Electric power material warehouse supplementing method and system

Abstract

The invention discloses an electric power material library supplementing method and system, and the method comprises the steps: carrying out the characteristic analysis of electric power materials based on basic data, obtaining material characteristics, determining the index factors of the electric power materials based on the basic data, carrying out the clustering analysis of the electric power materials according to the index factors, and obtaining material types; determining an inventory management strategy of electric power materials according to the material category, constructing a project description word vector and a project material feature vector of a power grid project based on the basic data, performing clustering analysis on the power grid project according to the project description word vector and the project material feature vector to obtain a project type, and determining a demand prediction model corresponding to the project type according to the material feature; and on the basis of the material characteristics, the project description word vectors and the project material characteristic vectors, the demand prediction model is used for carrying out demand prediction on the electric power materials, and an intelligent library supplementing strategy of the electric power materials is determined on the basis of an inventory management strategy and a demand prediction value, so that the scientificity, responsiveness and intelligent level of material management are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of material management, and in particular to a method and system for replenishing electric power materials. Background Art

[0002] With the annual increase in investment in power grid construction, the material demand of power grid companies is growing rapidly. From the perspective of power material management, power materials come in a wide variety of types and specifications, with varying degrees of standardization. Furthermore, diverse project types and varying responsiveness requirements for material demand are required. Simply relying on project demand units to submit demand without accumulating the specific material demand characteristics of project types and structures can lead to inventory backlogs for some commonly used materials, panic ordering, and pre-purchase purchases of some materials. Demand forecasting based solely on inventory consumption data is inadequate to meet the diverse needs of various business types, such as infrastructure changes, operations and maintenance, and emergency reserves. Furthermore, power material consumption data often exhibits a high proportion of intermittent or sudden peaks in demand. Consequently, the simplistic application of classic inventory models (assuming stable demand) to static reserve quotas cannot avoid high inventory and stock-outs caused by high volatility. While world-class companies today have comprehensive supply chain planning systems, power grid companies have yet to clarify the relationship between demand planning and replenishment planning, and between replenishment planning and procurement and distribution planning.

[0003] Various types of power materials have characteristics such as obvious seasonality, high correlation with projects, special use for certain types of projects, large changes in technology upgrades or technology trends. When the warehouse material management department needs to know the above information, it finds that the inventory is insufficient and there is only a short time left to replenish the inventory. It also often happens that a large amount of inventory is prepared according to demand, but the demand unit no longer uses it or even the technology is eliminated, resulting in a large amount of stagnant inventory.

[0004] After years of informatization, a centralized power material management information system, centered around the ECP (Enterprise Collaboration Platform) and ERP (Enterprise Resource Planning) systems, has been gradually established, effectively supporting material management operations. However, inventory quota setting still relies on manual experience to determine whether to submit replenishment requests and the quantity to be replenished, with warehouse staff conducting comprehensive approvals. Currently, distribution network project plans cannot be accurately submitted, and changes in requirements and the combined submission of material requirements for multiple projects result in significant material surpluses and waste. Currently, there is no effective tool to support users in intelligently calculating inventory quotas and making replenishment recommendations. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for replenishing electric power materials, which can improve the scientificity, responsiveness and intelligence level of material management.

[0006] In order to solve the above technical problems, a technical solution adopted by the present invention is: A method for replenishing power materials, comprising the steps of: Obtain basic data related to power grid projects and power supplies; Performing a characteristic analysis on the electric power material based on the basic data to obtain material characteristics, and determining an index factor of the electric power material based on the basic data; Performing cluster analysis on the power materials according to the indicator factors to obtain material categories, and determining an inventory management strategy for the power materials according to the material categories; Constructing a project description word vector and a project material feature vector of the power grid project based on the basic data, and performing cluster analysis on the power grid project according to the project description word vector and the project material feature vector to obtain a project type; Determining a demand forecast model for the electric power material corresponding to the project type according to the material characteristics, and performing demand forecasting on the electric power material using the demand forecasting model based on the material characteristics, the project description word vector, and the project material feature vector to obtain a demand forecast value; An intelligent replenishment strategy for the electric power materials is determined based on the inventory management strategy and the demand forecast value.

[0007] In order to solve the above technical problems, another technical solution adopted by the present invention is: A power material replenishment system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented: Obtain basic data related to power grid projects and power supplies; Performing a characteristic analysis on the electric power material based on the basic data to obtain material characteristics, and determining an index factor of the electric power material based on the basic data; Performing cluster analysis on the power materials according to the indicator factors to obtain material categories, and determining an inventory management strategy for the power materials according to the material categories; Constructing a project description word vector and a project material feature vector of the power grid project based on the basic data, and performing cluster analysis on the power grid project according to the project description word vector and the project material feature vector to obtain a project type; Determining a demand forecast model for the electric power material corresponding to the project type according to the material characteristics, and performing demand forecasting on the electric power material using the demand forecasting model based on the material characteristics, the project description word vector, and the project material feature vector to obtain a demand forecast value; An intelligent replenishment strategy for the electric power materials is determined based on the inventory management strategy and the demand forecast value.

[0008] The beneficial effects of the present invention are as follows: based on basic data, the characteristics of electric power materials are analyzed to obtain material characteristics, and based on the basic data, the index factors of electric power materials are determined, the electric power materials are clustered according to the index factors to obtain material categories, and the inventory management strategy of electric power materials is determined according to the material categories, the project description word vector and project material feature vector of the power grid project are constructed based on the basic data, and the power grid project is clustered according to the project description word vector and the project material feature vector to obtain the project type, and the demand forecast model of electric power materials corresponding to the project type is determined according to the material characteristics, and the demand forecast model is used based on the material characteristics, the project description word vector and the project material feature vector. Demand for power materials is forecasted to obtain demand forecast values. Based on inventory management strategies and demand forecast values, intelligent replenishment strategies for power materials are determined. Through cluster analysis of power materials and combining the characteristics and management priorities of different types of materials, differentiated inventory management strategies are designed to effectively optimize inventory structure and material guarantee efficiency. Material portraits (material characteristics) and project portraits (project description word vectors and project material feature vectors) are integrated to dynamically predict future material demand. At the same time, intelligent replenishment strategies are determined in combination with inventory management strategies and demand forecast values, and a flexible material supply system that combines forecast-driven and real-time adjustment is constructed, thereby improving the scientificity, responsiveness and intelligence of material management. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a flowchart of the steps of a method for replenishing electric power materials according to an embodiment of the present invention; Figure 2 This is a structural diagram of a power material replenishment system according to an embodiment of the present invention; Figure 3 Schematic diagram of the distribution of forecast deviations of all materials in the method for replenishing electric power materials in an embodiment of the present invention. DETAILED DESCRIPTION

[0010] To illustrate the technical content, achieved objectives and effects of the present invention in detail, the following description is given in conjunction with the embodiments and accompanying drawings.

[0011] Please refer to Figure 1 A method for replenishing electric power materials includes the following steps: Obtain basic data related to power grid projects and power supplies; Performing a characteristic analysis on the electric power material based on the basic data to obtain material characteristics, and determining an index factor of the electric power material based on the basic data; Performing cluster analysis on the power materials according to the indicator factors to obtain material categories, and determining an inventory management strategy for the power materials according to the material categories; Constructing a project description word vector and a project material feature vector of the power grid project based on the basic data, and performing cluster analysis on the power grid project according to the project description word vector and the project material feature vector to obtain a project type; Determining a demand forecast model for the electric power material corresponding to the project type according to the material characteristics, and performing demand forecasting on the electric power material using the demand forecasting model based on the material characteristics, the project description word vector, and the project material feature vector to obtain a demand forecast value; An intelligent replenishment strategy for the electric power materials is determined based on the inventory management strategy and the demand forecast value.

[0012] From the above description, it can be seen that the beneficial effects of the present invention are: based on the basic data, the characteristics of the power materials are analyzed to obtain the material characteristics, and based on the basic data, the index factors of the power materials are determined, the power materials are clustered according to the index factors to obtain the material categories, and the inventory management strategy of the power materials is determined according to the material categories, the project description word vector and the project material feature vector of the power grid project are constructed based on the basic data, and the power grid project is clustered according to the project description word vector and the project material feature vector to obtain the project type, and the demand forecast model of the power materials corresponding to the project type is determined according to the material characteristics, and the demand is used based on the material characteristics, the project description word vector and the project material feature vector. The forecasting model forecasts the demand for power materials and obtains the demand forecast value. Based on the inventory management strategy and the demand forecast value, the intelligent replenishment strategy for power materials is determined. Through cluster analysis of power materials, combined with the characteristics and management priorities of different types of materials, differentiated inventory management strategies are designed to effectively optimize the inventory structure and material guarantee efficiency. The material portrait (material characteristics) and project portrait (project description word vector and project material feature vector) are integrated to dynamically predict the future demand for materials. At the same time, the intelligent replenishment strategy is determined by combining the inventory management strategy and the demand forecast value, and a flexible material supply system that combines forecast-driven and real-time adjustment is constructed, thereby improving the scientificity, responsiveness and intelligence of material management.

[0013] Furthermore, the characteristic analysis of the electric power material based on the basic data to obtain material characteristics includes: Obtaining demand data or supply data of the electric power material within a preset time period from the basic data; calculating a coefficient of variation of the electric power material based on the demand data or the supply data, and determining volatility of the electric power material based on the coefficient of variation; Obtaining the inventory value proportion, material value, demand and inventory turnover rate of the power materials from the basic data; Determining the importance of the power material based on the inventory value ratio, the material value, the demand and the inventory turnover rate; Obtaining the total investment amount of the power grid project and the amount of the power materials used from the basic data; Determine the project relevance of the power materials using regression analysis and Pearson correlation coefficient based on the total investment amount and the withdrawal amount; Material characteristics are obtained according to the volatility, the importance, and the project relevance.

[0014] From the above description, we can see that analyzing the volatility of power materials is mainly to assess whether there are significant fluctuations in the demand or supply of a certain material, and then to judge its stability. By conducting an importance analysis of power materials, we can quickly identify key materials in corporate operations, which helps to optimize procurement and inventory allocation. By conducting a project relevance analysis of power materials, project managers or material management personnel can make more informed decisions during the material procurement and inventory management process.

[0015] Furthermore, the calculating the coefficient of variation of the electric power material according to the demand data or the supply data includes: ; Where, CV represents the coefficient of variation, represents the standard deviation of the demand data or the supply data, represents the mean value of the demand data or the supply data.

[0016] From the above description, we can see that the coefficient of variation can effectively measure the degree of dispersion of data, and thus the coefficient of variation can be used to accurately determine the degree of volatility of power materials.

[0017] Furthermore, the indicator factors include purchase price, proportion of purchase amount, investment amount in the current year, average monthly number of withdrawals, average monthly withdrawal quantity, average monthly withdrawal amount, consumption fluctuation coefficient, number of suppliers and average supply cycle; The power materials are clustered according to the index factors to obtain material categories including: The power materials are clustered using K-means according to the purchase unit price, the proportion of the purchase amount, the investment amount in the current year, the average monthly number of withdrawals, the average monthly withdrawal quantity, the average monthly withdrawal amount, the consumption fluctuation coefficient, the number of suppliers and the average supply cycle to obtain material categories.

[0018] From the above description, we can see that the indicator factors consider three aspects: material capital cost, consumption characteristics, and supply guarantee. K-means clustering of power materials based on multidimensional impact indicators can quickly and scientifically classify inventory materials, laying a good foundation for differentiated demand forecasting and dynamic inventory control.

[0019] Furthermore, constructing the project description word vector and project material feature vector of the power grid project based on the basic data includes: Acquiring project description information of the power grid project from the basic data; Segmenting the project description information to obtain a keyword list; Perform one-hot encoding and position encoding on each keyword in the keyword list to obtain a project description word vector; Obtaining the material usage, actual material usage quantity and material design quantity of the power grid project from the basic data; Constructing an initial project material vector based on the material usage; Normalizing the initial project material vector to obtain a project material vector; Calculate the material deviation rate based on the actual quantity of the materials used and the designed quantity of the materials; A project material feature vector is obtained according to the project material vector and the material deviation rate.

[0020] As can be seen from the above description, one-hot encoding can clearly distinguish different keywords, and positional encoding can better understand the order of keywords in a sentence. Combining one-hot encoding and positional encoding to obtain project description word vectors not only reflects the core characteristics of the project but also enhances the model's understanding of word order information. Then, based on the material usage, a project material vector is constructed and the material deviation rate is calculated to obtain the project material feature vector, which is used to identify similarities between projects.

[0021] Furthermore, the project correlation includes strong correlation, medium correlation and weak correlation; Determining the demand forecast model for the power material corresponding to the project type according to the material characteristics includes: If the project correlation is the strong correlation, determining that the demand forecast model for the electric power materials corresponding to the project type is a linear regression model; If the project correlation is the medium correlation or the weak correlation, it is determined that the demand forecasting model for the electric power material corresponding to the project type is an extreme gradient boosting regression model.

[0022] From the above description, we can see that for strongly correlated power materials, the linear regression model is used as the demand forecasting model. The linear regression model has a clear structure and strong interpretability, and is suitable for these materials that are closely linked to the scale of funds. For medium and weakly correlated power materials, the extreme gradient boosting (XGBoost) regression model is used as the demand forecasting model. The XGBoost regression model has powerful nonlinear fitting capabilities and automatic feature selection mechanisms, which can more effectively capture the hidden influencing factors behind the materials, thereby improving the accuracy and robustness of the overall prediction.

[0023] Furthermore, the performing demand forecasting on the power material using the demand forecasting model based on the material characteristics, the project description word vector, and the project material feature vector to obtain a demand forecast value includes: Performing demand forecasting on the power material using the linear regression model according to the material characteristics to obtain a first demand forecast value; The extreme gradient boosting regression model is used to perform demand forecasting on the power material according to the project description word vector and the project material feature vector to obtain a second demand forecast value.

[0024] From the above description, it can be seen that the power materials used in the linear regression model belong to materials with strong project correlation. Demand forecasting can be achieved directly based on the material characteristics. When using the XGBoost regression model to forecast the demand for power materials, it is achieved based on the project description word vector and the project material feature vector. In this way, without the need for complete project implementation information, accurate material demand estimates can be quickly obtained based only on project text and material configuration information.

[0025] Furthermore, it also includes: Obtain the project planned start time and start time deviation label of the power grid project; Determine the theoretical project start time according to the project planned start time and the start time deviation label; Obtain the supplier's theoretical plan approval time, theoretical contract execution time, and theoretical delivery time; Determine the theoretical demand capture time based on the theoretical project start time, the theoretical plan approval time, the theoretical contract execution time, and the theoretical supply time; The intelligent replenishment strategy for the electric power materials determined based on the inventory management strategy and the demand forecast value includes: An intelligent inventory replenishment strategy for the electric power material is determined based on the theoretical demand capture time, the inventory management strategy, and the demand forecast value.

[0026] As can be seen from the above description, when determining the intelligent replenishment strategy, in addition to based on inventory management strategies and demand forecast values, the theoretical demand capture time obtained based on the project's theoretical start time, theoretical plan approval time, theoretical fulfillment execution time, and theoretical delivery time is also introduced. This effectively improves the scientific nature and foresight of time management and reduces material supply delays caused by unreasonable time arrangements.

[0027] Furthermore, the cluster analysis of the power grid project is performed based on the project description word vector and the project material feature vector to obtain project types including: K-means clustering is performed on the power grid project according to the project description word vector and the project material feature vector to obtain the project type.

[0028] From the above description, we can see that K-means clustering is performed on power grid projects based on project description word vectors and project material feature vectors. K-means clustering is an unsupervised iterative dynamic clustering algorithm. It is easy to implement, has fast processing speed for large data sets, high efficiency, easy convergence, multi-factor consideration, and high accuracy. It can accurately and quickly classify power grid projects.

[0029] Please refer to Figure 2 Another embodiment of the present invention provides an electric power material replenishment system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step in the above-mentioned electric power material replenishment method is implemented.

[0030] The above-mentioned method and system for replenishing electric power materials of the present invention can be applied to electric power material management scenarios, and are described below through specific implementation methods: Please refer to Figure 1 and Figure 3 , embodiment 1 of the present invention is: A method for replenishing power materials, comprising the steps of: S1. Obtain basic data related to power grid projects and power materials.

[0031] Specifically, original data related to power grid projects and electric power materials are obtained, and the original data are cleaned, extracted and normalized to obtain basic data related to power grid projects and electric power materials.

[0032] In an optional embodiment, the original data may include an outbound list, an inbound list, and a design list. The basic data may include material master information, project data, purchase batch data, historical inbound and outbound data, inventory information, etc.

[0033] S2. Analyzing the characteristics of the electric power material based on the basic data to obtain material characteristics, and determining the index factor of the electric power material based on the basic data, specifically including S21-S28: S21. Obtain demand data or supply data of the electric power material within a preset time period from the basic data.

[0034] S22. Calculate a coefficient of variation of the electric power material according to the demand data or the supply data, and determine volatility of the electric power material according to the coefficient of variation.

[0035] The calculating of the coefficient of variation of the electric power material according to the demand data or the supply data includes: ; Where, CV represents the coefficient of variation, represents the standard deviation of the demand data or the supply data, represents the mean value of the demand data or the supply data.

[0036] Determining the volatility of the electric power material according to the coefficient of variation includes: If the coefficient of variation is less than or equal to a first preset value, it is determined that the electric power material has low volatility; If the coefficient of variation is greater than a first preset value and less than or equal to a second preset value, it is determined that the electric power material has medium volatility; If the coefficient of variation is greater than a second preset value, it is determined that the electric power material has high volatility.

[0037] In an optional implementation, the first preset value is 1, and the second preset value is 2.

[0038] Low volatility indicates that demand or supply is relatively stable. Medium volatility indicates that the purchasing rhythm needs to be paid attention to. CV ≤1): A quantitative replenishment strategy (such as the EOQ economic order quantity model) can be adopted to maintain a low inventory. CV ≤2) of power supplies: A safety stock + regular replenishment strategy can be adopted, and procurement can be adjusted according to demand forecasts. High volatility ( CV >2) Power supplies: It is recommended to use flexible inventory strategies, such as VMI (vendor managed inventory) and JIT (just-in-time), to reduce the risk of overstocking and stockouts.

[0039] S23. Obtain the inventory value ratio, material value, demand, and inventory turnover rate of the power materials from the basic data.

[0040] S24. Determine the importance of the power material according to the inventory value ratio, the material value, the demand, and the inventory turnover rate.

[0041] Specifically, the initial importance of the electric power material is first determined according to the inventory value ratio, and then the importance of the electric power material is determined based on the initial importance according to the material value, the demand and the inventory turnover rate.

[0042] For example, if power supplies account for approximately 80% of the inventory value but are relatively small in quantity, the initial importance of these power supplies is determined to be Class A, representing the highest value and most critical component. These supplies are directly related to the company's core business, and a disruption in the supply chain could lead to production stagnation or significant losses. For example, large transformers, high-voltage switchgear, and critical cable materials have long procurement cycles and are expensive, requiring strict management. This is often achieved through precise demand forecasting, long-term procurement contracts, and vendor-managed inventory (VMI) to ensure a continuous supply.

[0043] If the inventory value of power supplies accounts for approximately 15% and demand is moderate, the initial importance of these power supplies is determined to be Class B. Class B materials are of intermediate importance and their management strategy falls between Class A and Class C. This class of supplies includes spare parts for conventional power equipment, such as small and medium-sized transformers, wires and cables, and circuit breakers. While these supplies are not as valuable as Class A materials, shortages can still impact construction progress or maintenance efficiency. Therefore, regular inventory checks and safety stock management are typically implemented to avoid backlogs while ensuring timely supply.

[0044] If power supplies account for approximately 5% of inventory value but are often the largest in quantity, the initial importance of these power supplies is determined to be Class C, which has the lowest value. These supplies typically include consumables, low-value hardware, fasteners, insulating tape, and labor protection supplies. They are low-priced but frequently used. The management strategy for these supplies primarily relies on bulk purchasing, regular replenishment, and automated inventory management to avoid excessive manpower investment and reduce storage costs.

[0045] On the basis of ABC classification, the importance level of materials can be further divided according to their value, demand and inventory turnover rate for more refined management.

[0046] If power supplies are generally high-value or in high demand, but the quantity is relatively small, accounting for approximately 20% of all supplies, then the importance of these power supplies is determined to be very important (Importance Level 1). These supplies are not only expensive, but also have long supply cycles and are difficult to procure. A shortage can severely impact project progress or the normal operation of the power system. Examples include core substation equipment, specialized high-voltage switches, and critical protective devices. For these supplies, companies typically adopt supply chain diversification, long-term procurement contracts, and real-time inventory monitoring to ensure a continuous supply of critical materials.

[0047] Power supplies are generally of moderate value or demand, with moderate quantities, accounting for approximately 20% to 50% of the supply, and are classified as moderately important (Importance Level 2). These supplies are essential to daily operations, but short-term shortages are unlikely to result in severe consequences. Examples include common cables, small and medium-sized power equipment, backup transformers, and common circuit breakers. Management strategies for these supplies typically employ periodic replenishment and safety stock arrangements to ensure a stable supply while mitigating the risk of inventory overstocking.

[0048] If power supplies are generally low-value but high in quantity, accounting for more than 50% of the inventory, they are considered unimportant (importance level 3). These supplies are often consumables or low-value spare parts, such as terminal blocks, cable ties, fasteners, lubricants, and gloves. The management of these supplies focuses on reducing inventory management costs and improving automated replenishment efficiency. Common strategies include bulk purchasing, just-in-time (JIT) delivery, and intelligent inventory management to reduce manual intervention and optimize operating costs.

[0049] S25. Obtain the total investment amount of the power grid project and the amount of the power materials used from the basic data.

[0050] S26. Determine the project relevance of the power materials using regression analysis and Pearson correlation coefficient based on the total investment amount and the withdrawal amount.

[0051] The project correlation includes strong correlation, medium correlation and weak correlation.

[0052] Specifically, the correlation between the independent variable (the total investment amount of the power grid project) and the dependent variable (the amount of power materials used) is constructed, with R 2 The project relevance of power materials is determined based on the coefficient (coefficient of determination) and combined with the Pearson correlation coefficient.

[0053] Based on linear regression, its basic idea is to find the best fitting line so that the sum of the squares of the distances from all data points to the fitting line is minimized. The linear regression model is expressed as: ; Where, Y Indicates the amount of power materials used. X represents the total investment amount of the power grid project, Represents the model intercept, that is, the expected value of the withdrawal amount when the total investment amount is 0, Represents the regression coefficient, that is, the change in the withdrawal amount when the total investment amount increases by one unit. represents the error term, which is the part that the model cannot explain.

[0054] The R² coefficient in regression analysis is an indicator used to measure the fitting effect of the linear regression model. R² indicates the degree to which the model explains the changes in the dependent variable. The closer the value is to 1, the more the model can explain the changes in the dependent variable and the better the model fitting effect. The calculation formula of the R² coefficient is: ; Where, represents the true observation value, represents the predicted value of the linear regression model, represents the mean of the dependent variable.

[0055] An R² coefficient of 0 indicates that the model explains no variation, while an R² of 1 indicates that the model perfectly explains all variation in the dependent variable. In a regression analysis of the total investment amount of power grid projects and the amount of power materials used, a higher R² indicates a greater impact of project investment on the amount of materials used.

[0056] To further understand the strength of the relationship between the total investment amount of power grid projects and the amount of power supplies used, we used the Pearson correlation coefficient. The Pearson correlation coefficient ranges from -1 to +1, with the following meanings: +1: Perfect positive correlation, indicating that as the independent variable increases, the dependent variable also increases. -1: Perfect negative correlation, indicating that as the independent variable increases, the dependent variable also decreases. 0: No correlation, indicating that there is no linear relationship between the independent and dependent variables.

[0057] Through regression analysis and Pearson correlation coefficient, the correlation between materials and projects can be labeled to help project managers or material management personnel make more informed decisions during material procurement and inventory management.

[0058] In an optional embodiment, if the Pearson correlation coefficient is close to +1 or -1, and the R² coefficient reaches a third preset value, indicating that there is a strong correlation between project investment and material requisition, the project correlation of power materials is determined to be a strong correlation. In this case, material demand can be directly predicted through project investment. If the Pearson correlation coefficient is between 0.5 and 0.8, and the R² coefficient is greater than the fourth preset value and less than the third preset value, it indicates that there is a certain relationship between project investment and material requisition, but it may still be affected by other factors. In this case, the project correlation of power materials is determined to be moderate. If the Pearson correlation coefficient is close to 0 and the R² value is less than or equal to the fourth preset value, it means that the relationship between project investment and material requisition is weak, and it may be necessary to introduce other variables or a more complex model for analysis. In this case, the project correlation of power materials is determined to be weak.

[0059] Generally speaking, important materials are strongly linearly correlated with the total amount of project investment, while unimportant materials have a weaker correlation.

[0060] S27. Obtaining material characteristics according to the volatility, the importance, and the project relevance.

[0061] S28. Determine the index factor of the electric power material based on the basic data.

[0062] Among them, the indicator factors include purchase unit price (unit: yuan), proportion of purchase amount (unit: %), investment amount in the year (unit: yuan), average monthly number of withdrawals (unit: times), average monthly withdrawal quantity (unit: unit of measurement), average monthly withdrawal amount (unit: yuan), consumption fluctuation coefficient (unit: %), number of suppliers (unit: piece) and average supply cycle (unit: day).

[0063] S3. Performing cluster analysis on the power materials according to the indicator factors to obtain material categories, and determining the inventory management strategy of the power materials according to the material categories, specifically including S31-S32: S31. Perform K-means clustering on the power materials based on the purchase unit price, the proportion of the purchase amount, the investment amount in the current year, the average monthly number of withdrawals, the average monthly withdrawal quantity, the average monthly withdrawal amount, the consumption fluctuation coefficient, the number of suppliers and the average supply cycle to obtain material categories.

[0064] Among them, the material categories include strategically important materials, business-type high-risk materials and value-based low-risk materials.

[0065] S32. Determine the inventory management strategy of the power materials according to the material category, as shown in Table 1.

[0066] Table 1 Inventory management strategies for power materials of different material categories

[0067] Table 1 reflects the attribute characteristics of different types of materials and the corresponding inventory management strategies. The purpose is to develop appropriate inventory management plans based on the characteristics of the materials to optimize inventory levels, reduce costs and ensure the stability of the material supply chain.

[0068] Strategically important materials: These materials are high-value, require high consumption, and have strong, ongoing demand, posing a higher supply risk. They are often involved in core production or business processes, and a supply disruption could have serious consequences. Strategically important materials require extreme caution in management, with strict inventory control while ensuring that stock-outs are never an issue. Due to significant demand fluctuations, forecast accuracy is crucial, so a fixed order quantity model is employed, and regular inventory checks are required to ensure timely replenishment.

[0069] Fixed Order Quantity Control Model with Continuous Inspection Specifically: E=e× ; s=p×N×1 / M+E; S=q×N×1 / M+s; Q=e× ; Where E represents the expected out-of-stock quantity, e represents the unit out-of-stock cost, represents the service level, which is usually the complement of the out-of-stock probability, s represents the safety stock, that is, the additional inventory maintained before the next order arrives to cope with demand fluctuations, p represents the unit holding cost, N represents the average demand within the order cycle, M represents the length of the order cycle, S represents the reorder point, that is, the inventory level that triggers a new order, q represents the ordering cost, that is, the fixed cost incurred each time an order is placed, and Q represents the order quantity.

[0070] By continuously checking the company's existing inventory, a safety stock is calculated based on a standard table of demand forecasts and service levels. Based on the safety stock, the upper and lower limits of inventory are dynamically calculated. If the current inventory level falls below the lower limit during the check, a replenishment signal is triggered, and the order quantity is kept constant at Q each time. This control model is applicable to the first category of strategically important materials. The inventory management of the first category of materials in power companies has basic characteristics such as high value, high liquidity, and high supply risk. These materials have high requirements, requiring both guaranteed stock-outs and reasonable inventory control.

[0071] Business-related high-risk materials: These materials are typically low-value, consumed quickly, and have cyclical demand. They also face higher supply risks and may be affected by market or supply chain fluctuations. Inventory management for these materials also requires significant attention, particularly ensuring sufficient inventory to avoid stockouts. Accurate demand forecasts are also crucial, and to ensure a continuous supply of inventory, a fixed order point control model is often employed. Unlike strategically important materials, supply risk management for these materials focuses more on preventing supply disruptions.

[0072] Fixed order point control model for continuous inspection Specifically: ; ； ; This control model primarily utilizes a continuous check method, calculating a maximum inventory limit based on safety stock and using this maximum inventory limit as a fixed order point. When the check indicates that the current inventory level falls below the fixed order point S, a replenishment signal is triggered. Each order quantity Q fluctuates based on actual inventory levels. This control model is primarily applicable to Category II high-risk business materials, which have moderate value, high liquidity, and high supply risk.

[0073] Valuable, low-risk materials: These materials are high in value but low in consumption, posing a lower supply risk. Their demand is sporadic and infrequent, resulting in a lesser impact on the supply chain. Because these materials have low consumption and a stable supply, they require less attention, and inventory levels can be maintained at a moderate level. Because demand forecasts require less precision, management adopts a periodic inspection and control model, regularly checking inventory levels and deciding whether to restock based on actual needs.

[0074] Periodic Inspection Control Model Specifically: ; ; ; This control model primarily utilizes periodic inspections, typically monthly, based on a maximum safety stock. When inspections are completed, restocking begins, but the maximum safety stock cannot be exceeded. The quantity ordered each time depends on the material demand forecast. This control model is primarily applicable to Category III, low-risk, value-added materials, characterized by high value and low supply risk. The inventory control strategy does not require a fixed order point; it only requires an inspection period and a maximum safety stock.

[0075] S4. Constructing a project description word vector and a project material feature vector of the power grid project based on the basic data, and performing cluster analysis on the power grid project based on the project description word vector and the project material feature vector to obtain a project type, specifically including S41-S49: S41. Obtain project description information of the power grid project from the basic data.

[0076] S42: Segment the project description information to obtain a keyword list.

[0077] S43. Perform one-hot encoding and position encoding on each keyword in the keyword list to obtain a project description word vector.

[0078] In natural language processing (NLP) tasks, word embedding is a core technology for representing textual information. One-hot encoding is a simple yet effective word embedding method. It is a discrete feature representation method used to convert categorical variables into binary vectors. For a keyword set for a power grid project, each keyword can be mapped to a unique index and converted into a corresponding one-hot vector.

[0079] For example, suppose the project description of a power grid project contains the following keywords: substation, transmission line, distribution network, photovoltaic access, and smart grid. If the size of each code is 5, the one-hot codes for substation, transmission line, and photovoltaic access can be: substation = [1,0,0,0,0]; transmissionlines = [0,1,0,0,0]; PV_Access = [0,0,0,1,0]; One-Hot encoding can clearly distinguish different keywords, but it has problems of high dimensionality and sparsity. In addition, One-Hot encoding cannot reflect the order information of words in the text. In the project description information of the power grid project, the relative position of different keywords may affect their meaning. For example, "substation expansion" and "substation expansion" have similar meanings in reading, but One-Hot encoding cannot distinguish between the two. "Distribution network upgrade" and "transmission line transformation" may involve different power grid levels, and position information can help distinguish these projects. Therefore, it is necessary to introduce position encoding so that the model can better understand the order information of keywords in the sentence. The core idea of ​​position encoding is to generate learnable position information based on fixed mathematical functions (such as sine and cosine functions). Assuming that the length of the project description sequence is n, the position encoding of each position 𝑖 can be expressed as: PE (i,2j) =sin( i / 10000 2j / d); PE (i,2j+1) =cos( i / 10000 2j / d ); Where, i Indicates the position of the word in the project description (numbering starts from 0). j represents the general dimension of the word vector (because sine and cosine calculate odd and even positions respectively), d Represents the dimension of the word vector.

[0080] The introduction of positional encoding enables the One-Hot word vector to not only represent the information of the vocabulary itself, but also incorporate word order features, thereby improving the model's ability to understand project descriptions.

[0081] S44. Obtaining the material usage, actual material usage quantity, and material design quantity of the power grid project from the basic data.

[0082] S45. Construct an initial project material vector according to the material usage.

[0083] Each project will use different types of materials, such as cables, transformers, switchgear, towers, etc. N There are 10 kinds of materials (for example), and the initial project material vector constructed according to the material usage of a project can be expressed as: V 项目 =( x 1, x 2, x 3,…, x N ); Where, x N Indicates the project N The amount of materials used.

[0084] S46. Normalize the initial project material vector to obtain a project material vector.

[0085] Since the units and magnitudes of different materials may vary greatly (for example, cables are counted in meters, while transformers are counted in units), normalization is required to ensure that the dimensions of the material vector are of the same magnitude.

[0086] The normalization process is Min-Max normalization, specifically: ; Where, Indicates the first i The value of the data point, Represents the normalized i The value of the data point, represents the minimum value of all data points in the dataset, Represents the maximum value of all data points in the dataset.

[0087] In this way, all material values ​​are mapped to [0,1], making the influence of different materials more balanced.

[0088] S47. Calculate the material deviation rate based on the actual quantity of the materials used and the designed quantity of the materials.

[0089] During the construction and operation and maintenance of power grid projects, material management directly impacts construction progress, cost control, and resource utilization efficiency. Analyzing the deviation between actual material usage and designed quantities not only optimizes the material supply chain but also enhances project management and enables refined management. The material deviation rate identifies discrepancies between actual and designed quantities.

[0090] The material deviation rate is specifically: .

[0091] S48. Obtain a project material feature vector according to the project material vector and the material deviation rate.

[0092] By constructing a project's material vectors and descriptive word vectors, we can not only comprehensively characterize the project's material usage model, but also capture the semantic characteristics contained in the project's textual information, thereby enabling automated project classification. The material vectors reflect the frequency of use and investment structure of various materials in the project, while the descriptive word vectors mine semantic differences based on textual information such as project names and content. The fusion of the two can effectively distinguish different types of projects.

[0093] S49. Perform K-means clustering on the power grid project according to the project description word vector and the project material feature vector to obtain a project type.

[0094] The project types include conventional projects and projects with insufficient deviation control during the execution phase. The project clustering results are shown in Table 2.

[0095] Table 2 Item clustering results

[0096] Table 2 summarizes the typical characteristics of various types of projects during their execution by describing their characteristics. Overall, most projects have relatively few types of important materials, indicating that the project structure itself is relatively simple and that key resources are concentrated. However, at the execution level, there are significant differences in deviation control and time management among different projects. Some projects mentioned "good deviation control" in their descriptions. In these projects, the discrepancy between material collection and design is small, and the plan execution is relatively smooth; while other projects are characterized by "large deviations between material collection and design," reflecting the presence of many unplanned changes during the execution process and problems with management and coordination. In addition, in terms of start time, projects generally have the phenomenon of "starting early" or "starting late," indicating that there is generally a certain degree of time deviation between actual execution and the original plan.

[0097] S5. Determining a demand forecast model for the electric power material corresponding to the project type according to the material characteristics, and using the demand forecast model to forecast demand for the electric power material based on the material characteristics, the project description word vector, and the project material feature vector to obtain a demand forecast value, specifically including S51-S54: S51: If the material characteristic is the strong correlation, determine that the demand forecast model for the electric power material corresponding to the project type is a linear regression model.

[0098] S52: If the material characteristic is the medium correlation or the weak correlation, determine that the demand forecasting model for the electric power material corresponding to the project type is an extreme gradient boosting regression model.

[0099] S53: Use the linear regression model to perform demand forecasting on the power material according to the material characteristics to obtain a first demand forecast value.

[0100] Specifically, the linear regression model is used to perform demand forecasting on the electric power materials according to the project correlation to obtain a first demand forecast value.

[0101] For example, through material property analysis, it can be found that 10% of materials have a strong linear correlation with the total project investment amount. These materials can be used to construct a correlation formula between the total investment amount and the material requisition amount. When a new project emerges, the total project investment amount can be directly used and combined with the correlation formula to obtain the material requisition quantity.

[0102] S54. Perform demand forecasting on the electric power material using the extreme gradient boosting regression model according to the project description word vector and the project material feature vector to obtain a second demand forecast value.

[0103] The XGBoost regression model is an ensemble learning algorithm based on the gradient boosting framework. Its core concept is to iteratively train weak learners (typically regression trees), fitting each round of predictions to the residuals of the previous round to gradually optimize the model's predictive performance. Compared with traditional regression methods, the XGBoost regression model offers significant advantages in accuracy, training speed, and resistance to overfitting. It is particularly suitable for scenarios with high feature dimensions and complex data distribution.

[0104] The first step in model training is preparing input features. The input consists of two parts: a project material feature vector and a project description word vector. The project material feature vector reflects the historical proportion of each type of material invested in the project and provides a quantitative description of the project's material structure. The project description word vector encodes the project text fields to extract implicit semantic information. These two vectors are then normalized or standardized, concatenated, and combined as the complete project input features, which are then fed into the XGBoost regression model.

[0105] During the model training phase, the system uses historical projects as training samples, and the target variable is the usage amount or quantity of specific materials. The XGBoost regression model can identify which feature dimensions in the project material feature vector and the project description word vector contribute most to the prediction target by automatically constructing feature splitting nodes. At the same time, through the built-in regularization mechanism and pruning strategy, the overfitting risk of the model is effectively suppressed, and the prediction stability on new projects is improved. After the model training is completed, the project material feature vector and project description word vector of the project to be predicted are used as input to the trained XGBoost regression model, and the predicted values ​​of various target materials can be output. This process makes it possible to quickly obtain accurate material demand estimates based only on project text and preliminary material configuration information without the need for complete project implementation information, providing strong technical support for pre-procurement decisions.

[0106] like Figure 3 As shown, Figure 3 The deviation rate distribution of the usage of different materials in various projects predicted by the demand forecasting model shows that the overall deviation rate of the model is concentrated below 0.2, and the overall deviation is less than 15%. In general, the prediction effect of the model can meet practical use.

[0107] In an optional embodiment, the method further includes: Obtain the project planned start time and start time deviation label of the power grid project; Determine the theoretical project start time according to the project planned start time and the start time deviation label; Obtain the supplier's theoretical plan approval time, theoretical contract execution time, and theoretical delivery time; The theoretical demand capture time is determined based on the theoretical project start time, the theoretical plan approval time, the theoretical contract execution time and the theoretical supply time.

[0108] In an optional embodiment, the method further includes: A supply lead time model is constructed, and the material submission time is calculated based on the supply lead time model.

[0109] The supply lead time model includes analysis of the time interval between order generation and fulfillment and analysis of the time interval between fulfillment arrival and start of warehousing.

[0110] In power grid supply chain management, lead orders are purchase orders that are planned and reserved in advance to ensure the timely supply of critical materials before demand. Order generation time refers to the time when the purchase order is created in the system, while fulfillment time typically refers to the time when the supplier actually completes delivery. A smaller time interval between these two indicates a faster supply response; a larger time interval indicates potential issues with untimely planning, poor supply, or delayed delivery.

[0111] Different suppliers have different delivery times. Some suppliers' delivery times fluctuate slightly, while others fluctuate slightly more. This can be judged through volatility analysis. The stability of the supplier's supply wave can be judged through the volatility of the coefficient of variation. For example, 0<coefficient of variation≤0.6 indicates small volatility, which means that the data is relatively stable and the supply process is relatively controllable; 0.6<coefficient of variation≤1 indicates small volatility, which means that the data has certain fluctuations but is still within an acceptable range and is slightly unstable; 1<coefficient of variation≤2 indicates large volatility and obvious data fluctuations, which may involve periodic changes or external interference factors; coefficient of variation>2 indicates large volatility and extremely unstable data, which requires special attention and may be abnormal fluctuations or process problems.

[0112] Arrival of contracted goods refers to the time when the supplier actually delivers the materials to their destination, while the start of warehousing is when the purchaser or user registers the arrival of the goods in the system, inspects them for storage, or confirms receipt. The time interval between these two is a key indicator for assessing the efficiency of internal receiving and acceptance processes and whether there are any delays or backlogs.

[0113] This system integrates the time dimension with the schedule deviation factor, focusing on the deviation between the project's planned start time and its historical start time. By establishing a start time deviation labeling system, the system can identify the time fluctuation range commonly found in the actual execution of different project types. Based on this deviation information and combined with the original planned nodes, the system calculates the theoretical start time for each project. Furthermore, the system incorporates the material supply lead time into its calculations, deducing when the materials should be submitted to ensure smooth delivery to the site before the project's critical construction milestones. This process effectively improves the scientific and forward-looking nature of time management and reduces material supply delays caused by irrational scheduling.

[0114] S6. Determine an intelligent replenishment strategy for the electric power materials based on the inventory management strategy and the demand forecast value.

[0115] In an optional implementation, the intelligent replenishment strategy for the electric power materials is determined based on the theoretical demand capture time, the inventory management strategy, the demand forecast value, and the material submission time.

[0116] This invention breaks away from the traditional manual material replenishment model. It leverages machine learning, data mining, and strategy optimization to create a closed-loop system that integrates project clustering, demand forecasting, construction start time estimation, and replenishment strategy development. This not only improves the accuracy and timeliness of material supply but also establishes a new supply chain management model for enterprises that is highly automated, responsive, and capable of sustainable optimization.

[0117] Please refer to Figure 2 , the second embodiment of the present invention is: A power material replenishment system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step of the power material replenishment method in the first embodiment is implemented.

[0118] In summary, the present invention provides a method and system for replenishing electric power materials, which performs characteristic analysis on electric power materials based on basic data to obtain material characteristics, and determines index factors of electric power materials based on the basic data, performs cluster analysis on electric power materials according to the index factors to obtain material categories, and determines inventory management strategies for electric power materials according to the material categories, constructs project description word vectors and project material feature vectors of power grid projects based on basic data, and performs cluster analysis on power grid projects according to the project description word vectors and project material feature vectors to obtain project types, determines demand forecasting models for electric power materials corresponding to project types according to material characteristics, and uses demand forecasting models to forecast demand for electric power materials based on material characteristics, project description word vectors, and project material feature vectors to obtain demand forecast values, and determines intelligent replenishment strategies for electric power materials based on inventory management strategies and demand forecast values, thereby Through cluster analysis of power materials, combined with the characteristics and management priorities of different types of materials, differentiated inventory management strategies are designed to effectively optimize inventory structure and material guarantee efficiency, and the material portrait and project portrait are integrated to dynamically predict future material demand. At the same time, the intelligent replenishment strategy is determined by combining inventory management strategies and demand forecast values, and a flexible material supply system that combines forecast-driven and real-time adjustment is constructed, thereby improving the scientificity, responsiveness and intelligence of material management. In addition, power materials using the linear regression model belong to materials with strong project correlation, and demand prediction can be achieved directly based on material characteristics. When using the XGBoost regression model to predict the demand for power materials, it is achieved based on the project description word vector and the project material feature vector. In this way, without the need for complete project implementation information, accurate material demand estimates can be quickly obtained based on project text and material configuration information.

[0119] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's description and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for replenishing electric power materials, characterized in that: Including steps: Obtain basic data related to power grid projects and power supplies; Performing a characteristic analysis on the electric power material based on the basic data to obtain material characteristics, and determining an index factor of the electric power material based on the basic data; Performing cluster analysis on the power materials according to the indicator factors to obtain material categories, and determining an inventory management strategy for the power materials according to the material categories; Constructing a project description word vector and a project material feature vector of the power grid project based on the basic data, and performing cluster analysis on the power grid project according to the project description word vector and the project material feature vector to obtain a project type; Determining a demand forecast model for the electric power material corresponding to the project type according to the material characteristics, and performing demand forecasting on the electric power material using the demand forecasting model based on the material characteristics, the project description word vector, and the project material feature vector to obtain a demand forecast value; An intelligent replenishment strategy for the electric power materials is determined based on the inventory management strategy and the demand forecast value.

2. A method for replenishing electric power materials according to claim 1, characterized in that: The characteristic analysis of the power material based on the basic data to obtain the material characteristics includes: Obtaining demand data or supply data of the electric power material within a preset time period from the basic data; calculating a coefficient of variation of the electric power material based on the demand data or the supply data, and determining volatility of the electric power material based on the coefficient of variation; Obtaining the inventory value proportion, material value, demand and inventory turnover rate of the power materials from the basic data; Determining the importance of the power material based on the inventory value ratio, the material value, the demand and the inventory turnover rate; Obtaining the total investment amount of the power grid project and the amount of the power materials used from the basic data; Determine the project relevance of the power materials using regression analysis and Pearson correlation coefficient based on the total investment amount and the withdrawal amount; Material characteristics are obtained according to the volatility, the importance, and the project relevance.

3. A method for replenishing electric power materials according to claim 2, characterized in that: Calculating the coefficient of variation of the electric power material according to the demand data or the supply data includes: ； Where, CV represents the coefficient of variation, represents the standard deviation of the demand data or the supply data, represents the mean value of the demand data or the supply data.

4. The method for replenishing electric power materials according to claim 1, characterized in that: The index factors include purchase price, proportion of purchase amount, investment amount in the current year, average monthly number of withdrawals, average monthly withdrawal quantity, average monthly withdrawal amount, consumption fluctuation coefficient, number of suppliers and average supply cycle; The power materials are clustered according to the index factors to obtain material categories including: The power materials are clustered using K-means according to the purchase unit price, the proportion of the purchase amount, the investment amount in the current year, the average monthly number of withdrawals, the average monthly withdrawal quantity, the average monthly withdrawal amount, the consumption fluctuation coefficient, the number of suppliers and the average supply cycle to obtain material categories.

5. The method for replenishing electric power materials according to claim 1, characterized in that: The constructing of the project description word vector and the project material feature vector of the power grid project based on the basic data includes: Acquiring project description information of the power grid project from the basic data; Segmenting the project description information to obtain a keyword list; Perform one-hot encoding and position encoding on each keyword in the keyword list to obtain a project description word vector; Obtaining the material usage, actual material usage quantity and material design quantity of the power grid project from the basic data; Constructing an initial project material vector based on the material usage; Normalizing the initial project material vector to obtain a project material vector; Calculate the material deviation rate based on the actual quantity of the materials used and the designed quantity of the materials; A project material feature vector is obtained according to the project material vector and the material deviation rate.

6. A method for replenishing electric power materials according to claim 2, characterized in that: The project correlation includes strong correlation, medium correlation and weak correlation; Determining the demand forecast model for the power material corresponding to the project type according to the material characteristics includes: If the project correlation is the strong correlation, determining that the demand forecast model for the electric power materials corresponding to the project type is a linear regression model; If the project correlation is the medium correlation or the weak correlation, it is determined that the demand forecasting model for the electric power material corresponding to the project type is an extreme gradient boosting regression model.

7. A method for replenishing electric power materials according to claim 6, characterized in that: The performing demand forecasting on the power material using the demand forecasting model based on the material characteristics, the project description word vector, and the project material feature vector to obtain a demand forecast value includes: Performing demand forecasting on the power material using the linear regression model according to the material characteristics to obtain a first demand forecast value; The extreme gradient boosting regression model is used to perform demand forecasting on the power material according to the project description word vector and the project material feature vector to obtain a second demand forecast value.

8. The method for replenishing electric power materials according to claim 1, characterized in that: Also includes: Obtain the project planned start time and start time deviation label of the power grid project; Determine the theoretical project start time according to the project planned start time and the start time deviation label; Obtain the supplier's theoretical plan approval time, theoretical contract execution time, and theoretical delivery time; Determine the theoretical demand capture time based on the theoretical project start time, the theoretical plan approval time, the theoretical contract execution time, and the theoretical supply time; The intelligent replenishment strategy for the electric power materials determined based on the inventory management strategy and the demand forecast value includes: An intelligent inventory replenishment strategy for the electric power material is determined based on the theoretical demand capture time, the inventory management strategy, and the demand forecast value.

9. The method for replenishing electric power materials according to claim 1, characterized in that: The cluster analysis of the power grid project based on the project description word vector and the project material feature vector is performed to obtain the project types including: K-means clustering is performed on the power grid project according to the project description word vector and the project material feature vector to obtain the project type.

10. A power material replenishment system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, each step of the method for replenishing electric power materials according to any one of claims 1 to 9 is implemented.

Images