Class-based optimal inventory management method and device realized based on advanced intelligent algorithm, electronic equipment and storage medium

Abstract

The invention provides a category-based optimal inventory management method and device realized based on an advanced intelligent algorithm, electronic equipment and a storage medium, and the method comprises the steps: carrying out the data screening of first supply chain related data, and obtaining second supply chain related data; obtaining candidate feature data; performing prediction based on the candidate feature data to obtain a target material demand value in an extreme scene; and determining a target function and a constraint condition of inventory management based on the target material demand value, and performing scheme retrieval on a plurality of candidate inventory management schemes according to the target function and the constraint condition to obtain a recommended inventory management scheme. Through data screening, complementation and feature extraction, a reliable basis is provided for demand prediction and scheme optimization; extreme scene data are supplemented, corresponding requirements are predicted, and supply chain toughness is improved; and demand prediction and scheme optimization are realized by means of an advanced intelligent algorithm, manual experience decision is replaced, and the efficiency and optimality of inventory management are greatly improved.

Description

Technical Field

[0001] This application relates to the field of inventory management technology, and in particular to a category-based optimal inventory management method, apparatus, electronic device, and storage medium based on advanced intelligent algorithms. Background Technology

[0002] The power industry involves a wide variety of materials and complex management processes, making traditional experience-based management models ill-suited to handle massive amounts of data and dynamic demands. In recent years, technologies centered on data lakes, the Internet of Things (IoT), and machine learning have been reshaping the logic of building material category management strategy libraries. The following analysis, based on domestic enterprise practices, provides a detailed breakdown from technical architecture to scenario implementation.

[0003] For high-value categories such as power transmission and transformation equipment, China Southern Power Grid has developed an intelligent operation and maintenance control platform that has constructed a full lifecycle strategy library. This system innovatively adopts a three-tiered strategy model: the lifecycle model optimizes overhaul and replacement plans through economic evaluation; the annual model generates component-level maintenance plans by combining equipment importance and lifespan curves; and the real-time model responds to sudden state transitions through a database of six strategy matching mechanisms. By introducing new mechanical performance parameters and dynamic threshold range analysis, the equipment hazard detection rate has been improved to 99%, and the annual plan execution rate has reached 100%, fully demonstrating the strategy library's adaptability to complex scenarios.

[0004] Industry case studies demonstrate that the technical implementation of a materials category management strategy library requires grasping three core points: a data lake + middleware architecture to solve the problem of cross-domain data connectivity; the practice of State Grid Anhui shows that this architecture can improve the efficiency of cross-regional materials allocation by more than 3 times; clustering and model prediction constitute a dual engine for strategy generation; Dezhou Power Supply has reduced the duplication of nine types of materials through this mechanism; and the Internet of Things + unmanned warehouse form an execution closed loop, improving the emergency material response speed by 60%. Currently, these technical solutions have been promoted in eight provincial power grid companies, saving a total of over 120 million yuan in operating costs, providing a replicable technical path for the digital transformation of materials management in the power industry.

[0005] However, the above three-level strategy model of "lifetime - annual - real-time" does not address the risk of missing core data: the model relies on "new mechanical performance parameters" (such as equipment vibration frequency and insulation strength), and these parameters need to be collected in real time by sensors. If the sensors fail (such as the transmission line sensors being damaged by lightning), the real-time model will lose data input and will be unable to identify potential equipment problems. Furthermore, it does not mention "alternative solutions when data is missing" (such as supplementing the model based on historical data of similar equipment), which may lead to a "gap" in the operation and maintenance strategy. This contradicts the achievement of "99% equipment problem identification rate" in the case study and exposes the model's excessive reliance on the stability of the data source. Summary of the Invention

[0006] This application aims to at least partially address one of the technical problems in the related art.

[0007] Therefore, this application proposes a method, apparatus, electronic device, and storage medium.

[0008] One embodiment of this application proposes a category-based optimal inventory management method based on advanced intelligent algorithms, including: Data related to the first supply chain is filtered and supplemented to obtain data related to the second supply chain; Feature extraction is performed on the data related to the second supply chain to obtain candidate feature data; Based on the candidate feature data, a prediction is made to obtain the target material demand value under extreme scenarios; Based on the target material demand value, the objective function and constraints of inventory management are determined. Based on the objective function and constraints, multiple candidate inventory management schemes are searched to obtain a recommended inventory management scheme.

[0009] Optionally, the step of filtering and supplementing the first supply chain-related data to obtain the second supply chain-related data includes: The data related to the first supply chain are uniformly labeled according to their respective regions; Identify the missing values ​​in the first supply chain-related data, supplement the missing values, and supplement the first supply chain-related data under extreme scenarios.

[0010] Optionally, determining the missing values ​​in the first supply chain-related data includes: Remove invalid values ​​from the first supply chain-related data; For the real-time data in the first supply chain-related data after removal, if the proportion of valid data is lower than a preset first proportion threshold, then data is determined to be missing. If the proportion of valid data in the first supply chain-related data after removal is lower than a preset second proportion threshold, then data is determined to be missing.

[0011] Optionally, the process of supplementing the missing values ​​includes: In response to the data missing rate being lower than a preset first missing threshold, data completion is performed using a multi-factor correction based on the average of adjacent time periods and the device operating status; wherein, the data missing rate is determined based on the proportion of valid data; In response to the data missing rate being greater than or equal to the first missing threshold and lower than the preset second missing threshold, a standardized preprocessing and optimization algorithm is used to complete the data. In response to the data missing rate being greater than or equal to the second missing threshold, data completion is performed using multi-source data fusion and long short-term memory network model prediction.

[0012] Optionally, the supplementary first supply chain-related data under extreme scenarios includes: Target events with similarity scores higher than a preset similarity threshold are selected from the meteorological database; Extract the category demand fluctuation coefficient corresponding to the target event; The supplementary data is determined based on the demand fluctuation coefficient of the aforementioned product category.

[0013] Optionally, the step of predicting based on the candidate feature data to obtain the target material demand value under extreme scenarios includes: Prediction is performed based on the candidate feature data to obtain the initial material demand value; The initial material demand value is perturbed and optimized based on the simulated annealing algorithm. An initial temperature, temperature decay strategy and minimum temperature threshold are designed. A new correction coefficient is generated by the dynamic perturbation step size. It is determined whether to accept the new correction coefficient. The globally optimal correction coefficient is saved during the iteration process to obtain the corrected target demand prediction value.

[0014] Optionally, the objective function includes:

[0015]

[0016] in, For inventory holding costs, To replenishment costs, Cost is the cost of stockouts, and Service is the service level. The constraints are as follows: Supply capacity constraints ,in, For the quantity of materials replenished in a single transaction, This represents the supplier's maximum monthly production capacity. Inventory capacity constraints ,in, For the current inventory of material i, This represents the maximum capacity of the warehouse. Service Level Constraints ,in, This is the lower limit of the service level.

[0017] Optionally, the step of retrieving multiple candidate inventory management solutions based on the objective function and constraints to obtain a recommended inventory management solution includes: The candidate inventory management schemes are encoded as chromosomes, and a first number of chromosomes are generated. Through iterative selection, crossover, and mutation operations, a second number of candidate inventory management schemes with the optimal objective function are selected as initial solutions. Based on the initial solution, the probability of accepting inferior solutions is controlled by temperature decay, and the recommended inventory management scheme is finally obtained through convergence.

[0018] Another embodiment of this application proposes a category-based optimal inventory management device based on advanced intelligent algorithms, comprising: The preprocessing module is used to filter and supplement the data related to the first supply chain to obtain the data related to the second supply chain. The feature extraction module is used to extract features from the second supply chain-related data to obtain candidate feature data; The prediction module is used to make predictions based on the candidate feature data to obtain the target material demand value under extreme scenarios. The strategy determination module is used to determine the objective function and constraints of inventory management based on the target material demand value, and to perform a scheme retrieval on multiple candidate inventory management schemes according to the objective function and constraints to obtain a recommended inventory management scheme.

[0019] Another embodiment of this application proposes an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method described in the foregoing aspect.

[0020] Another embodiment of this application proposes a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the foregoing aspect.

[0021] Another embodiment of this application proposes a chip including processing circuitry configured to perform the method described in one aspect above.

[0022] Another embodiment of this application proposes a computer program product that, when executed by a processor, implements the method described in the foregoing aspect.

[0023] The product category-based optimal inventory management method, device, electronic device, chip, and storage medium proposed in this application, based on advanced intelligent algorithms, can achieve the following beneficial effects: Data-driven accuracy: By filtering, completing, and extracting features, it solves the problem of "poor data quality" in traditional inventory management, providing a reliable foundation for demand forecasting and solution optimization; Adaptability to extreme scenarios: Specifically supplements data for extreme scenarios and predicts corresponding demand, avoiding the risk of inventory disruption / backlog under unconventional events and improving supply chain resilience; High efficiency of intelligent algorithms: By leveraging advanced intelligent algorithms to achieve demand forecasting and solution optimization, the efficiency and optimality of inventory management can be significantly improved, replacing human experience-based decision-making.

[0024] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0025] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a category-based optimal inventory management method based on advanced intelligent algorithms, provided as an embodiment of this application; Figure 2 A schematic diagram of the structure of a category-based optimal inventory management device based on an advanced intelligent algorithm provided in this application embodiment; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a chip proposed in an embodiment of this application. Detailed Implementation

[0026] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0027] The following description, with reference to the accompanying drawings, describes a category-based optimal inventory management method, apparatus, electronic device, chip, and storage medium based on advanced intelligent algorithms implemented according to embodiments of this application.

[0028] Figure 1 This is a flowchart illustrating an optimal inventory management method based on product category, implemented using an advanced intelligent algorithm, provided in an embodiment of this application.

[0029] As one implementation, the category-based optimal inventory management method based on advanced intelligent algorithms in this application embodiment can be configured in a category-based optimal inventory management device based on advanced intelligent algorithms. This category-based optimal inventory management device based on advanced intelligent algorithms can be applied to any electronic device so that the electronic device can perform the category-based optimal inventory management function based on advanced intelligent algorithms.

[0030] Among them, electronic devices can be any device with computing capabilities, such as mobile terminals, which can be hardware devices with various operating systems, touch screens and / or displays, such as mobile phones, tablets, personal digital assistants, wearable devices, etc.

[0031] As another implementation, the category-based optimal inventory management method based on advanced intelligent algorithms in this application embodiment can also be executed by a chip with processing capabilities. The chip includes an image signal processing chip (ISP), a central processing unit (CPU), an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a field-programmable gate array (FPGA), a system on a chip (SOC), a reduced instruction set computer (RISC), etc., which will not be listed here.

[0032] It should be noted that all data collection operations related to users in this application are conducted with the user's authorization and in strict compliance with relevant laws and regulations such as privacy and security.

[0033] like Figure 1 As shown, the method may include the following steps: Step 101: Filter and supplement the data related to the first supply chain to obtain the data related to the second supply chain; Step 102: Extract features from the second supply chain-related data to obtain candidate feature data; Step 103: Make a prediction based on the candidate feature data to obtain the target material demand value under extreme scenarios; Step 104: Determine the objective function and constraints for inventory management based on the target material demand value, and perform a scheme retrieval on multiple candidate inventory management schemes according to the objective function and constraints to obtain a recommended inventory management scheme.

[0034] In this embodiment, the advanced intelligent algorithm system is a combination of algorithms covering the entire process of "data preprocessing - feature extraction - demand forecasting - solution optimization", specifically including: Data preprocessing: multi-source data fusion algorithm, LSTM missing value completion algorithm; Feature extraction: Principal Component Analysis (PCA), Random Forest feature importance algorithm; Demand Forecasting: Spatiotemporal Series Forecasting Model (ST-LSTM); Solution optimization: a hybrid algorithm combining genetic algorithm and simulated annealing.

[0035] First-tier supply chain related data: refers to the initial, unfiltered set of raw data collected by the enterprise, specifically including: Demand data: Product category sales and order volume in different regions and time periods; Inventory data: Real-time inventory and historical inventory turnover rate for each warehouse; Supply data: Suppliers' monthly production capacity and replenishment response time; Warehouse data: maximum capacity and storage costs of each warehouse.

[0036] Extreme scenarios: These refer to unconventional events with a probability of occurrence of less than 5% but an impact of more than 80%, specifically categorized as follows: Natural disasters: Regional demand fluctuations caused by typhoons, rainstorms, earthquakes, etc. Social categories: Sudden outbreaks of epidemics and large-scale sporting events leading to a surge in demand for supplies; Industry-specific: Policy adjustments (such as environmental protection-related production restrictions) and changes in supply constraints caused by supply chain disruptions.

[0037] Explanation of principles This method achieves intelligent management of the entire process "from data to solution" through four core steps. Taking the regional inventory management of a fast-moving consumer goods company as an example: Data filtering and supplementation (first → second supply chain data) Unified regional labeling: The "Chengdong Warehouse" in City A and City B will be labeled as "330102 - Chengdong Warehouse (City A)" and "310105 - Chengdong Warehouse (City B)" respectively (the first 6 digits are the administrative region code), eliminating data confusion caused by duplicate regional names; Missing value completion: For a certain period of time in City A where demand data is missing (only 70% of the data is valid), the demand data for that period is completed by using "the average of the three adjacent days + the operating status of the sales terminal equipment (normal)". Extreme scenario data supplementation: Historical typhoon events in City A were screened from the meteorological database (90% similarity), and the fluctuation coefficient of fast-moving consumer goods demand under these events (2.3 times) was extracted to complete the "Demand data of City A under typhoon scenarios".

[0038] Feature extraction (second data → candidate feature data) uses the "random forest feature importance algorithm" to extract core features from the second data, including: Demand-related factors: 7-day demand volatility, demand amplification factor in extreme scenarios; Supply-related factors: Supplier capacity utilization rate, replenishment response time; Inventory-related: Current inventory turnover rate, warehouse capacity utilization rate.

[0039] Redundant features such as "sales terminal number" are filtered out to obtain candidate feature data.

[0040] Demand prediction in extreme scenarios (candidate features → target demand value) Based on the ST-LSTM model, the input candidate feature data yields "Weekly demand value of fast-moving consumer goods in City A under normal scenarios (1000 boxes)"; The simulated annealing algorithm was used for optimization: the initial temperature was set to 100, the temperature decay coefficient was 0.95, and the minimum temperature was 10. The initial demand value was dynamically perturbed (the step size was gradually reduced from 5% to 1%), and after 50 iterations, the globally optimal correction coefficient (1.8) was obtained. Finally, it was corrected to "the weekly demand value of fast-moving consumer goods in City A under the typhoon scenario (1800 boxes)".

[0041] Solution retrieval and optimization (target requirements → recommended solution): Construct an objective function of "minimizing cost + maximizing service level", combined with constraints of "supplier monthly capacity (2000 boxes), warehouse capacity (2500 boxes), and service level minimum (95%)", and use a genetic algorithm for initial screening (generating 100 candidate solutions, iterating 20 times and retaining 10 better solutions), and then use simulated annealing to converge to obtain the recommended solution: "single replenishment quantity of 1500 boxes, and inventory threshold set at 500 boxes".

[0042] Beneficial effects At the data level: it solves the problems of "data fragmentation, missing data, and single scenario" in traditional inventory management, and improves the completeness of second supply chain data by 80%, providing a reliable foundation for subsequent decision-making; In terms of forecasting: the demand forecasting error under extreme scenarios has been reduced from the traditional 30% to less than 5%, avoiding the risk of inventory shortages under unconventional events; At the solution level: the efficiency of solution retrieval has been improved from "3 days / category" to "1 hour / 10 categories" by manual methods, while achieving a two-way optimization of "cost reduction of 15% + service level improvement of 98%", taking into account both economic efficiency and supply chain resilience.

[0043] Optionally, the step of filtering and supplementing the first supply chain-related data to obtain the second supply chain-related data includes: The data related to the first supply chain are uniformly labeled according to their respective regions; Identify the missing values ​​in the first supply chain-related data, supplement the missing values, and supplement the first supply chain-related data under extreme scenarios.

[0044] In this embodiment, warehouse data from 30 cities across the country are collected using a unified regional tagging system and re-tagged according to the rule of "administrative region code + data type + timestamp". Original label: "Beijing Warehouse - 20230801 - Inventory" → New label: "110100-02-20230801"; Original label: "Shanghai Sales Point - 20230801 - Demand" → New label: "310100-01-20230801".

[0045] The tagged data is stored in a unified database, supporting one-click retrieval of data across regions.

[0046] Missing value imputation was performed to validate the marked Beijing warehouse inventory data. Remove invalid values: Filter out the outlier "Inventory = -100 boxes" (which does not conform to business logic); Missing data detected: This data is real-time data, and the effective data ratio is only 85% (below the first ratio threshold of 90%), therefore it is determined to be missing data; Complete missing values: Use "average inventory of 3 adjacent days (500 boxes) + warehouse temperature and humidity status (normal)" to complete the missing data (500 boxes).

[0047] Additional data on extreme scenarios, specifically for heavy rain scenarios in the Beijing area: The National Meteorological Science Data Center screened "Beijing 7 in 2012" 21 "Rainstorm" events (92% similarity); Extract the demand volatility coefficient (2.1 times) and supplier capacity reduction ratio (30%) for the retail category under this event. Based on the current normal demand in Beijing (1000 boxes / day) and supplier capacity (2000 boxes / day), demand data (2100 boxes / day) and supply data (1400 boxes / day) under the rainstorm scenario are calculated and added to the second supply chain data.

[0048] Beneficial effects Regional data standardization: eliminated the confusion in cross-regional data identification, and improved data retrieval efficiency from "2 hours / time" to "10 seconds / time"; Improved data integrity: After missing values ​​were filled in, the data validity rate increased from 75% to 98%, providing high-quality data for subsequent predictions; Comprehensive scenario coverage: The addition of extreme scenario data enables the inventory solution to cope with unconventional events, increasing the emergency adaptability rate of the solution from 30% to 95%.

[0049] Optionally, determining the missing values ​​in the first supply chain-related data includes: Remove invalid values ​​from the first supply chain-related data; For the real-time data in the first supply chain-related data after removal, if the proportion of valid data is lower than a preset first proportion threshold, then data is determined to be missing. If the proportion of valid data in the first supply chain-related data after removal is lower than a preset second proportion threshold, then data is determined to be missing.

[0050] In this embodiment, the traditional data missing detection method suffers from the problem of "not distinguishing between two types": Indiscriminate use of different data types: Using the same threshold to determine missing data for real-time data (such as the current day's demand) and historical data (such as the demand from the same period last year) leads to "a small amount of missing real-time data being ignored, and a large amount of missing historical data being over-completed"; The system fails to distinguish between invalid and missing values: it misclassifies "demand is 0" (valid data) as a missing value or ignores cases where "demand has been empty for 5 consecutive days" (true missing value), resulting in a missing value detection accuracy of less than 60%.

[0051] These issues resulted in a data deviation of over 20% after completion, directly impacting the accuracy of demand forecasting.

[0052] Glossary First proportion threshold: The lower limit of the effective data proportion for real-time data. It is set according to the characteristics of the industry. In the FMCG industry, it is usually 90% (because real-time demand affects the replenishment decision on the same day, requiring higher data integrity).

[0053] The second proportion threshold is the lower limit of the effective data proportion for historical data, which is usually 70% (because historical data is used for trend analysis, a small number of missing data has a small impact on the overall data).

[0054] Valid data ratio: The calculation formula is "number of valid data records ÷ total number of data records × 100%", where valid data refers to values ​​that conform to business logic (such as demand ≥ 0, inventory ≤ warehouse capacity).

[0055] Explanation of principles Taking the determination of missing supply chain data of an electronics manufacturing company as an example: Collect real-time demand data (12 time periods per day) and historical demand data (120 time periods from the same period last year) for a certain product category after removing invalid values: Real-time data: Includes "Demand = -50 units" (invalid value) and "Demand = 0 units" (valid value); Historical data: Includes "Demand = 1000 units" (valid value) and "Demand = 1500 units (exceeding maximum capacity)" (invalid value).

[0056] After removing the invalid values ​​mentioned above, a dataset containing only valid data is obtained.

[0057] After missing data detection and processing, the real-time data consists of 10 valid time periods (out of a total of 12 time periods): Valid data percentage = 10 ÷ 12 × 100% ≈ 83.3%; If the percentage is lower than the first percentage threshold (90%), it is determined to be "real-time data missing".

[0058] After handling the missing data detection, there are a total of 80 valid historical time periods (out of a total of 120 time periods): Valid data percentage = 80 ÷ 120 × 100% ≈ 66.7%; If the percentage is lower than the second percentage threshold (70%), it is judged as "missing historical data".

[0059] Beneficial effects Precise data type differentiation: This avoids misjudging real-time data from historical data, and improves the accuracy of real-time data missing data detection from 60% to 95%. Invalid and missing values ​​are separated: the calculation of the proportion of valid data is more accurate, and the error in missing value detection is reduced from 25% to less than 5%; Reasonable allocation of completion resources: Different completion strategies can be adapted to address different types of missing data, thereby reducing completion costs.

[0060] Invalid values ​​(such as out-of-limit values ​​caused by sensor malfunctions or system default values ​​during equipment downtime) must be removed through data preprocessing before missing value detection is performed. - Real-time data (such as SCADA equipment consumption, sensor real-time readings): Based on the equipment data acquisition frequency (usually 1-5 minutes / time), data is considered missing if there are no valid updates for 12 consecutive times or more (i.e., a cumulative total of 1 hour). For example, if a transformer oil temperature sensor uploads one data point every 5 minutes, and there are no data updates for 12 consecutive acquisition points (60 minutes), it is marked as missing data.

[0061] -Historical data (such as monthly demand, equipment consumption ledgers): Using the calendar month as the statistical period, days with a cumulative total of 0 valid data records for ≥3 days, or a monthly valid data percentage of less than 90% (i.e., the equivalent duration of missing data is ≥3 days), are considered missing. For example, if a certain product category has 3 days in May with no outbound records at all, or only 1 record per day (normally it should be 4 records / day), resulting in a valid data percentage of 82%, both are considered missing.

[0062] Optionally, the process of supplementing the missing values ​​includes: In response to the data missing rate being lower than a preset first missing threshold, data completion is performed using a multi-factor correction based on the average of adjacent time periods and the device operating status; wherein, the data missing rate is determined based on the proportion of valid data; In response to the data missing rate being greater than or equal to the first missing threshold and lower than the preset second missing threshold, a standardized preprocessing and optimization algorithm is used to complete the data. In response to the data missing rate being greater than or equal to the second missing threshold, data completion is performed using multi-source data fusion and long short-term memory network model prediction.

[0063] In this embodiment, the first missing threshold is the upper limit of the low missing rate, which is usually set to 10% (applicable to simple completion strategies).

[0064] Second missing threshold: The lower limit of high missing rate, usually set to 30% (applicable to complex completion strategies).

[0065] Multi-factor correction of equipment operating status: The complete value is corrected based on the operating status of the data acquisition equipment (such as normal, fault, low battery). For example, when the equipment is faulty, the complete value needs to be reduced by 10% (because the fault may cause the data acquisition to be too low).

[0066] Standardized preprocessing includes steps such as "data normalization (scaling values ​​to the 0-1 range) and noise reduction (filtering out random fluctuations)" to improve the accuracy of algorithm completion.

[0067] Multi-source data fusion: Integrating related data from the same region, such as when supplementing demand data for a certain product category, fusion of population flow data and competitor sales data from that region.

[0068] Explanation of principles Taking the completion of warehouse data of a logistics company as an example: Imputation with low missing rate (missing rate 8% < first missing threshold 10%): The missing rate of inventory data in a certain warehouse is 8%. The "average of adjacent time periods" method is used: the average inventory of the three time periods before and after the missing time period (1000 boxes) is taken. Equipment status correction: The warehouse temperature and humidity equipment shows "normal" during this period and no correction is required; The final completion value was 1000 boxes, and the completion time was only 2 seconds.

[0069] Imputation based on missing data rate (10% ≤ missing data rate 25% < second missing data threshold 30%): A warehouse's required data has a missing data rate of 25%. Standardized preprocessing: Normalize the demand data to the 0-1 range and filter out random fluctuations; Optimized algorithm completion: A linear regression algorithm is used, with "date and regional population" as independent variables, to complete the missing required data; The deviation of the completed data is controlled within 10%.

[0070] High missing rate completion (missing rate 40% ≥ second missing threshold 30%): The missing rate of supply data in a certain warehouse is 40%. Multi-source data fusion: Integrating the supplier's historical production capacity data and the production capacity data of suppliers in the same industry; LSTM model prediction: Using the fused multi-source data as input, train an LSTM model to predict missing supply data; The deviation of the completed data is controlled within 5%.

[0071] Beneficial effects Reduced completion costs: Using a simple strategy for data with low missing rates, computational costs are reduced by 60%; Improved completion accuracy: For high missing rate data, LSTM + multi-source fusion reduced completion bias from 30% to 5%; Business scenario adaptation: By combining device status and multi-source data, the completed results are more in line with the actual business situation.

[0072] Hierarchical completion strategy Differential imputation is implemented based on the missing data rate (missing data volume / total data volume to be collected). After imputation, the effectiveness must be verified by "residual test" (absolute value of residual ≤ mean ± 2 standard deviations). - Mild missing data (missing rate < 10%): The missing data is filled using "average of adjacent time periods + multi-factor correction of equipment operating status". First, take the average of the valid data from the two windows of equal duration before the missing time period (e.g., if 2 hours are missing, take the first 2 hours and the last 2 hours), then multiply it by the equipment real-time status correction coefficient (correction factors include load rate, ambient temperature, and operating mode, with weights of 60%, 30%, and 10% respectively). Example: A 10kV transformer current sensor is missing 2 hours of data. The average for the first 2 hours is 80A, and the average for the last 2 hours is 84A. The real-time load rate is 90% (standard load rate is 100%), and the ambient temperature is 35℃ (base temperature is 25℃, and the correction coefficient decreases by 5% for every 10℃ increase). Then the filled value = (80 + 84) / 2 × 90% × (1 - (35 - 25) × 5% / 10) = 82 × 0.9 × 0.95 = 69.69A.

[0073] - Moderate missing data (10% ≤ missing rate < 30%): Imputation is performed using "standardized preprocessing + optimized KNN algorithm". First, the data is Z-score standardized (to eliminate the influence of units). Then, based on "Euclidean distance + device similarity", 5 best neighbors are selected (same type of equipment, power deviation ≤ 5%, service life difference ≤ 2 years, maintenance frequency deviation ≤ 10%). The weighted average is calculated according to the similarity weight (similarity = 1 / Euclidean distance, weights sum to 1). Example: A water pump has a 15% missing flow rate. Five similar water pumps with the same flow rate during the same period were selected: 12 m³ / h, 11.8 m³ / h, 12.2 m³ / h, 11.7 m³ / h, and 12.1 m³ / h. The similarity weights are 0.22, 0.21, 0.23, 0.18, and 0.16, respectively. The completed value is 12 × 0.22 + 11.8 × 0.21 + 12.2 × 0.23 + 11.7 × 0.18 + 12.1 × 0.16 = 11.98 m³ / h.

[0074] - Severe missing data (missing rate ≥ 30%): Imputation is achieved using "multi-source data fusion + LSTM model prediction". First, backup data sources are activated (manual readings from manual inspection records, concurrent data from similar equipment within a 3km radius of the same area, and related consumption data from the supply chain system). These are then integrated with existing complete data to form a feature set (including time-series features, equipment attribute features, and environmental features), which is then input into the LSTM model for prediction imputation. The model employs an "offline pre-training + online incremental training" mode, updating the training set daily to adapt to dynamic data changes. The prediction error must be controlled within SMAPE ≤ 8%.

[0075] Optionally, the supplementary first supply chain-related data under extreme scenarios includes: Target events with similarity scores higher than a preset similarity threshold are selected from the meteorological database; Extract the category demand fluctuation coefficient corresponding to the target event; The supplementary data is determined based on the demand fluctuation coefficient of the aforementioned product category.

[0076] In this embodiment, fluctuations in demand for supplies under extreme scenarios represent a "black swan" risk in supply chain management. Traditional inventory management relies on normal historical data and lacks a correlation model between "extreme events and demand fluctuations," leading to significant discrepancies between demand forecasts and actual demand when extreme scenarios occur. For example, during a sudden outbreak of an epidemic, the demand for protective equipment may surge tenfold, but such scenarios are not recorded in regular data, rendering inventory plans completely ineffective. The core pain point lies in the lack of methods to extract scenario characteristics from external event databases and quantify demand fluctuations, thus failing to provide effective data support for inventory decisions under extreme scenarios.

[0077] Glossary Meteorological Database / Event Database: A structured database that stores historical extreme events (such as typhoons, rainstorms, and epidemics), including fields such as "spatiotemporal attributes (occurrence area, duration), impact intensity (such as typhoon level, rainfall), and related data (such as changes in material demand under this event)".

[0078] Similarity matching dimensions: used to measure the similarity between historical events and the current scene, including "spatial dimension (the degree of matching between the administrative level of the event location and the current location), temporal dimension (the degree of matching between the duration of the event and the prediction period), and intensity dimension (the degree of matching of the quantitative impact of the event)".

[0079] Category demand fluctuation coefficient: The ratio of actual demand for a certain category of goods to normal demand under extreme scenarios is a core indicator for quantifying the degree of demand amplification / contraction under extreme scenarios.

[0080] Explanation of principles The core of this claim is to supplement extreme scenario data through the theoretical link of **"event matching - fluctuation coefficient extraction - data completion"**, and its theoretical logic is based on the "reusability of demand fluctuations in historical extreme events": The theoretical basis for target event selection is that the impact of extreme events exhibits a correlation of "region-intensity-duration," meaning that extreme events in the same region with similar intensity and duration exhibit consistent demand fluctuation patterns for the same category of goods. Therefore, by calculating similarity across three dimensions—"space + time + intensity" (such as the cosine similarity algorithm)—historical events with a matching degree ≥ a threshold to the current scenario are selected from the event database to ensure their reference value for demand fluctuations.

[0081] The theoretical logic behind the extraction of the volatility coefficient is that there is a positive correlation between the demand for a product category and the intensity of the impact of extreme events (e.g., the higher the typhoon level, the greater the volatility coefficient of the demand for flood control materials). By statistically analyzing the values ​​of the "normal demand" and "extreme scenario demand" for this product category during historical target events, the volatility coefficient is calculated. This coefficient is a quantitative amplification factor for the demand in extreme scenarios, and its theoretical basis is the "linear relationship between demand elasticity and event intensity".

[0082] The theoretical method for generating supplementary data is to take the normal demand data of the current scenario as a benchmark and multiply it by the fluctuation coefficient to obtain the demand data under extreme scenarios. Its theoretical essence is "demand extrapolation based on historical patterns" - that is, assuming that the demand fluctuation pattern of this category of goods is consistent with the historical target events under the current scenario, thereby generating supplementary data that conforms to extreme scenarios.

[0083] Beneficial effects Theoretical level: A quantitative correlation model of "extreme events - demand fluctuations" was established, which solved the theoretical gap of lack of data on extreme scenarios in traditional inventory management; In practice: The supplemented extreme scenario data enables inventory plans to adapt to unconventional demands in advance, reducing demand forecasting deviation in extreme scenarios from over 30% to less than 5%. In terms of efficiency: Based on the reusability of historical events, the time cost of data completion has been reduced from "1 week of manual analysis per scenario" to "1 hour of automatic generation by algorithm per scenario".

[0084] For scenarios where historical data is scarce, such as once-in-a-century cold waves, typhoons, and heavy rainfall, a "historical similar event matching + Monte Carlo simulation" approach is used to generate highly reliable virtual demand data. - Similar event matching: Screening events with a similarity of ≥85% from the meteorological database over the past 50 years (matching dimensions include disaster type, duration, affected area, and extreme value intensity, such as the minimum temperature of a cold wave and the maximum wind speed of a typhoon). - Basic parameter extraction: Extract core parameters such as the category demand fluctuation coefficient during similar events (e.g., the demand for heating equipment increases by 300% during a cold wave), inventory consumption rate, and replenishment delay days; - Monte Carlo simulation: Using the baseline parameters as the mean, a random fluctuation range of ±20% is set to generate 1000 sets of virtual demand sequences (covering daily demand, hourly consumption, and peak demand times). After removing outlier sequences exceeding 3 standard deviations, the mean is used as the complete data. For example, the average daily demand for heaters matched to three strong cold wave events is 120, 115, and 125 units respectively. After 1000 simulations, the final virtual average daily demand is 119 units, and the peak demand is 180 units (occurring in the early morning of the third day of the event).

[0085] Optionally, the step of predicting based on the candidate feature data to obtain the target material demand value under extreme scenarios includes: Prediction is performed based on the candidate feature data to obtain the initial material demand value; The initial material demand value is perturbed and optimized based on the simulated annealing algorithm. An initial temperature, temperature decay strategy and minimum temperature threshold are designed. A new correction coefficient is generated by the dynamic perturbation step size. It is determined whether to accept the new correction coefficient. The globally optimal correction coefficient is saved during the iteration process to obtain the corrected target demand prediction value.

[0086] In this embodiment, 12 core features, including "monthly consumption volatility", "supplier delivery delay rate", and "extreme weather impact coefficient", are extracted as inputs for subsequent prediction.

[0087] Demand forecasting correction (addressing the defect of "forecast failure in extreme scenarios") Basic Forecast: Outputs monthly material demand forecasts (Dbase) under typical scenarios using an LSTM model. Algorithm Correction: Dbase is perturbed and optimized using simulated annealing algorithm. The core of this optimization is... The "temperature regulation exploration-convergence" process identifies the optimal correction coefficient δ for demand forecasting under extreme scenarios (such as cold waves and typhoons).

[0088] (1) Temperature parameter design (core control variable) The "temperature" of simulated annealing determines the intensity of solution space exploration. Parameters must be designed in conjunction with the characteristics of power demand fluctuations (±20% in extreme scenarios) to avoid overexploration or premature convergence. 1. Initial temperature T0 The initial temperature is set to T0 = 500, and the value is determined based on: -Referring to the maximum fluctuation range required in extreme scenarios (±20%), T0 needs to be large enough to cover the entire fluctuation range, ensuring that the initial stage can accept "inferior solutions" with large deviations (such as the prediction deviation corresponding to δ=0.2), and avoiding missing the optimal solution; -Preliminary experiments verified that when T0 < 400, the probability of missing the optimal value of δ (approximately 0.3) under extreme cold wave scenarios reaches 30%; when T0 = 500, the omission probability drops to < 5%, balancing exploration efficiency and accuracy.

[0089] 2. Temperature decay strategy An exponential decay function is used to control the temperature decrease, balancing "expansion breadth" and "convergence speed," as shown in the following formula: T k+1 = λ * T k; where: T k Let T0 be the temperature of the k-th iteration, T0 = 500; the attenuation coefficient λ = 0.95, and the temperature decreases by 5% after each iteration. - Early stage (T>100): The temperature drops slowly, preserving the ability to explore a wide range of δ (e.g., δ can quickly jump between -0.2 and 0.2), adapting to the uncertainty of extreme scenario requirements; - Later stage (T≤100): The temperature decreases faster, and the search is gradually focused on the fine search around the optimal δ (e.g., δ is fine-tuned between 0.15 and 0.25) to avoid invalid iterations.

[0090] 3. Minimum temperature threshold T min Set T min =10 -3 When T k <T min Stop temperature decay at time: - At this point, the temperature is extremely low, and the probability of accepting a "bad solution" is close to 0 (e.g., if the deviation increases by 0.01, the probability of acceptance is <10). -5 The algorithm has converged to a stable solution; - Avoid excessively low temperatures that could cause iterations to "stagnate" (e.g., T=10). -4 (At that time, only 1 to 2 δ values ​​can be explored in 100 iterations), thus improving computational efficiency.

[0091] (2) Optimization of the disturbance mechanism (adapting to the scenario of demand correction) The perturbation objective is to generate new correction coefficients δ new The step size needs to be dynamically adjusted based on the current temperature to ensure the effectiveness and rationality of the exploration. 1. Constraints on the range of values ​​for δ Based on demand statistics for extreme scenarios in the power industry (such as a 30% increase in cable demand due to a cold wave and a 25% increase in demand for tower accessories due to a typhoon), δ is limited. [-0.2, 0.3]: - Negative δ (-0.2~0): Adapts to special situations where demand decreases in extreme scenarios (such as high temperatures causing some outdoor maintenance to be suspended, resulting in a decrease in the demand for insulation materials). - Positive δ (0~0.3): Demand growth covering the vast majority of extreme scenarios (cold waves, typhoons, blizzards).

[0092] 2. Dynamic perturbation step size formula New correction factor δ new From the current δ current The temperature-related perturbation term is constituted by the following formula: ,in: rand() generates a random number uniformly distributed in the range [0,1]. (rand() * 2 - 1) achieves "symmetric positive and negative perturbation"; the step size coefficient α = 0.0004, calibrated through pre-experimentation. -When T k When the initial temperature is 500, the perturbation step size is approximately 0.2 (500 * 0.0004), and δ can jump within a wide range of [-0.2, 0.3] to quickly explore extreme values; -When T k When the value is 10 (later temperature), the perturbation step size is approximately 0.004, and δ is only slightly adjusted near the current value (e.g., δ...). current =0.2, δ new Fine convergence is achieved (between 0.196 and 0.204).

[0093] 3. Boundary handling rules If δ new If the value exceeds the range of [-0.2, 0.3], use "reflection boundary" correction: -Example 1: delta current =0.3, δ after perturbation new =0.31, then corrected to δ new =0.29 (same magnitude reverse offset); - Example 2: delta current =-0.2, δ after perturbation new =-0.22, then corrected to δ new =-0.18, to ensure that δ always conforms to the actual needs logic.

[0094] (3) Determine whether to accept the new solution δ using the Metropolis criterion. new The core principle is "good solutions must be accepted, and inferior solutions are accepted according to probability," balancing exploration and convergence. 1. Definition of the objective function (energy function) With the objective of minimizing the relative deviation between the predicted value and the actual requirements of the simulated extreme scenario, the energy function E (deviation value) is defined as follows: ,in: - D pred =D base * (1+δ) (corrected forecast demand); - D actual To "simulate real-world demand in extreme scenarios": Using actual demand data from similar historical events (such as the 2021 cold wave), Monte Carlo simulations were used to generate 100 sets of virtual real-world values, and the average was taken as D. actual (e.g., cable D in a cold wave scenario) actual =120 sets).

[0095] 2. There are two cases regarding the acceptance logic. Let the current solution energy be E current (corresponding to δ current ), and the new solution energy be E new (corresponding to δ new ): - Case 1: E new ≤ E current (the new solution is better), directly accept δ new , update the current solution to δ current = δ new , E current = E new ; Example: E current = 0.08 (δ = 0.15), E new = 0.06 (δ = 0.18), directly accept (δ = 0.18).

[0096] - Case 2: E new > E current (the new solution is a worse solution) Calculate the acceptance probability P. If rand() < P, then accept; otherwise, reject. The formula is as follows: , the physical meaning of the acceptance probability: the higher the temperature, the higher the tolerance for "worse solutions", avoiding premature convergence to local optima; the lower the temperature, the lower the tolerance, gradually converging to the optimal solution. Example (cold wave scenario): - Current T k = 100, E current} = 0.08 (δ = 0.15), E new = 0.10 (δ = 0.22); - Acceptance probability P = exp(-(0.10 - 0.08) / 100) = exp(-0.0002) ≈ 0.9998, likely to accept, retaining the possibility of exploring larger δ; - If T k = 10, P = exp(-0.02 / 10) = 0.9802, the acceptance probability decreases, only a small probability of accepting a larger δ deviation.

[0097] 3. Optimal solution preservation mechanism During the iteration process, the "global optimal solution" is saved in real-time (that is, δ corresponding to the historical minimum E best ), even if a worse solution is accepted later, it will not overwrite the global optimal solution, ensuring that the final output δ is the optimal value throughout the iteration.

[0098] Optionally, the objective function includes:

[0099]

[0100] in, For inventory holding costs, To replenishment costs, Cost is the cost of stockouts, and Service is the service level. The constraints are as follows: Supply capacity constraints ,in, For the quantity of materials replenished in a single transaction, This represents the supplier's maximum monthly production capacity. Inventory capacity constraints ,in, For the current inventory of material i, This represents the maximum capacity of the warehouse. Service Level Constraints ,in, This is the lower limit of the service level.

[0101] In this embodiment, traditional inventory management often focuses solely on cost control, neglecting the impact of service levels on supply chain stability. Furthermore, the definitions of constraints are typically vague, making it difficult to adapt to inventory optimization needs across multiple product categories and constraints. For example, some inventory solutions only aim for the lowest total cost, but fail to consider supplier capacity limits or warehouse capacity constraints, leading to replenishment failures in actual implementation; or, due to a lack of service level constraints, frequent stockouts impact downstream demand. In addition, existing objective functions lack a clear breakdown of cost components, failing to accurately quantify the weighting relationships of the three core costs—inventory holding, replenishment, and stockouts—and making it difficult to balance the conflict between cost and customer satisfaction through service level indicators. Therefore, there is an urgent need for an objective function that simultaneously encompasses both cost and service level, along with clear multi-dimensional constraints, to support the precise optimization of inventory management solutions.

[0102] Objective function: This method contains a combined function of two core objectives: minimizing total cost (minCost) and maximizing service level (maxService). By quantifying the relationship between cost and service level, it achieves bidirectional optimization of inventory management.

[0103] Inventory holding costs ( ): Costs incurred to maintain inventory, including storage fees, interest on capital tied up, inventory losses, and other expenses directly related to the quantity of inventory.

[0104] Replenishment costs ( ): Costs incurred each time a replenishment is initiated, including procurement fees, logistics and transportation costs, supplier communication costs, and other expenses related to the frequency / batch of replenishment.

[0105] Out-of-stock costs ( Costs incurred due to insufficient inventory to meet demand, including customer churn losses, premiums for emergency restocking, and compensation for order defaults.

[0106] Service Level: A measure of the ability of inventory to meet demand. The higher the service level, the lower the risk of stockouts at the point of cost.

[0107] Supply capacity constraints: Limiting the quantity of a single material replenishment to no more than the supplier's maximum monthly production capacity to prevent suppliers from being unable to deliver due to over-capacity replenishment.

[0108] Inventory capacity constraint: This constraint limits the current inventory of a single item to no more than the maximum capacity of the warehouse, preventing inventory buildup from exceeding the warehouse's carrying capacity.

[0109] Service level constraint: This constraint requires that the actual service level not be lower than the preset lower limit, ensuring a minimum level of customer satisfaction for the inventory solution.

[0110] Total cost is the sum of inventory holding costs, replenishment costs, and stockout costs. The goal is to minimize this value to control inventory input. Service level is calculated by subtracting the proportion of stockout costs to total costs from 1. The goal is to maximize this value to improve demand fulfillment rate. This design binds "cost control" with "service assurance" to avoid extreme decisions under a single objective (such as sacrificing service to reduce costs or excessively stockpiling inventory to ensure service).

[0111] Supply capacity constraints: By limiting the quantity of each replenishment of a single item, we ensure that the replenishment quantity is within the supplier's production capacity, thus avoiding order non-fulfillment. Inventory capacity constraints: By limiting current inventory, inventory is prevented from exceeding the physical / management capacity of the warehouse, thus avoiding waste of warehousing resources or the risk of warehouse overload. Service Level Constraints: By setting a minimum service level, ensure that the inventory plan at least meets the preset customer demand guarantee standards.

[0112] Beneficial effects Balancing costs and service: The dual objective function takes into account both "cost reduction" and "service maintenance", avoiding the extreme risks of traditional single-objective solutions (such as excessive inventory reduction leading to stockouts, or excessive inventory increasing costs), and achieving a win-win situation of economy and stability in inventory management.

[0113] Precise implementation of constraints: The three types of constraints limit the feasibility of the solution from three dimensions: supplier capacity, warehouse capacity, and service bottom line. This ensures that the optimized inventory solution is executable in the actual supply chain scenario and avoids the problem of "theoretically optimal but unimplementable".

[0114] Clear and controllable quantitative indicators: Inventory holding, replenishment, and stockout costs are broken down into independent variables, and service levels are quantified through cost ratios. This makes it easier for companies to accurately assess the impact weight of various costs, while also intuitively monitoring service level compliance and improving the precision of inventory management.

[0115] Optionally, the step of retrieving multiple candidate inventory management solutions based on the objective function and constraints to obtain a recommended inventory management solution includes: The candidate inventory management schemes are encoded as chromosomes, and a first number of chromosomes are generated. Through iterative selection, crossover, and mutation operations, a second number of candidate inventory management schemes with the optimal objective function are selected as initial solutions. Based on the initial solution, the probability of accepting inferior solutions is controlled by temperature decay, and the recommended inventory management scheme is finally obtained through convergence.

[0116] In this embodiment, the solution process includes: Initialize the population: Generate 50 chromosomes (stockpile strategy), safety stock. Range of values Replenishment cycle Value range [7, 30] days (regular replenishment cycle for power supplies); Iterative optimization (number of iterations = 100): Selection: Calculate the objective function value (minimum Cost + maximum Service) for each chromosome, and filter the top 30 high-quality chromosomes; Crossover: High-quality chromosomes are crossovered with a 50% probability (e.g., the first 5 material parameters of chromosome A are recombinated with the last 5 parameters of chromosome B). Mutation: The chromosome after crossover is mutated at a probability of 1% (e.g., the safety stock of cables is randomly adjusted from 100 sets to 105 sets). Output: After the iteration is complete, select the 10 inventory strategies with the best objective function as the initial solutions for the simulated annealing algorithm.

[0117] To achieve the above embodiments, this application also proposes a category-based optimal inventory management device based on advanced intelligent algorithms.

[0118] Figure 2 This is a schematic diagram of a category-based optimal inventory management device based on an advanced intelligent algorithm, provided as an embodiment of this application.

[0119] like Figure 2 As shown, the device may include: The preprocessing module 210 is used to filter and supplement the data related to the first supply chain to obtain the data related to the second supply chain. Feature extraction module 220 is used to extract features from the second supply chain related data to obtain candidate feature data; The prediction module 230 is used to make predictions based on the candidate feature data to obtain the target material demand value under extreme scenarios. The strategy determination module 240 is used to determine the objective function and constraints of inventory management based on the target material demand value, and to perform a scheme retrieval on multiple candidate inventory management schemes according to the objective function and constraints to obtain a recommended inventory management scheme.

[0120] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of this embodiment, and will not be repeated here.

[0121] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing method embodiments.

[0122] To implement the above embodiments, this application also proposes a computer program product having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in the foregoing method embodiments.

[0123] To implement the above embodiments, this application also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method described in the foregoing method embodiments.

[0124] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. For example, the electronic device 800 may be a mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, etc.

[0125] Reference Figure 3 The electronic device 800 may include one or more of the following components: processing component 802, memory 804, power component 806, multimedia component 808, audio component 810, input / output (I / O) interface 812, sensor component 814, and communication component 816.

[0126] Processing component 802 typically controls the overall operation of electronic device 800, such as operations associated with display, telephone calls, data communication, camera operation, and recording operations. Processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the methods described above. Furthermore, processing component 802 may include one or more modules to facilitate interaction between processing component 802 and other components. For example, processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.

[0127] Memory 804 is configured to store various types of data to support the operation of electronic device 800. Examples of such data include instructions for any application or method operating on electronic device 800, contact data, phonebook data, messages, pictures, videos, etc. Memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0128] Power component 806 provides power to various components of electronic device 800. Power component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 800.

[0129] Multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 808 includes a front-facing camera and / or a rear-facing camera. When the electronic device 800 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0130] Audio component 810 is configured to output and / or input audio signals. For example, audio component 810 includes a microphone (MIC) configured to receive external audio signals when electronic device 800 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 804 or transmitted via communication component 816. In some embodiments, audio component 810 also includes a speaker for outputting audio signals.

[0131] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0132] Sensor assembly 814 includes one or more sensors for providing state assessments of various aspects of electronic device 800. For example, sensor assembly 814 can detect the on / off state of electronic device 800, the relative positioning of components such as the display and keypad of electronic device 800, changes in position of electronic device 800 or a component of electronic device 800, the presence or absence of user contact with electronic device 800, orientation or acceleration / deceleration of electronic device 800, and temperature changes of electronic device 800. Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 814 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0133] Communication component 816 is configured to facilitate wired or wireless communication between electronic device 800 and other devices. Electronic device 800 can access wireless networks based on communication standards, such as WiFi, 4G, or 5G, or combinations thereof. In one exemplary embodiment, communication component 816 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 816 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0134] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0135] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, which can be executed by a processor 820 of an electronic device 800 to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0136] To implement the above embodiments, this application also proposes a chip, including: the chip includes a processing circuit configured to perform the methods provided in the foregoing embodiments.

[0137] Figure 4 This is a schematic diagram of the structure of a chip according to an embodiment of this application. See also... Figure 4 The diagram shown is a schematic representation of the structure of chip 1100, but it is not limited to this.

[0138] Chip 1100 includes processing circuitry 1101, which is configured to perform any of the above methods.

[0139] In some embodiments, chip 1100 further includes one or more interface circuits 1102. Optionally, the interface circuit 1102 is connected to memory 1103, and the interface circuit 1102 can be used to receive signals from memory 1103 or other devices, and the interface circuit 1102 can be used to send signals to memory 1103 or other devices. For example, the interface circuit 1102 can read instructions stored in memory 1103 and send the instructions to processing circuit 1101.

[0140] In some embodiments, the interface circuit 1102 performs at least one of the communication steps such as sending and / or receiving in the above method, while the processing circuit 1101 performs other steps.

[0141] In some embodiments, the terms interface circuit, interface, transceiver pin, transceiver, etc., can be used interchangeably.

[0142] In some embodiments, chip 1100 further includes one or more memories 1103 for storing instructions. Optionally, all or part of the memories 1103 may be located outside of chip 1100.

[0143] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0144] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0145] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0146] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0147] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0148] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes one or a combination of the steps of the method embodiments.

[0149] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0150] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A category-based optimal inventory management method based on advanced intelligent algorithms, characterized in that, include: Data related to the first supply chain is filtered and supplemented to obtain data related to the second supply chain; Feature extraction is performed on the data related to the second supply chain to obtain candidate feature data; Based on the candidate feature data, a prediction is made to obtain the target material demand value under extreme scenarios; Based on the target material demand value, the objective function and constraints of inventory management are determined. Based on the objective function and constraints, multiple candidate inventory management schemes are searched to obtain a recommended inventory management scheme.

2. The method according to claim 1, characterized in that, The process of filtering and supplementing the data related to the first supply chain to obtain the data related to the second supply chain includes: The data related to the first supply chain are uniformly labeled according to their respective regions; Identify the missing values ​​in the first supply chain-related data, supplement the missing values, and supplement the first supply chain-related data under extreme scenarios.

3. The method according to claim 2, characterized in that, Determining the missing values ​​in the first supply chain-related data includes: Remove invalid values ​​from the first supply chain-related data; For the real-time data in the first supply chain-related data after removal, if the proportion of valid data is lower than a preset first proportion threshold, then data is determined to be missing. If the proportion of valid data in the first supply chain-related data after removal is lower than a preset second proportion threshold, then data is determined to be missing.

4. The method according to claim 3, characterized in that, The process of supplementing the missing values ​​includes: In response to the data missing rate being lower than a preset first missing threshold, data completion is performed using a multi-factor correction based on the average of adjacent time periods and the device operating status; wherein, the data missing rate is determined based on the proportion of valid data; In response to the data missing rate being greater than or equal to the first missing threshold and lower than the preset second missing threshold, a standardized preprocessing and optimization algorithm is used to complete the data. In response to the data missing rate being greater than or equal to the second missing threshold, data completion is performed using multi-source data fusion and long short-term memory network model prediction.

5. The method according to claim 3, characterized in that, The supplementary data related to the first supply chain under extreme scenarios includes: Target events with similarity scores higher than a preset similarity threshold are selected from the meteorological database; Extract the category demand fluctuation coefficient corresponding to the target event; The supplementary data is determined based on the demand fluctuation coefficient of the aforementioned product category.

6. The method according to any one of claims 1-5, characterized in that, The prediction based on the candidate feature data to obtain the target material demand value under extreme scenarios includes: Prediction is performed based on the candidate feature data to obtain the initial material demand value; The initial material demand value is perturbed and optimized based on the simulated annealing algorithm. An initial temperature, temperature decay strategy and minimum temperature threshold are designed. A new correction coefficient is generated by the dynamic perturbation step size, and it is determined whether to accept the new correction coefficient. The globally optimal correction coefficient is saved during the iteration process to obtain the corrected target demand prediction value.

7. The method according to claim 6, characterized in that, The objective function includes: in, For inventory holding costs, To replenishment costs, Cost is the cost of stockouts, and Service is the service level. The constraints are as follows: Supply capacity constraints ,in, For the quantity of materials replenished in a single transaction, This represents the supplier's maximum monthly production capacity. Inventory capacity constraints ,in, For the current inventory of material i, This represents the maximum capacity of the warehouse. Service Level Constraints ,in, This is the lower limit of the service level.

8. The method according to claim 7, characterized in that, The step of retrieving multiple candidate inventory management solutions based on the objective function and constraints to obtain a recommended inventory management solution includes: The candidate inventory management schemes are encoded as chromosomes, and a first number of chromosomes are generated. Through iterative selection, crossover, and mutation operations, a second number of candidate inventory management schemes with the optimal objective function are selected as initial solutions. Based on the initial solution, the probability of accepting inferior solutions is controlled by temperature decay, and the recommended inventory management scheme is finally obtained through convergence.

9. A product category-based optimal inventory management device based on advanced intelligent algorithms, characterized in that, include: The preprocessing module is used to filter and supplement the data related to the first supply chain to obtain the data related to the second supply chain. The feature extraction module is used to extract features from the second supply chain-related data to obtain candidate feature data; The prediction module is used to make predictions based on the candidate feature data to obtain the target material demand value under extreme scenarios. The strategy determination module is used to determine the objective function and constraints of inventory management based on the target material demand value, and to perform a scheme retrieval on multiple candidate inventory management schemes according to the objective function and constraints to obtain a recommended inventory management scheme.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method as described in any one of claims 1-8.

Images