Distribution network material multi-level cost decomposition and accounting method

Abstract

The invention discloses a distribution network material multi-level cost decomposition and accounting method, and the method comprises the following specific steps: constructing a multi-dimensional special BOM model, and building a structured database; the distribution network material cost is decomposed step by step, and cost data are integrated by using an ERP system to realize refined cost collection; carrying out regression analysis and principal component analysis on the collected cost data, identifying key factors influencing the cost, and predicting a cost change trend; constructing a multiple linear regression model, calculating the cost ratio of each element, and visually displaying the cost ratio by using a data visualization technology; a distribution network material cost database is established, cost data is updated in real time, a data mining technology is used for cost trend analysis, dynamic monitoring and management of cost are realized, and a cost structure is continuously optimized. According to the invention, refined, automatic and dynamic management of the distribution network material cost can be realized, the efficiency and precision of cost accounting are improved, and an effective tool is provided for cost control and management decision of distribution network materials.

Description

Technical Field

[0001] The present invention relates to the field of power system distribution network material management, and in particular to a multi-level cost decomposition and accounting method for distribution network materials. Background Art

[0002] With the development of the power industry, the cost management of distribution network materials, a fundamental component of grid construction and maintenance, has become increasingly important. Traditional cost accounting methods, often based on broad, general estimates, are unable to meet the demands of modern, refined management. Accurately and efficiently breaking down and calculating the costs of distribution network materials has become a major challenge within the industry, particularly in the face of a complex and volatile market environment and growing cost pressures.

[0003] In the past, cost analysis of distribution network materials primarily relied on bills of materials (BOMs) and basic cost estimation methods. These methods often lacked in-depth consideration of multi-dimensional factors such as process flow and resource consumption, resulting in inaccurate cost accounting and an inability to fully reflect the cost structure in actual production. Furthermore, due to the lack of effective data management and version control mechanisms, data consistency issues were prominent across different time points or projects, affecting the reliability and traceability of cost analysis results.

[0004] With the development of information technology, the application of ERP systems in enterprise resource planning has become increasingly widespread, making it possible to integrate cost data. However, simply applying ERP systems to cost accounting cannot completely solve the problem of insufficient cost granularity. Especially when dealing with complex product structures and diverse cost factors, how to effectively break down costs to the component, part, and even part level to achieve truly refined cost aggregation remains a pressing technical challenge.

[0005] At the same time, in order to further improve the effectiveness of cost management, not only accurate cost accounting is required, but also the ability to identify key factors affecting costs and predict cost change trends. This requires extracting valuable information from massive cost data, but current statistical analysis and data processing methods are difficult to achieve. Summary of the Invention

[0006] The purpose of the present invention is to provide a multi-level cost decomposition and accounting method for distribution network materials, which can not only realize the refined, automated and dynamic management of distribution network material costs, but also improve the efficiency and accuracy of cost accounting.

[0007] The technical solution of the present invention to solve the above technical problems is as follows:

[0008] The present invention provides a multi-level cost decomposition and accounting method for distribution network materials, comprising the following steps:

[0009] S1: Build a multi-dimensional dedicated BOM model and establish a structured database;

[0010] S2: Based on a dedicated BOM model, the distribution network material costs are broken down into components, parts, and spare parts step by step. Expense data is integrated into the ERP system to achieve refined cost aggregation.

[0011] S3: Using statistical analysis and data mining methods, we conduct regression analysis and principal component analysis on the collected cost data to identify key factors affecting costs and predict cost change trends;

[0012] S4: Based on the identified key cost factors, a multivariate linear regression model is constructed to accurately calculate the cost proportion of each factor and intuitively display it using data visualization technology;

[0013] S5: Establish a distribution network material cost database, update cost data in real time, use data mining technology to analyze cost trends, realize dynamic monitoring and management of costs, and continuously optimize the cost structure.

[0014] Optionally, in step S1, a multi-level dedicated BOM model including a bill of materials (BOM), process path, and resource consumption is constructed based on the physical properties, process flow, and resource constraint characteristics of the distribution network materials; database transaction management and version control technology are used to establish a structured database, which includes basic materials, process parameters, and resource unit prices, and supports multi-version data traceability and consistency maintenance.

[0015] Optionally, in step S1, the dedicated BOM model realizes cross-level data association between the three core dimensions of material layer, process layer and resource layer through a unique identifier (UUID). The material layer is associated with the process number of the process layer through the material code, and the process layer binds the resource layer data through the resource type to form an end-to-end structured data link; the dedicated BOM model can add customized attributes through extended fields.

[0016] Optionally, in step S1, a semantic version numbering rule is designed, where the version number adopts the format of Major.Minor.Patch, where Major indicates a major structural change, Minor corresponds to a process or resource field expansion, and Patch is used for data correction; a calculation formula for the version change impact coefficient Δ is defined, which guides priority decisions by quantifying the sensitivity of field changes to cost accounting;

[0017]

[0018] Among them, N 新增 N is the number of new fields; 删除 is the number of deleted fields; α and β are weight coefficients;

[0019] The data traceability is achieved through database triggers; through transaction management rules, when the material layer data changes, the process and resource data are automatically updated synchronously, and row-level locks are used to prevent concurrency conflicts.

[0020] Optionally, the step S2 is specifically:

[0021] Based on the three-layer structure of material layer, process layer and resource layer of the dedicated BOM model, the three-level cost decomposition logic of components, parts and components is constructed.

[0022] Define three-level cost decomposition fields and clarify the mapping relationship between cost components at each level and ERP system data sources;

[0023] Synchronize data with the ERP system in real time through the middleware interface to establish material coding mapping rules, process mapping rules, and resource consumption mapping rules;

[0024] Build a two-way verification mechanism based on the real-time data flow of the ERP system.

[0025] Optionally, in step S2,

[0026] The logic for constructing the three-level cost decomposition is:

[0027] Component-level cost: Taking a functional module as a unit, summarize the direct material costs of all the components it contains;

[0028] Component-level cost: Disassemble the assembly into specific parts and add up the processing costs and labor hours.

[0029] Part-level cost: Break it down to individual parts and link it to the purchase price, transportation costs, and inventory management costs in the ERP system;

[0030] The material code mapping rules, process mapping rules and resource consumption mapping rules are as follows:

[0031] (1) Material code mapping: The material code of the BOM model is associated with the material master data of the ERP system to obtain real-time purchase prices and inventory status, supporting part-level material cost and inventory management cost accounting;

[0032] (2) Process mapping: The process number in the BOM model is associated with the standard process path and process database in the ERP system to obtain the standard process flow, single-piece labor time and equipment matching parameters, supporting the calculation of component-level processing costs and equipment resource consumption;

[0033] (3) Resource consumption mapping: The resource layer data is connected to the ERP human resources and equipment resource modules to obtain labor hour rates, overtime rates and equipment energy consumption unit prices, and to calculate the resource consumption costs of each process in detail;

[0034] Component-level processing fee calculation:

[0035]

[0036] Where n is the number of processes involved in the component; labor rate i is the rate of the i-th process in the ERP system work order;

[0037] The bidirectional verification mechanism is constructed as follows:

[0038] (1) Forward aggregation: cost accounting tasks are distributed from the BOM model to the ERP module step by step;

[0039] (2) Reverse verification: When the ERP system detects that the material price increase exceeds the threshold, it triggers the BOM model version upgrade and recalculates the component-level cost impact.

[0040] Optionally, in step S3, principal component analysis (PCA) is used to perform dimensionality reduction processing on the collected cost data, extract core cost factors, and construct a multivariate linear regression model to quantify the contribution of each factor;

[0041]

[0042] Among them, Y is the total cost, X i is the input matrix, λ is the regularization coefficient, β i Reflects the impact of each factor on cost.

[0043] Optionally, in step S3, a dynamic prediction model is constructed based on time series analysis and ensemble learning:

[0044] Divide historical data into training and test sets in chronological order, and align external variables at quarterly frequencies;

[0045] The Prophet algorithm is used to capture the periodicity and holiday effects of cost data, and the LSTM neural network is combined to capture nonlinear trends. The model loss function is defined as:

[0046] Loss=α·MAE+(1-α)·RMSE+β·R 2 (5)

[0047] α = 0.6 is used to balance prediction error and model interpretability; β = 0.2 constrains model complexity.

[0048] Optionally, in step S4, a multiple linear regression model is established with the principal component factors and key original variables as independent variables and the total cost as the dependent variable.

[0049]

[0050] Among them, β i Solved by the least square method, it reflects the marginal contribution of each factor to the total cost, ∈ is the error term;

[0051] Define the factor cost ratio formula and dynamically analyze the contribution ratio of each factor to the total cost:

[0052]

[0053] Among them, PC i 、PC j Indicates the i-th and j-th principal component values, that is, the scores of each factor in PCA, β j represents the regression coefficient of the jth factor.

[0054] Optionally, in step S5,

[0055] The cost database is built based on a distributed cloud-native architecture, using a three-tier storage model: storing frequently updated data from the past three months, archiving historical cost data from the past three years, and storing static data for the long term.

[0056] Building a real-time data pipeline based on the Kafka-Flink stream processing framework: The ERP system captures cost change events through CDC (Change Data Capture) and sends them to a Kafka topic (topic = cost_update); Flink filters abnormal data in real time and verifies data consistency through UDF (User Defined Function) formula calls.

[0057]

[0058] The cleaned data is written to the cloud database in batches through the Flink JDBC connector;

[0059] An incremental learning algorithm is used to dynamically update the prediction model to achieve adaptive cost trend analysis: time series features, external variables, and BOM version features are extracted, and future cost trends are predicted based on the Prophet-LSTM hybrid model. The model loss function is defined as:

[0060] Loss=λ1·MAE+λ2·MAPE+λ3·R 2 (λ1=0.5, λ2=0.3, λ3=0.2) (9)

[0061] When the forecast error exceeds the threshold for three consecutive periods, the baseline recalculation is automatically triggered. The update formula is:

[0062] New baseline = α·old baseline + (1-α)·rolling prediction value (α=0.7) (10);

[0063] Build a multi-level early warning and automatic optimization mechanism: Use core indicators to alert abnormal events; locate cost anomaly drivers based on SHAP values, trigger preset optimization actions, and quantify the expected cost reduction effects:

[0064]

[0065] The present invention has the following beneficial effects:

[0066] The present invention proposes a multi-level cost decomposition and accounting method for distribution network materials. By constructing a special BOM model that includes dimensions such as materials, processes, and resources, and using database technology to ensure data consistency and traceability, it achieves refined cost aggregation and step-by-step decomposition. Combined with the ERP system to integrate cost data, statistical analysis and data mining are used to identify key cost factors and predict cost trends. At the same time, a multivariate linear regression model is used to accurately calculate the cost ratio and the results are displayed through visualization technology to provide decision support for cost control. In addition, a real-time updated cost database is established, and data mining technology is used for dynamic monitoring and management to continuously optimize the cost structure. This method significantly improves cost accounting accuracy and management efficiency, provides enterprises with a scientific and systematic cost control solution, enhances market competitiveness, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a flow chart of the multi-level cost decomposition and accounting method for distribution network materials provided by the present invention. DETAILED DESCRIPTION

[0068] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0069] Example

[0070] The present invention provides a multi-level cost decomposition and accounting method for distribution network materials, referring to Figure 1 As shown, the following steps are included:

[0071] S1: Build a dedicated BOM model based on the characteristics of distribution network materials, including dimensions such as materials, processes, and resources, and use database technology to implement version control to ensure data consistency and traceability;

[0072] The dedicated BOM (Bill of Materials) model is constructed using a multi-dimensional layered architecture, combined with database transaction management and version control technology to achieve structured storage and dynamic traceability of all-factor data of distribution network materials. The specific implementation steps are as follows:

[0073] 1. Multi-dimensional BOM model layered design

[0074] The dedicated BOM model includes three core dimensions: material, process, and resource. It uses a unique identifier (UUID) to link data across different hierarchies. The model structure is shown in Table 1.

[0075] Table 1 Multi-dimensional BOM model hierarchical field definition

[0076]

[0077] To support dynamic expansion, the model allows for the addition of customized attributes (such as environmental certifications and quality inspection standards) through extension fields. UUIDs (such as 550e8400-e29b-41d4-a716-446655440000) are used to ensure cross-system data uniqueness, avoiding data inconsistencies caused by multiple version conflicts in traditional BOM models. Table 1 shows the hierarchical field definitions and associations of the multi-dimensional BOM model. The material layer is associated with the process layer's process number through the material code. The process layer then binds the resource layer data through the resource type, forming an end-to-end structured data link.

[0078] 2. Version control and data consistency assurance

[0079] To ensure data consistency, this step uses database transaction management technology to design semantic version number rules and change traceability mechanisms. The version number adopts the format of Major.Minor.Patch (such as V2.1.3), where Major indicates a major structural change (such as adding a new layer), Minor corresponds to the process or resource field expansion, and Patch is used for data correction (such as updating the unit price). Formula (1) defines the calculation logic of the version change impact coefficient Δ, which guides priority decisions by quantifying the sensitivity of field changes to cost accounting. For example, when a new supplier field is added to the resource layer, Δ is calculated as 0.25 (affecting the cost accounting scope), indicating the sensitivity of this change to subsequent cost collection. Through transaction management rules, the system automatically triggers the synchronous update of process and resource data when the material layer data changes, and prevents concurrency conflicts through row-level locks to ensure strong consistency of cross-level data. The example of the version control table in Table 2 shows that when a new field is added, the weight coefficient α of the change impact coefficient Δ is set to 0.6, reflecting the direct impact of the new field on cost accounting.

[0080] Table 2 Version Control Table

[0081]

[0082]

[0083] Among them, N 新增 N is the number of new fields; 删除is the number of fields to be deleted; α and β are weight coefficients (default α=0.6, β=0.4).

[0084] 3. Data association and traceability implementation

[0085] Data association and traceability are achieved through UUID mapping tables and database triggers. The UUID mapping table establishes parent-child data associations. For example, the UUID (550e8400-e29b-41d4) of material code M001 is associated with the UUID (6ba7b810-9dad) of crimping process P001, which is further bound to the UUID (7a8b9c0d-ae0f) of resource R001, forming an end-to-end data link. When the cable supplier changes, the system automatically upgrades the version number (e.g., V1.0.0 → V1.1.0), triggering a transaction to update the supplier field of the associated process and recording the change impact coefficient Δ = 0.3 (due to the resource-level unit price adjustment).

[0086] Table 3 UUID mapping table

[0087]

[0088] S2: Based on a dedicated BOM model, the distribution network material costs are broken down into components, parts, and spare parts step by step. The ERP system is used to integrate material, processing, labor, and other cost data to achieve refined cost aggregation.

[0089] A multi-level cost decomposition approach based on a dedicated BOM model integrates hierarchical cost mapping rules with ERP system data, enabling the gradual aggregation of distribution network material costs from total cost to components, parts, and subassemblies. This approach breaks through the limitations of traditional ERP systems, which only support single-level cost accounting, and for the first time dynamically links process parameters and resource consumption with ERP financial data, forming a closed-loop cost data loop covering the entire lifecycle.

[0090] 1. Cost decomposition process design

[0091] Based on the three-layer structure of the dedicated BOM model (material layer, process layer, resource layer), a three-level cost decomposition logic is constructed:

[0092] (1) Component-level cost: Taking a functional module as a unit (e.g., a cable terminal assembly), the direct material costs of all components (e.g., crimp terminals, insulation sleeves) are summarized;

[0093] (2) Component-level cost: Disassembling the assembly into specific components (such as crimp terminals), adding the processing costs (such as stamping, tinning) and labor costs;

[0094] (3) Part-level cost: This is broken down into individual parts (such as copper core conductors), and is associated with the purchase price, transportation costs, and inventory management costs in the ERP system.

[0095] Table 4 defines the three-level cost decomposition fields and clarifies the mapping relationship between the cost components at each level and the ERP system data source.

[0096] Table 4 Cost breakdown field definition and ERP mapping

[0097]

[0098]

[0099] 2.ERP system data integration rules

[0100] Synchronize data with the ERP system in real time through the middleware interface and establish the following mapping rules:

[0101] (1) Material code mapping: The material code of the BOM model (such as M001) is associated with the material master data (Material Master) of the ERP system to obtain real-time purchase price and inventory status;

[0102] (2) Process mapping: The process number in the BOM model is associated with the standard process path and process database in the ERP system to obtain the standard process flow, single-piece labor time and equipment matching parameters, supporting the calculation of component-level processing costs and equipment resource consumption;

[0103] (3) Resource consumption mapping: Resource layer data (such as labor hours R001) is connected to the ERP human resources module to obtain the difference between the actual labor hour rate and the overtime rate.

[0104] Component-level processing fee calculation:

[0105]

[0106] Where n is the number of processes involved in the component; labor rate i is the rate of the i-th process in the ERP system work order (e.g., stamping ￥30 / hour, tinning ￥20 / hour).

[0107] 3. Dynamic cost collection and anomaly verification

[0108] Build a two-way verification mechanism based on the real-time data flow of the ERP system:

[0109] (1) Forward aggregation: cost accounting tasks are issued step by step from the BOM model to the ERP module, for example, obtaining copper price fluctuation data from the procurement module to update the part-level cost;

[0110] (2) Reverse verification: When the ERP system detects that the material price increase exceeds a threshold (such as ±10%), it triggers a BOM model version upgrade and recalculates the component-level cost impact.

[0111] Table 5 shows the cost aggregation results and anomaly verification logic of a cable terminal head component.

[0112] Table 5 Cost aggregation example and anomaly verification

[0113]

[0114]

[0115] When the processing fee exceeds the limit, the system automatically freezes the work order, notifies the process department to review the efficiency of the tinning process, and triggers formula (3) to recalculate the reasonable rate:

[0116]

[0117] S3: Using statistical analysis and data mining methods, we conduct regression analysis and principal component analysis on the collected cost data to identify key factors affecting costs and predict cost change trends;

[0118] By integrating multivariate statistical analysis with machine learning algorithms, a cost driver identification and dynamic forecasting model was constructed, enabling in-depth analysis and trend prediction of distribution network material costs. This approach breaks through the limitations of traditional cost analysis, which relies on empirical judgment. For the first time, it integrates process parameters, resource consumption, and external market data into a model, forming a data-driven cost forecasting system.

[0119] 1. Data-driven cost factor modeling

[0120] Based on the three-level cost data (components, parts, and components) collected by S2, a multi-source heterogeneous data fusion model is constructed:

[0121] (1) Variable definition: 15 internal variables, including material cost, processing hours, labor rate, and equipment depreciation rate, as well as 6 external variables, including copper price index and labor wage growth rate, are selected to form an input matrix X containing 21-dimensional features. The target variable is the total cost Y.

[0122] (2) Regression analysis: Ridge regression is used to solve the problem of multicollinearity and establish a quantitative relationship between cost driving factors and total cost:

[0123]

[0124] Among them, λ is the regularization coefficient (determined to be 0.1 through cross-validation), β iTable 6 shows the regression coefficients and significance levels of key factors, reflecting the impact of each factor on cost. For example, the coefficient β of the copper price index (CPI) is 1.32 (p < 0.01), indicating that its elasticity of impact on total cost is 1.32 times.

[0125] Table 6 Regression analysis results and factor significance

[0126]

[0127] (3) Principal Component Analysis (PCA): Dimensionality reduction is performed on the 21-dimensional features, and the first three principal components (cumulative contribution rate ≥ 85%) are extracted to reveal the core driving dimensions of cost fluctuations:

[0128] Principal component 1 (PC1): Material cost fluctuation (contribution rate 52%), including copper price, aluminum usage, etc.

[0129] Principal component 2 (PC2): process efficiency (contribution rate 23%), covering processing hours, equipment failure rate, etc.;

[0130] Principal component 3 (PC3): external environment (contribution rate 10%), such as labor wage index and exchange rate fluctuations.

[0131] 2. Cost change trend prediction model

[0132] Build a dynamic prediction model based on time series analysis and integrated learning:

[0133] (1) Data segmentation: historical data is divided into a training set (70%) and a test set (30%) in chronological order, and external variables (such as monthly CPI data) are aligned at a quarterly frequency.

[0134] (2) Model selection: The Prophet algorithm is used to capture the periodicity and holiday effects of cost data, and the LSTM neural network is combined to capture nonlinear trends. The model loss function is defined as:

[0135] Loss=α·MAE+(1-α)·RMSE+β·R 2 (5)

[0136] α = 0.6 is used to balance prediction error and model interpretability; β = 0.2 constrains model complexity.

[0137] S4: Based on the identified key cost factors, a multivariate linear regression model is constructed to accurately calculate the cost proportion of each factor and intuitively display it using data visualization technology to provide decision support for cost control;

[0138] Based on the quantitative analysis of key cost factors, a dynamic multivariate linear regression model is constructed. Through cost ratio calculation and visualization technology, transparent analysis of cost structures and decision support are achieved. This method breaks through the limitations of traditional cost analysis, which often involve isolated variables and ambiguous causal relationships. For the first time, it combines the principal component contribution with the regression coefficient to form an interpretable cost-driven model. Furthermore, it uses interactive visualization tools to transform complex data into decision-making rules.

[0139] 1. Core model construction

[0140] A multiple linear regression model was established with the principal component factors (PC1-PC3) and key original variables (such as material price index and process efficiency) as independent variables and total cost as the dependent variable.

[0141]

[0142] Among them, β i Solved using the least squares method, it reflects the marginal contribution of each factor to the total cost, with ∈ being the error term. The model quantifies the influence of factors through standardized coefficients (β′). For example, the material fluctuation factor β1′ = 0.65 indicates that it is more sensitive to cost fluctuations than other variables.

[0143] 2. Calculation of cost ratio

[0144] Define the factor cost ratio formula and dynamically analyze the contribution ratio of each factor to the total cost:

[0145]

[0146] This formula combines the standardized values of the principal components with the regression coefficients to accurately calculate the cost weights of factors such as materials, processes, and labor. For example, when the material factor accounts for more than 50%, it is directly mapped to the supply chain optimization priority.

[0147] S5: Establish a distribution network material cost database, update cost data in real time, use data mining technology to analyze cost trends, realize dynamic monitoring and management of costs, and continuously optimize the cost structure.

[0148] By building a cloud-based dynamic knowledge base and intelligent analysis engine, establishing a full lifecycle cost database, and combining streaming data processing with adaptive data mining technologies, this approach enables real-time monitoring, trend forecasting, and closed-loop structural optimization of distribution network material costs. This approach transcends the limitations of traditional databases, which only support static storage, and for the first time combines incremental learning with dynamic adjustment of cost baselines to form a self-optimizing cost management system.

[0149] 1. Cloud cost database architecture design

[0150] The cost database is built based on a distributed cloud-native architecture, using a three-tier storage model:

[0151] (1) Hot data layer: stores high-frequency updated data (such as real-time purchase prices and work order status) for the past three months, supporting millisecond-level query responses;

[0152] (2) Warm data layer: archives historical cost data from the past three years (such as project settlement records and supplier contracts) for trend analysis and model training;

[0153] (3) Cold data layer: long-term storage of static data such as regulations, policies, and material standards, reducing storage costs through columnar compression.

[0154] Table 7 defines the core data table structure. For example, the cost event table (cost_event) is associated with the BOM version number (bom_version) through a timestamp (timestamp), supporting cross-version cost tracing.

[0155] Table 7 Cost database core table structure

[0156]

[0157] 2. Real-time update mechanism for streaming data

[0158] Building real-time data pipelines based on the Kafka-Flink stream processing framework:

[0159] (1) Data collection: The ERP system captures cost change events (such as supplier quotation updates) through CDC (Change Data Capture) and sends them to a Kafka topic (topic = cost_update);

[0160] (2) Data cleaning: Flink filters abnormal data (such as negative unit prices) in real time and verifies data consistency by calling formula (8) through UDF (User Defined Function):

[0161]

[0162] (3) Data writing: The cleaned data is written to the cloud database in batches through the Flink JDBC connector, with a write latency of ≤500ms.

[0163] 3. Cost trend mining and dynamic baseline adjustment

[0164] Adopting incremental learning algorithms to dynamically update the forecast model and achieve adaptive cost trend analysis:

[0165] (1) Feature engineering: extracting time series features (such as moving average MA7), external variables (such as copper price index), and BOM version features (such as process complexity);

[0166] (2) Model training: Based on the Prophet-LSTM hybrid model, the future cost trend is predicted. The model loss function is defined as:

[0167] Loss=λ1·MAE+λ2·MAPE+λ3·R 2 (λ1=0.5, λ2=0.3, λ3=0.2) (9)

[0168] (3) Dynamic baseline adjustment: When the forecast error exceeds the threshold for three consecutive periods (e.g., MAPE>8%), the baseline recalculation is automatically triggered. The update formula is:

[0169] New baseline = α·old baseline + (1-α)·rolling prediction value (α=0.7) (10)

[0170] 4. Dynamic cost monitoring and closed-loop optimization

[0171] Build a multi-level early warning and automatic optimization mechanism:

[0172] (1) Real-time monitoring dashboard: Grafana is used to visualize core indicators such as cost deviation (e.g., current cost / baseline cost) and trend slope. Red alerts indicate abnormal events with deviations greater than 15%.

[0173] (2) Root Cause Analysis Engine: Locates cost anomaly drivers based on SHAP values. For example, when the material cost ratio suddenly increases, it is automatically linked to copper price fluctuations or BOM version changes.

[0174] (3) Optimization strategy execution: triggering preset optimization actions (such as switching suppliers, adjusting production batches), and quantifying the expected cost reduction effect through formula (11):

[0175]

[0176] Embodiments of the present invention further provide a storage medium, which stores a computer program. When executed by a processor, the computer program implements some or all of the steps of each embodiment of the method for multi-level cost decomposition and accounting of distribution network materials of the present invention. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0177] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus a necessary general-purpose hardware platform. Based on this understanding, the technical solutions in the embodiments of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention or certain portions of the embodiments.

[0178] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multi-level cost decomposition and accounting method for distribution network materials, characterized in that: The steps include: S1: Build a multi-dimensional dedicated BOM model and establish a structured database; S2: Based on a dedicated BOM model, the distribution network material costs are broken down into components, parts, and spare parts step by step. Expense data is integrated into the ERP system to achieve refined cost aggregation. S3: Using statistical analysis and data mining methods, we conduct regression analysis and principal component analysis on the collected cost data to identify key factors affecting costs and predict cost change trends; S4: Based on the identified key cost factors, a multivariate linear regression model is constructed to accurately calculate the cost proportion of each factor and intuitively display it using data visualization technology; S5: Establish a distribution network material cost database, update cost data in real time, use data mining technology to analyze cost trends, realize dynamic monitoring and management of costs, and continuously optimize the cost structure.

2. A multi-level cost decomposition and accounting method for distribution network materials according to claim 1, characterized in that: In step S1, a multi-level dedicated BOM model including a bill of materials (BOM), process path, and resource consumption is constructed based on the physical properties, process flow, and resource constraint characteristics of the distribution network materials; database transaction management and version control technology are used to establish a structured database, which includes basic materials, process parameters, and resource unit prices, and supports multi-version data traceability and consistency maintenance.

3. A multi-level cost decomposition and accounting method for distribution network materials according to claim 2, characterized in that: In step S1, the dedicated BOM model realizes cross-level data association among the three core dimensions of material layer, process layer and resource layer through a unique identifier (UUID). The material layer is associated with the process number of the process layer through the material code, and the process layer is then bound to the resource layer data through the resource type to form an end-to-end structured data link; the dedicated BOM model can add customized attributes through extended fields.

4. A multi-level cost decomposition and accounting method for distribution network materials according to claim 2, characterized in that: In step S1, a semantic version numbering rule is designed, and the version number adopts the format of Major.Minor.Patch, where Major indicates a major structural change, Minor corresponds to a process or resource field expansion, and Patch is used for data correction. A calculation formula for the version change impact coefficient Δ is defined. This coefficient guides priority decisions by quantifying the sensitivity of field changes to cost accounting. Among them, N 新增 N is the number of new fields; 删除 is the number of deleted fields; α and β are weight coefficients; The data traceability is achieved through database triggers; through transaction management rules, when the material layer data changes, the process and resource data are automatically updated synchronously, and row-level locks are used to prevent concurrency conflicts.

5. A multi-level cost decomposition and accounting method for distribution network materials according to claim 1, characterized in that: The step S2 is specifically as follows: Based on the three-layer structure of material layer, process layer and resource layer of the dedicated BOM model, the three-level cost decomposition logic of components, parts and components is constructed. Define three-level cost decomposition fields and clarify the mapping relationship between cost components at each level and ERP system data sources; Synchronize data with the ERP system in real time through the middleware interface to establish material coding mapping rules, process mapping rules, and resource consumption mapping rules; Build a two-way verification mechanism based on the real-time data flow of the ERP system.

6. A multi-level cost decomposition and accounting method for distribution network materials according to claim 5, characterized in that: In the step S2, The logic for constructing the three-level cost decomposition is: Component-level cost: Taking a functional module as a unit, summarize the direct material costs of all the components it contains; Component-level cost: Disassemble the assembly into specific parts and add up the processing costs and labor hours. Part-level cost: Break it down to individual parts and link it to the purchase price, transportation costs, and inventory management costs in the ERP system; The material code mapping rules, process mapping rules and resource consumption mapping rules are as follows: (1) Material code mapping: The material code of the BOM model is associated with the material master data of the ERP system to obtain real-time purchase prices and inventory status, supporting part-level material cost and inventory management cost accounting; (2) Process mapping: The process number in the BOM model is associated with the standard process path and process database in the ERP system to obtain the standard process flow, single-piece labor time and equipment matching parameters, supporting the calculation of component-level processing costs and equipment resource consumption; (3) Resource consumption mapping: The resource layer data is connected to the ERP human resources and equipment resource modules to obtain labor hour rates, overtime rates and equipment energy consumption unit prices, and to calculate the resource consumption costs of each process in detail; Component-level processing fee calculation: Where n is the number of processes involved in the component; labor rate i is the rate of the i-th process in the ERP system work order; The bidirectional verification mechanism is constructed as follows: (1) Forward aggregation: cost accounting tasks are distributed from the BOM model to the ERP module step by step; (2) Reverse verification: When the ERP system detects that the material price increase exceeds the threshold, it triggers the BOM model version upgrade and recalculates the component-level cost impact.

7. A multi-level cost decomposition and accounting method for distribution network materials according to claim 1, characterized in that: In step S3, principal component analysis (PCA) is used to reduce the dimension of the collected cost data, extract the core cost factors, and construct a multivariate linear regression model to quantify the contribution of each factor; Among them, Y is the total cost, X i is the input matrix, λ is the regularization coefficient, β i Reflects the impact of each factor on cost.

8. A multi-level cost decomposition and accounting method for distribution network materials according to claim 1, characterized in that: In step S3, a dynamic prediction model is constructed based on time series analysis and integrated learning: Divide historical data into training and test sets in chronological order, and align external variables at quarterly frequencies; The Prophet algorithm is used to capture the periodicity and holiday effects of cost data, and the LSTM neural network is combined to capture nonlinear trends. The model loss function is defined as: Loss=α·MAE+(1-α)·RMSE+β·R 2 (5) α = 0.6 is used to balance prediction error and model interpretability; β = 0.2 constrains model complexity.

9. A multi-level cost decomposition and accounting method for distribution network materials according to claim 1, characterized in that: In step S4, a multiple linear regression model is established with the principal component factors and key original variables as independent variables and the total cost as the dependent variable. Among them, β i Solved by the least square method, it reflects the marginal contribution of each factor to the total cost, ∈ is the error term; Define the factor cost ratio formula and dynamically analyze the contribution ratio of each factor to the total cost: Among them, PC i 、PC j Indicates the i-th and j-th principal component values, that is, the scores of each factor in PCA, β j represents the regression coefficient of the jth factor.

10. A multi-level cost decomposition and accounting method for distribution network materials according to claim 1, characterized in that: In the step S5, The cost database is built based on a distributed cloud-native architecture, using a three-tier storage model: storing frequently updated data from the past three months, archiving historical cost data from the past three years, and storing static data for the long term. Building a real-time data pipeline based on the Kafka-Flink stream processing framework: The ERP system captures cost change events through CDC (Change Data Capture) and sends them to a Kafka topic (topic = cost_update); Flink filters abnormal data in real time and verifies data consistency through UDF (User Defined Function) formula calls. The cleaned data is written to the cloud database in batches through the Flink JDBC connector; An incremental learning algorithm is used to dynamically update the prediction model to achieve adaptive cost trend analysis: time series features, external variables, and BOM version features are extracted, and future cost trends are predicted based on the Prophet-LSTM hybrid model. The model loss function is defined as: Loss=λ1·MAE+λ2·MAPE+λ3·R 2 (λ1=0.5,λ2=0.3,λ3=0.2) (9) When the forecast error exceeds the threshold for three consecutive periods, the baseline recalculation is automatically triggered. The update formula is: New baseline = α·old baseline + (1-α)·rolling prediction value (α=0.7) (10); Build a multi-level early warning and automatic optimization mechanism: Use core indicators to alert abnormal events; locate cost anomaly drivers based on SHAP values, trigger preset optimization actions, and quantify the expected cost reduction effects:

Images