correcting the out-of-stock rate threshold based on the inventory backlog change trend to obtain the corrected out-of-stock rate threshold; performing inventory control processing based on the corrected out-of-stock rate threshold until the end of the inventory cycle, so that the inventory backlog of the power grid is less than or equal to a predetermined inventory backlog threshold.

[0068] It can be understood that when determining the inventory adjustment strategy, a Markov decision process is used to judge the maximum transition probability. The Markov process has the characteristic of no aftereffect, which simplifies the decision-making process and makes the strategy update process more flexible. During the decision-making process, the service level (i.e., obtained from the out-of-stock rate threshold and characterized) and the reduction of two types of funds (i.e., inventory backlog) are considered as decision evaluation criteria. Since the service level and the reduction of two types of funds may conflict, a Lyapunov function is introduced to balance the objective conflict between the two. Based on the change trend of the power grid inventory cycle, the change trend of the power grid inventory is determined, the out-of-stock rate threshold in different cycles is adjusted, and the target inventory is adjusted using this threshold to balance the out-of-stock rate level and the inventory backlog level.

[0069] It should be noted that the above reduction of the two types of funds refers to reducing accounts receivable and inventory backlogs through reasonable inventory adjustment strategies, and improving the efficiency of fund use. Taking the reduction of the two types of funds as the goal of inventory adjustment can effectively balance the relationship between service level and the reduction of the two types of funds, enabling inventory management to reduce inventory backlogs under the condition of ensuring service level and improving inventory management efficiency.

[0070] Through the above step S102, based on the historical material requirements in the power grid, the target model is processed to determine the prediction error of the target model. The target model is trained based on material requirements in different dimensions, and the historical material requirements are the material requirements before the current moment; step S104, the target inventory is determined according to the prediction error and the stock-out rate threshold; step S106, based on the current demand at the current moment, the target model is processed to obtain the next predicted demand after the current moment; step S108, based on the target inventory, the current inventory at the current moment, and the next predicted demand, the inventory adjustment strategy is determined. It can achieve the purpose of dynamically adjusting the target inventory of the power grid based on the prediction error and the stock-out rate threshold, improving the prediction accuracy, reducing inventory backlogs, achieving the technical effect of improving the accuracy of inventory control, and thus solving the technical problem of unsatisfactory accuracy of inventory control in the related technologies.

[0071] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation manner for precise control of power grid inventory. In the implementation of the present invention, inventory control mainly includes three parts: the construction of the target prediction model, data preprocessing, and the determination of the inventory adjustment strategy. Figure 2 It is a schematic diagram of an optional inventory control method provided according to an embodiment of the present invention. The following will explain the processing flow.

[0072] Construct a hybrid target prediction model of a bidirectional neural network and an adversarial network for data training and prediction of safety inventory. The target prediction model includes an adversarial network and a bidirectional neural network. The adversarial network includes a generator and a discriminator, and the discrimination criterion is to minimize the loss functions lossD and lossT. The adversarial network screens, optimizes, and fills the preprocessed training set data by generating and discriminating the processed data, providing a higher-quality and more accurate data set for the bidirectional neural network. The bidirectional neural network captures the forward and backward dependencies of time series data and trains the network with the data set provided by the adversarial network as input to obtain the target inventory prediction result, improving the prediction accuracy. The adversarial network and the bidirectional neural network are used in combination to enhance the prediction ability for random and uncertain prediction requirements, thereby improving the prediction accuracy and the robustness of the model.

[0073] When performing data preprocessing, the wavelet analysis method is used to divide the material demand dimension into planned demand and random demand. Among them, the planned demand includes the predetermined demand and the supply chain demand, and the random demand includes the disaster demand and the operation and maintenance demand. The historical true material demand is calculated through the above four material demands, and the training data set of the adversarial network is obtained after excluding the data that does not require the use of safety stock. Then, the adversarial network is used to discriminate, screen, and fill the data set to obtain a more high-quality and accurate data set to train the bidirectional neural network, enhance the prediction ability of the network, and improve the prediction accuracy of the target inventory.

[0074] When predicting the target inventory, it can be determined based on the prediction error and the stockout rate threshold. For the prediction error, based on the true material demand at the current moment (i.e., the current demand), the bidirectional neural network is used to obtain the predicted demand at the next moment, and then the prediction error is calculated according to the difference between the true demand and the predicted demand. After accumulating the prediction errors of the lead time, the standard deviation to be estimated is obtained, and the predicted error observation values at each moment are calculated using this standard deviation to be estimated to guide the determination of the target inventory. For the stockout rate threshold, its inventory service level and inventory safety level are inversely related. The larger the stockout rate, the worse the inventory service level and the lower the inventory safety level.

[0075] As shown in Table 1, the daily inventory is divided into six levels, and the special-level inventory strategy is implemented in case of emergencies. It can be seen from Table 1 that the higher the material level, the better the inventory service level, the better the service level, the lower the stockout rate, and the higher the power grid safety factor; then, based on the procurement lead time of the power grid and the inventory cycle demand, the standard safety inventory (i.e., the predetermined inventory) is determined. The measurement of the predetermined inventory is generally carried out on a monthly basis, and the recommended procurement time is the first ten days of each month. If the procurement plan is executed in the middle of the month, it is recommended to increase the measurement by the first predetermined ratio, such as 10%. If the procurement plan is executed in the last ten days of the month, it is recommended to increase the measurement by the second predetermined ratio, such as 20%. The above first predetermined ratio is less than the second predetermined ratio.

[0076] Table 1 Classification Table of Inventory Quota Management

[0077]

[0078] The inventory service level α can be determined by the inventory safety factor z. According to the above preferred correlation, z = 1, corresponding to α = 0.84, that is, z 0.84 = 1. Correspondingly, z 0.85 = 1.04.

[0079] There is also a correlation between the above inventory service level α and the stockout rate threshold. The relationship between the inventory service level α and the stockout rate threshold is:

[0080] Stockout rate threshold = 1 - α

[0081] Figure 3 is a strategic flowchart of an optional inventory control method provided according to an embodiment of the present invention. As Figure 3 shown, when determining the adjustment strategy, the true material demand at the current moment (i.e., the current demand) is used to obtain the predicted demand F at the next moment t+1 ; then, the sum of the incoming inventory and the current inventory is taken the difference with the next predicted demand F t+1 to obtain the inventory deviation amount. The inventory deviation amount is compared with the target inventory S t (i.e., the safety inventory ) to determine the inventory adjustment strategy, that is, whether to execute a procurement action. If the inventory deviation amount is less than or equal to the target inventory, then according to the target inventory S t and update the target inventory to obtain the updated target inventory S' t , and take the difference between the updated target inventory S' t and the sum of the current inventory and the incoming inventory to determine the procurement quantity W t i and execute the procurement action; if the inventory deviation amount is greater than the target inventory, then do not execute the procurement action, that is, W t i = 0. Where i is the ordinal number. For example, there are n kinds of materials, and the i-th kind of material. L is the lead time, and Li is the lead time of the i-th kind of material. t generally refers to the current moment, and t + 1 is the next moment. k is the subscript, representing the statistical quantity in the next few months.

[0082] When determining the inventory adjustment strategy, the Markov decision process is adopted to judge the maximum transition probability. The Markov process has the characteristic of no aftereffect, which simplifies the decision-making process and makes the strategy update process more flexible. In the decision-making process, the service level and the reduction of two funds (i.e., inventory backlog) are considered as the decision evaluation criteria. Since the service level and the reduction of two funds may conflict, the Lyapunov function is introduced to balance the objective conflict between the two. Table 2 shows the target inventory data of the power grid in a certain year (taking 2021 as an example). This data is the target inventory data of 2021 predicted based on the historical material demand data in the previous years (such as 2017 - 2020) of the above-mentioned certain year. Based on the inventory cycle change trend, the power grid inventory change trend is determined, the out-of-stock rate threshold in different cycles is adjusted, and the target inventory is adjusted by using this threshold to balance the out-of-stock rate level and the inventory backlog level. It should be noted that the specific values in Table 2 are only for illustrative purposes and are not specifically limited.

[0083] Table 2 Target inventory data of the power grid in 2021

[0084]

[0085]

[0086] The inventory control method can be applied to a software application, and the software application can adopt the strategy shown in Table 3. As shown in Table 3, inventory management is generally divided into four time nodes, namely initialization, daily, monthly, quarterly, annual reports, and year-end closing and carry-forward. At the initialization time node, the main tasks of inventory management are to determine material classification, determine inventory quota categories, build a material historical pattern model, and initialize data. The business operations are material data maintenance, setting and maintaining the initial and target values of the reduction of two types of funds, setting safety inventory, upper and lower limits of inventory, and maintaining material adaptation model parameters. The inputs and outputs of the model are enterprise resource planning and the wide table of quota management data respectively.

[0087] At the daily time node, the main tasks of inventory management are to maintain outbound and inbound information, monitor reorder points, and calculate economic order quantities. The business operations are to import demand plans and maintain in-transit material data. The inputs and outputs of the model are enterprise resource planning and the inventory quota management ledger respectively.

[0088] At the monthly, quarterly, and annual report time nodes, the main task of inventory management is to maintain macro information of inventory quota management. The business operations are monthly reports (including this month, last month, and next month) of material in and out and plan information, quarterly reports, annual reports (including data at the beginning and end of the year), overall transactions, and analysis of inventory reduction levels. The inputs and outputs of the model are enterprise resource planning and monthly reports, annual reports, and database tables respectively.

[0089] At the year-end closing and carry-forward time node, the main tasks of inventory management are to carry forward the information of inbound and outbound items and correct parameter values. The business operations are to move the historical data time window, maintain model data, and adjust parameters. The inputs and outputs of the model are the inventory quota management table and the new inventory quota management table, historical table respectively.

[0090] Table 3 Time Nodes and Management Strategies of Inventory Quota Management

[0091]

[0092]

[0093] The above optional implementation methods achieve at least the following effects: The wavelet analysis method is used to separate the material demand dimension, improving the data quality of material demand. At the same time, the hybrid prediction model of bidirectional neural network and adversarial network takes into account both the randomness and the planning of material demand, improving the prediction accuracy of the target inventory period. The Markov decision process balances the contradiction between service level and inventory backlog, making the decision result more accurate.

[0094] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0095] In this embodiment, an inventory control device is also provided. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the terms "module" and "device" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0096] According to an embodiment of the present application, an embodiment of a device for implementing an inventory control method is also provided. Figure 4 It is a schematic diagram of an inventory control device according to an embodiment of the present application, as Figure 4 shown, the above-mentioned inventory control device includes: a first prediction module 402, a target inventory determination module 404, a second prediction module 406, and an inventory adjustment module 408. The device will be described below.

[0097] The first prediction module 402 is used to process based on the historical material requirements in the power grid by using a target model, and determine the prediction error of the target model, where the target model is trained based on material requirements in different dimensions, and the historical material requirements are the material requirements before the current moment;

[0098] The target inventory determination module 404 is connected to the first prediction module 402 and is used to determine the target inventory according to the prediction error and the out-of-stock rate threshold;

[0099] The second prediction module 406 is connected to the target inventory determination module 404 and is used to process based on the current demand at the current moment by using the target model to obtain the next predicted demand after the current moment;

[0100] The inventory adjustment module 408 is connected to the second prediction module 406 and is used to determine an inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand.

[0101] In an inventory control device provided by an embodiment of the present application, by setting a first prediction module 402, which is used to process based on historical material requirements in the power grid by using a target model to determine the prediction error of the target model, where the target model is trained based on material requirements in different dimensions, and the historical material requirements are the material requirements before the current moment; a target inventory determination module 404, connected to the first prediction module 402, is used to determine the target inventory according to the prediction error and the shortage rate threshold; a second prediction module 406, connected to the target inventory determination module 404, is used to process based on the current demand at the current moment by using the target model to obtain the next predicted demand after the current moment; an inventory adjustment module 408, connected to the second prediction module 406, is used to determine the inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand. The purpose of improving the inventory prediction accuracy and reducing inventory backlog is achieved, the technical effect of improving the inventory control accuracy is realized, and thus the technical problem of unsatisfactory inventory control accuracy existing in the related technology is solved.

[0102] It should be noted that the above-mentioned various modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following manner: the above-mentioned various modules can be located in the same processor; or, the above-mentioned various modules are located in different processors in any combination.

[0103] It should be noted here that the above-mentioned first prediction module 402, target inventory determination module 404, second prediction module 406, and inventory adjustment module 408 correspond to steps S102 to S108 in the embodiment. The examples and application scenarios implemented by the above-mentioned modules and the corresponding steps are the same, but are not limited to the content disclosed in the above-mentioned embodiment. It should be noted that the above-mentioned modules can run in a computer terminal as part of the device.

[0104] It should be noted that the optional or preferred implementation manners of this embodiment can refer to the relevant descriptions in the embodiment, and will not be repeated here.

[0105] The above-mentioned inventory control device may further include a processor and a memory. The first prediction module 402, target inventory determination module 404, second prediction module 406, inventory adjustment module 408, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to implement corresponding functions.

[0106] The processor contains a kernel, which retrieves the corresponding program units from the memory. One or more kernels can be set. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM), and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash RAM (flash random access memory), and the memory includes at least one memory chip.

[0107] An embodiment of the present application provides a non-volatile storage medium, on which a program is stored, and when the program is executed by a processor, an inventory control method is implemented.

[0108] An embodiment of the present application provides an electronic device, which includes a processor, a memory, and a program stored on the memory and executable on the processor. When the processor executes the program, the following steps are implemented: based on the historical material requirements in the power grid, processing is performed using a target model to determine the prediction error of the target model, where the target model is trained based on material requirements in different dimensions, and the historical material requirements are the material requirements before the current moment; determining the target inventory according to the prediction error and the out-of-stock rate threshold; based on the current demand at the current moment, processing is performed using the target model to obtain the next predicted demand after the current moment; and determining an inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand. The device in this article can be a server, a PC, etc.

[0109] The present application also provides a computer program product, which, when executed on a data processing device, is adapted to execute a program initialized with the following method steps: based on the historical material requirements in the power grid, processing is performed using a target model to determine the prediction error of the target model, where the target model is trained based on material requirements in different dimensions, and the historical material requirements are the material requirements before the current moment; determining the target inventory according to the prediction error and the out-of-stock rate threshold; based on the current demand at the current moment, processing is performed using the target model to obtain the next predicted demand after the current moment; and determining an inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand.

[0110] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0111] The present invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowchart illustrations and / or block diagrams, and combinations of flows and / or blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing apparatus create means for implementing the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0112] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0113] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in multiple blocks.

[0114] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0115] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory. The memory is an example of computer-readable media.

[0116] A computer-readable medium includes permanent and non-permanent, removable and non-removable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic tape disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0117] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0118] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, system, or computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0119] The above are only embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for inventory control, characterized in that: include: Based on the historical material demand in the power grid, a target model is used for processing to determine the prediction error of the target model, wherein the target model is trained based on material demands of different dimensions, and the historical material demand is the material demand before the current moment; Determine the target inventory according to the forecast error and out-of-stock rate threshold; Based on the current demand at the current moment, the target model is used for processing to obtain the next predicted demand after the current moment; An inventory adjustment strategy is determined based on the target inventory, the current inventory at the current moment, and the next predicted demand.

2. The method according to claim 1, characterized in that The target model includes a bidirectional neural network and an adversarial network. The target model is used for processing based on the historical material demand in the power grid to determine the prediction error of the target model, including: Based on the historical material demand, the adversarial network is used for processing to obtain training data; Based on the training data, the bidirectional neural network is used for processing to obtain a prediction result, wherein the prediction result includes a time series association of the historical material demand; The forecast error is determined based on an error between the historical forecast demand indicated by the forecast result and the historical material demand.

3. The method according to claim 2, characterized in that The adversarial network includes a generator and a discriminator. Based on the historical material demand, the adversarial network is used for processing to obtain training data, including: Processing the historical material demand to obtain preprocessing data; The generator is used to process the preprocessed data to generate simulation data that conforms to the material demand distribution characteristics in the power grid; Using the discriminator, performing discriminative processing on the simulation data and the pre-processed data to determine the generation capability of the generator; In the case where the generation capability meets a predetermined condition, the training data is obtained based on the preprocessed data using a generator of the generation capability.

4. The method according to claim 3, characterized in that The processing of the historical material demand to obtain pre-processed data includes: The historical material demand is processed by wavelet analysis method to obtain demand data of multiple dimensions, wherein the demand data of multiple dimensions at least includes: reservation demand data, operation and maintenance demand data, disaster demand data, and supply chain demand data; The pre-processed data is obtained by processing the demand data based on the multiple dimensions.

5. The method according to claim 1, characterized in that The step of determining the target inventory according to the prediction error and the out-of-stock rate threshold includes: Determining the stock-out rate threshold and the power grid safety factor according to the material level of the power grid; Determining the scheduled inventory based on the procurement lead time of the power grid and the inventory cycle requirements; The target inventory is determined based on the out-of-stock rate threshold, the power grid security factor, the scheduled inventory, and the prediction error.

6. The method according to claim 1, characterized in that The determining of the inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand includes: Determining the next inventory after the current moment; Determining an inventory deviation amount based on the current inventory, the next inventory, and the next forecast demand; When the inventory deviation amount is less than or equal to the target inventory, determining the order quantity based on the difference between the target inventory and the current inventory; Based on the order quantity, a purchase action is determined to be performed.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: Determine the inventory backlog change trend based on the inventory cycle of the power grid; Based on the inventory backlog change trend, the out-of-stock rate threshold is revised to obtain a revised out-of-stock rate threshold; Inventory control processing is performed based on the corrected out-of-stock rate threshold until the end of the inventory cycle is reached, so that the inventory backlog of the power grid is less than or equal to a predetermined inventory backlog threshold.

8. An inventory control device, characterized in that: include: A first prediction module is used to determine a prediction error of the target model by processing the historical material demand in the power grid using a target model, wherein the target model is trained based on material demands of different dimensions, and the historical material demand is the material demand before the current moment; A target inventory determination module, used to determine the target inventory according to the prediction error and the out-of-stock rate threshold; A second prediction module is used to process the current demand at the current moment using the target model to obtain the next predicted demand after the current moment; The inventory adjustment module is used to determine an inventory adjustment strategy based on the target inventory, the current inventory at the current moment, and the next predicted demand.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the inventory control method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the inventory control method described in any one of claims 1 to 7.

Images