Power grid supply chain distribution network protocol inventory material demand prediction method based on time sequence model

Abstract

The invention discloses a power grid supply chain distribution network protocol inventory material demand prediction method based on a time sequence model, and belongs to the field of data processing. Comprising the steps of data collection and preprocessing; determining a dimension establishment model; model prediction; and correcting the model. The method comprises the following steps: introducing a time sequence model into power grid supply chain demand prediction, and establishing a time sequence model suitable for power grid supply chain distribution network protocol inventory material demands; by means of a big data analysis method and an artificial intelligence algorithm, an association rule of project and demand prediction is mined, a material demand characteristic portrait is formed by using a data analysis method, and through historical data cleaning, model establishment and event management, the material characteristic portrait is carded, analyzed, diagnosed and evaluated in an all-around and multi-view manner. The operation capability and the data applicability of an algorithm model library are improved, a material demand prediction mode is optimized and evaluated, and a support is provided for a power grid supply chain material purchase supply decision. The method can be widely applied to the field of power grid supply chain distribution network protocol inventory material demand prediction and management.

Description

Technical Field

[0001] The present invention belongs to the field of data processing, and in particular relates to a method for predicting inventory material demand for a power grid supply chain distribution network protocol. Background Art

[0002] Material management is an important part of enterprise management. It is the planning, organization and control of the ordering, transportation, storage and supply of various materials needed in the production process of the enterprise.

[0003] Good material management is conducive to the rational use and conservation of materials, improving product quality, reducing production costs, accelerating capital turnover, and increasing corporate profits.

[0004] As the front end of supply chain management, power grid material planning management is the basis for material supply security.

[0005] Power grid material planning management is an important part of power grid enterprise management, and its management level directly affects the efficiency and long-term development of the enterprise.

[0006] With the rapid development of the power industry and the advancement of smart grid construction, the management of power grid supply chain has become increasingly important. As an important part of the power grid supply chain, the accuracy of power grid material planning (also known as distribution network agreement inventory or agreement inventory) is directly related to the stable operation and cost control of the power grid.

[0007] The forecast management of agreement inventory demand is to forecast the scale of procurement of major materials and services required in the professional fields of production safety, power grid development and marketing services of the company in a certain period of the future. Its purpose is to make full use of historical data, combine the company's annual budget control target and project planning and other influencing factors, apply scientific and reasonable methods, obtain practical demand data, and provide decision-making reference for annual work arrangements.

[0008] In recent years, although various power companies have conducted preliminary exploratory research and business practices in the annual demand forecasting of agreement inventory materials, there are still some shortcomings and deficiencies, and there is still a certain gap from achieving the goal. Specifically, the forecast accuracy of distribution network agreement inventory materials is not high enough, and the research on forecasting models and influencing factors needs to be further deepened. Due to the lack of big data analysis methods and scientific forecasting models, the demand plan deviates greatly from the actual needs, resulting in the subsequent supply stage agreement matching execution rate being too high or too low, which to a certain extent affects the supply guarantee of power materials.

[0009] The invention patent application with application publication number CN106600032A and application publication date 2017.04.26 discloses "a method and device for predicting demand for inventory materials", which includes: according to the initial investment plan, obtaining the planned investment amount in each project and the engineering attributes corresponding to each project; according to the engineering attributes corresponding to each project, obtaining the material subcategories contained in each project, searching the preset corresponding relationship list of material subcategories and engineering attributes and prediction models, and obtaining the prediction model corresponding to each material subcategory; inputting the planned investment amount into the prediction model corresponding to the corresponding material subcategory, and calculating the predicted demand of each material subcategory in each project; adding the predicted demand of each material subcategory in different projects, and calculating the total predicted demand of each material subcategory. The method and device for predicting demand for inventory materials can improve the efficiency and accuracy of inventory material prediction by pre-constructing corresponding prediction models for different material subcategories under each project. However, since it only pre-builds corresponding prediction models for different material categories under each project, the actual implementation process requires a large amount of mathematical modeling work, and the accuracy of the prediction model requires a verification process. Once external conditions (such as changes in the supply price of a single material or the supply channel of a certain material) change, the early modeling and prediction models must be started again. The entire project implementation process takes a long time from the beginning to the achievement of results, which affects the actual user experience and usage effect.

[0010] The invention patent application with application publication number CN111882262A and application publication date 2020.11.03 discloses "a method for accurately predicting demand for agreed inventory materials", including step 1: according to the production plan, obtain the total investment amount of the production plan; step 2: count the inventory records, and record the inventory reserve information remaining in the warehouse in detail according to the items required by the production plan; step 3: list the items required for the production plan, and count the number of materials required for each project; step 4: subtract the number of inventory from the required project materials, and count the funds required for each project and summarize them. It helps to reduce cost investment and save resources by counting the inventory quantity; all data are input into the computer program and calculated using the clustering algorithm and the calculation results are saved in ES. While predicting, the material procurement market price data input method can accurately predict the actual demand for materials and avoid large differences between actual funds and planned funds. However, this technical solution is based on inventory quantity statistics. If the inventory data cannot be updated in time or is temporarily unavailable, the data source and implementation basis of the solution will be affected. In addition, it uses a "clustering" algorithm for calculation and saves the calculation results to ES (Elasticsearch, a non-relational database for distributed documents, built on Apache Lucene, designed to process large amounts of data and support complex data analysis; it can be widely used in full-text search, real-time analysis and other scenarios, and is particularly good at processing unstructured or semi-structured data). It has high technical requirements for computer hardware and computing tools, and the computing power of the algorithm model library and the applicability of the database are affected to a certain extent.

[0011] Based on the current technical conditions and management level, how to use modern big data analysis methods and artificial intelligence algorithms to explore the correlation between projects and demand forecasts, use data analysis methods to optimize demand forecasting models, and provide support for material procurement and supply decisions is a technical problem that urgently needs to be solved in the actual power grid material planning management work. Summary of the invention

[0012] The technical problem to be solved by the present invention is to provide a method for predicting the demand for inventory materials in the power grid supply chain distribution network agreement based on a time series model. It uses modern big data analysis methods and artificial intelligence algorithms to mine the correlation between projects and demand forecasts, uses data analysis methods to form a material demand characteristic portrait, and through historical data cleaning, model building, and event management, comprehensively and multi-perspectively combs, analyzes, diagnoses, and evaluates the material characteristic portrait, optimizes the demand forecast model, and provides support for material procurement and supply decisions.

[0013] The technical solution of the present invention is to provide a method for predicting the demand for distribution network agreement inventory materials in a power supply chain based on a time series model, including the prediction and management of agreement inventory demand, which is characterized in that the prediction and management of agreement inventory demand are carried out according to the following steps:

[0014] 1) Basic data collection and data preprocessing:

[0015] Basic data collection includes system connection data and protocol preparation data;

[0016] The system connection data shall at least include: evaluation material master data, project unit master data, project master data, warehouse master data, outbound data, and framework agreement winning status;

[0017] The agreement preparation data includes at least the bid package master data, procurement catalog, bid package materials, new and old material comparison, project investment plan, and rolling forecast new and old material comparison;

[0018] After the basic data collection is completed, the data is cleaned and operations including removing negative values ​​and outliers are performed to complete the preprocessing of the basic data;

[0019] 2) Determine the dimensions:

[0020] Mining the correlation between projects and demand forecasts, using data analysis methods to form a profile of material demand characteristics, using the standard package dimension of outbound data after data cleaning and preprocessing, and monthly summary quantity as parameters, to conduct a profile analysis to evaluate material characteristics;

[0021] 3) Build the model:

[0022] Introduce time series models into the forecast of power supply chain demand. Based on the results of the portrait analysis of the material characteristics, use three types of model combinations: linear and nonlinear seasonal combination model, trend and seasonal autoregressive combination model, and discontinuous and seasonal combination model. According to the analysis and calculation of power data, make corresponding selections from the model library, and determine the most suitable model combination and related parameters for power data through parameter tuning to complete the intelligent selection of material forecasting models.

[0023] 4) Model prediction:

[0024] Analyze the characteristics of the assessed materials and locate the characteristics of the assessed materials from the perspective of seasonality, volatility, continuity, and importance; select a model based on the characteristics of the assessed materials; and finally output the predicted values ​​of the assessment material package dimensions through the model;

[0025] 5) Model modification:

[0026] By comparing and analyzing the material demand forecast results and actual delivery data, the forecast deviation is statistically calculated and the accuracy analysis is performed. Based on the correspondence between the forecast deviation index and the confidence level, the demand forecast results are put into practice to provide technical support for the procurement and supply decisions of power grid supply chain materials.

[0027] Specifically, the method for predicting the demand for inventory materials in the distribution network agreement of the power supply chain based on the time series model introduces the time series model into the prediction of the demand for the power supply chain, broadens the traditional time series algorithm modeling, and establishes a time series model suitable for the demand for inventory materials in the distribution network agreement of the power supply chain; by using modern big data analysis methods and artificial intelligence algorithms, the correlation between projects and demand forecasts is mined, and a material demand characteristic portrait is formed using data analysis methods. Through historical data cleaning, model building, and event management, the material characteristic portrait is comprehensively and multi-perspectively sorted out, analyzed, diagnosed, and evaluated, the computing power and data applicability of the algorithm model library are improved, the material demand forecasting model is optimized and evaluated, and support is provided for the procurement and supply decision-making of materials in the power supply chain.

[0028] Furthermore, the material property profile analysis at least includes continuity analysis, volatility analysis, and seasonality analysis.

[0029] Furthermore, the forecast deviation=(forecasted outbound amount-actual outbound amount) / actual outbound amount.

[0030] Furthermore, according to the corresponding relationship between the forecast deviation index and the confidence level, the confidence levels of various materials are divided into four levels according to the forecast deviation results:

[0031] a. Forecast deviation <= 20%, confidence level is I;

[0032] b. 20% < prediction deviation <= 40%, the confidence level is II;

[0033] c. 40% < prediction deviation <= 80%, the confidence level is III;

[0034] d. Prediction deviation>80%, confidence level is IV.

[0035] Specifically, for materials with confidence level I, combined with the batch demand coverage and the remaining framework agreement share, the demand forecasting model is used to calculate the results, and the agreement inventory procurement plan is directly formulated after quantity-price conversion;

[0036] For materials with confidence levels II and III, combined with the batch demand coverage and the remaining framework agreement share, the demand forecast model calculation results are combined with the demand submission, and after quantity and price conversion, the agreement inventory procurement plan is jointly formulated;

[0037] For materials with credibility level IV, manual submission is adopted directly, and after quantity and price conversion, an agreement inventory purchase plan is formulated.

[0038] Specifically, the method of determining the model combination and related parameters that best suit the power data through parameter tuning at least includes:

[0039] 1) When selecting a model, consider the model's indicators including goodness of fit and white noise test of residuals;

[0040] 2) Use methods including least squares and maximum likelihood estimation to estimate model parameters using historical data;

[0041] 3) Using test methods including white noise test of residuals, cross validation, and goodness-of-fit test of the model, statistical tests were performed on the parameter estimation results to verify the accuracy and effectiveness of the model;

[0042] 4) Adjust and optimize the model based on the test results to improve the prediction accuracy and stability of the model.

[0043] Furthermore, the preprocessing of the basic data includes:

[0044] 1) Data processing includes negative value removal and outlier processing;

[0045] Remove negative values: Determine whether the current month is a negative value. If it is a negative value, set the current month to 0 and add it to the previous month to get the previous month's delivery value. If the sum is still a negative value, it is still set to 0, and the process is repeated until a non-negative value appears.

[0046] Outlier processing: Based on the monthly summary quantity of the outbound data package dimension, the average value of the outbound volume of the month before and after the month + 3 times the standard deviation X is used as the boundary. Data with a quantity exceeding X is considered abnormal data, and an outlier reminder service is performed for verification; the outlier is judged, and if it is determined to be system abnormal data, it is fed back to the ERP system for adjustment and resynchronization of data;

[0047] 2) Characteristic analysis and division;

[0048] Use the monthly summary quantity of the outbound data package dimension after data cleaning and preprocessing as a parameter to conduct material characteristic profiling analysis;

[0049] Material property profile analysis at least includes continuity analysis, volatility analysis, and seasonality analysis;

[0050] Seasonality and importance are classified as high, medium and low respectively; continuity and volatility are classified as high, relatively high, medium, relatively low and low respectively;

[0051] 3) Model construction and revision;

[0052] For the demand forecasting model, the “top-down” forecasting logic is integrated with the “bottom-up” forecasting model to carry out model construction and correction;

[0053] The "top-down" forecasting logic mentioned above includes using project investment plans and historical project material delivery data to analyze the correlation between project types, voltage levels and major equipment materials, and sort out the corresponding rules between project types and material requirements; and comprehensively consider factors such as annual growth rate and project approval rate to decompose and derive the expected usage quantity and amount of each material subcategory;

[0054] The "bottom-up" forecasting model uses data including historical material delivery data, bid packages, and new and old material comparison tables. On the basis of analyzing the volatility, seasonality, importance, continuity, and project relevance of materials, it uses multiple forecasting models that can handle both continuous and intermittent demand for combined forecasting, so as to achieve automation and intelligence in the selection of forecasting models. It uses simulation "if-else" technology, combined with demand forecasting results, to establish a material intelligent allocation strategy library for different business scenarios, forming a "scenario-resource-strategy" operation model. It uses an automatic optimization algorithm to intelligently weight the forecast results of different forecasting models, effectively improving the accuracy of the demand forecasting model.

[0055] The method for predicting inventory material demand of power grid supply chain distribution network protocol based on time series model described in the technical solution of the present invention adopts simulation "if-else" technology, establishes a material intelligent allocation strategy library for different business scenarios, forms a "scenario-resource-strategy" operation model, and intelligently weights the prediction results of different prediction models through automatic optimization algorithm, thereby effectively improving the accuracy of the demand prediction model.

[0056] The method for predicting the demand for inventory materials in the power supply chain distribution network agreement based on the time series model described in the technical solution of the present invention introduces the time series model into the prediction of the demand for the power supply chain in terms of data modeling and algorithm engineering, broadens the traditional time series algorithm modeling, incorporates more efficient models that can automatically optimize parameters, and uses the combined application of the algorithm model library to improve the prediction accuracy, so as to improve the computing power and data applicability of the algorithm model library;

[0057] At the system operation level, the remaining share of the agreed inventory, the remaining in-transit inventory and the inventory in stock are comprehensively considered to form a "one-stop" online operation mode for bidding and procurement plan submission and supply plan, and the decision-making and instructions are connected through "cloud" deployment to achieve online decision support, real-time monitoring of material supply and demand, and improve the front-end and back-end coordination capabilities;

[0058] Simulate and emulate materials to dynamically improve the level of resource optimization. Use simulation "if-else" technology, combined with demand forecasting results, to establish a smart material allocation strategy library for different business scenarios, form a "scenario-resource-strategy" operation model, adapt to the optimal material strategy, thereby realizing intelligent coordination of material allocation and further improving the level of material services.

[0059] Compared with the prior art, the advantages of the present invention are:

[0060] 1. The technical solution of the present invention has greatly improved the accuracy of demand forecasting on the basis of existing research on material demand forecasting projects, which is mainly reflected in data modeling and algorithm engineering. It broadens the traditional time series algorithm modeling and incorporates more efficient models with automatic parameter optimization, including Prophet, Auto-Arima, TBATS, etc., which effectively improves the computing power and data applicability of the algorithm model library, thereby improving the prediction accuracy of the combined application of the algorithm model library.

[0061] 2. The technical solution of the present invention improves the front-end and back-end coordination capabilities by real-time monitoring of material supply and demand; at the system operation level, it comprehensively considers the remaining share of the agreed inventory, the remaining inventory in transit and the inventory in stock, and forms a "one-stop" online operation mode for bidding and procurement plan submission and supply plan, connects the decision-making and instructions through "cloud" deployment, and realizes online decision support.

[0062] 3. The technical solution of the present invention improves the level of resource optimization by simulating the dynamics of materials. It uses the simulation "if-else" technology, combined with the results of demand forecasting, to establish a material intelligent allocation strategy library for different business scenarios, forming a "scenario-resource-strategy" operation model, adapting the optimal material strategy, thereby realizing intelligent coordination of material allocation and further improving the material service level. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic diagram of the overall path of the prediction method of the present invention;

[0064] Figure 2 It is a schematic diagram of the "top-down" prediction logic in the present invention;

[0065] Figure 3 It is a schematic diagram of the "bottom-up" prediction model in the present invention. DETAILED DESCRIPTION

[0066] The present invention will be further described below in conjunction with the accompanying drawings.

[0067] A time series is a collection of data points arranged in chronological order that can be used to predict future values. The time series model predicts future demand trends based on past observations and helps companies make more accurate demand forecasts and plans by analyzing the seasonality, trend, and cyclical characteristics of historical data.

[0068] In supply chain demand forecasting, commonly used time series models include moving average models, exponential smoothing models, autoregressive moving average models (ARMA), autoregressive integrated moving average models (ARIMA) and seasonal autoregressive integrated moving average models (SARIMA).

[0069] The moving average model is a method that uses the average of several past periods of data to predict future demand. It can smooth out outliers, reduce forecast errors, and is suitable for situations where demand changes relatively smoothly. However, the moving average model does not work well when demand has large fluctuations or seasonal changes.

[0070] Exponential smoothing is a method of smoothing past observations with different weights. By gradually reducing the weights, it pays more attention to recent observations and can better adapt to the changing trend of demand. Exponential smoothing is suitable for situations where the demand curve has a linear trend.

[0071] The autoregressive moving average model (ARMA) is a random process model based on time series, which combines the characteristics of autoregression and moving average. The ARMA model can capture the autocorrelation of time series and the correlation of moving average, and can better model multivariate time series.

[0072] The Autoregressive Inverse Moving Average (ARIMA) model is an extension of the ARMA model, which introduces a difference operation to handle non-stationary time series. The ARIMA model can handle the trend of time series and can also consider seasonal factors. When demand has significant seasonal changes, the ARIMA model can provide more accurate forecast results.

[0073] The Seasonal Autoregressive Integrated Moving Average (SARIMA) model is a further extension of the ARIMA model, specifically designed to handle seasonal time series. The SARIMA model can take into account seasonal trends and cyclical changes, and can provide more accurate forecasts when there is obvious seasonality in demand.

[0074] In addition to the models mentioned above, there are some other time series models that can be used to predict supply chain demand, such as seasonal exponential smoothing model (SES), grayscale model (GM), etc. These models have different characteristics and applicable scenarios. Enterprises can choose the appropriate model for prediction according to their own needs and data characteristics.

[0075] The technical solution of the present invention, through the use of modern big data analysis methods and artificial intelligence algorithms, explores the correlation between projects and demand forecasts, uses data analysis methods to form a material demand characteristic portrait, and through historical data cleaning, model building, and event management, comprehensively and multi-perspectively combs, analyzes, diagnoses, and evaluates the material characteristic portrait, optimizes the demand forecast model, and provides support for material procurement and supply decisions.

[0076] The technical solution of the present invention introduces the time series model into the prediction of power supply chain demand, broadens the traditional time series algorithm modeling, and incorporates more efficient models that can automatically optimize parameters, including Prophet, Auto-Arima, TBATS, etc., effectively improving the computing power and data applicability of the algorithm model library. By mining and analyzing historical data, a time series model suitable for the inventory material demand of the power supply chain distribution network agreement is established, and the model is used to predict demand. This method has high prediction accuracy and stability, and can provide strong support for the material management of the power grid.

[0077] like Figure 1 As shown in , the overall method path of the technical solution of the present invention is as follows:

[0078] (1) Data processing:

[0079] Data collection is divided into two types: system docking data and protocol preparation data. Among them, system docking data mainly includes: material master data, project unit master data, project master data, warehouse master data, outbound data, framework agreement winning situation, etc. Protocol preparation data includes bid package master data, procurement catalog, bid package materials, new and old material comparison, project investment plan, rolling forecast new and old material comparison, etc. At the same time, data cleaning is carried out to remove negative values ​​and outliers, etc., to complete the preprocessing of basic data.

[0080] (2) Determine the dimension:

[0081] The monthly summary (quantity) of the outbound data package dimensions after data cleaning and preprocessing is used as a parameter to conduct material property portrait analysis, which mainly includes continuity analysis, volatility analysis, seasonal analysis, etc. Based on the results of the material property portrait analysis, the intelligent selection of the material forecasting model is completed.

[0082] (3) Model building:

[0083] The basic data are all time series data. For bottom-up prediction, three types of model combinations are used: linear and nonlinear seasonal combination model, trend and seasonal autoregressive combination model, and discontinuity and seasonal combination model. According to the analysis and measurement of power data, the models in the model library will be selected accordingly, and the model combination and related parameters that are most suitable for power data will be determined through parameter tuning.

[0084] (4) Model prediction:

[0085] Analyze the characteristics of materials and locate the characteristics of materials from the perspective of seasonality, volatility, continuity, and importance. Then select a model based on the characteristics of the materials. Finally, output the predicted value of the standard package dimension through the model.

[0086] (5) Model modification:

[0087] By comparing and analyzing the demand forecast results with the actual delivery data, the forecast deviation is statistically calculated for accuracy analysis. Based on the corresponding relationship between the forecast deviation index and the confidence level, the demand forecast results are put into practical application.

[0088] Specifically, the above-mentioned forecast deviation = (forecasted outbound amount - actual outbound amount) / actual outbound amount;

[0089] During implementation, the confidence levels of various materials can be divided into four levels according to the corresponding relationship between the forecast deviation index and the confidence level and the forecast deviation results:

[0090] a. Prediction deviation <= 20%, confidence level I:

[0091] For materials with confidence level 1, the demand forecasting model is used to calculate the results based on the batch demand coverage and the remaining framework agreement share, and the agreement inventory procurement plan is directly formulated after quantity-price conversion;

[0092] b. 20% < prediction deviation <= 40%, confidence level II;

[0093] 40%<prediction deviation<=80%, confidence level III.

[0094] For materials with confidence levels 2 and 3, combined with the batch demand coverage and the remaining framework agreement share, the demand forecast model calculation results are combined with the demand submission, and after quantity and price conversion, the agreement inventory procurement plan is jointly formulated;

[0095] c, Grade IV: prediction bias >80%;

[0096] For materials with credibility level 4, manual submission is adopted directly, and after quantity and price conversion, an agreement inventory purchase plan is formulated.

[0097] Furthermore, in this technical solution, the processing of data (also referred to as implementation means) includes the following steps:

[0098] (1) Data processing:

[0099] Negative value removal: Determine whether the current month is a negative value. If it is a negative value, set the current month to 0 and add it to the previous month to get the previous month's delivery value. If the sum is still a negative value, it is still set to 0, and the process is repeated until a non-negative value appears.

[0100] Abnormal value processing: Based on the monthly summary (quantity) of the outbound data package dimension, the average value of the outbound volume of the month before and after the month + 3 times the standard deviation X is used as the boundary. Data with a quantity exceeding X is abnormal data, and abnormal value reminder business is carried out for verification. Judge the abnormal value, if it is determined to be system abnormal data, feedback to the ERP system for adjustment and resynchronization of data.

[0101] (2) Characteristic analysis division:

[0102] The monthly summary (quantity) of the outbound data package dimension after data cleaning and preprocessing is used as a parameter to conduct material characteristic profile analysis, which mainly includes continuity analysis, volatility analysis, seasonality analysis, etc. Seasonality and importance are divided into high, medium and low, and continuity and volatility are divided into high, relatively high, medium, relatively low and low.

[0103] (3) Model construction and modification:

[0104] like Figure 2 As shown in , in this technical solution, the demand forecasting model adopts a method of integrating the "top-down" forecasting logic with the "bottom-up" forecasting model.

[0105] a. "Top-down" prediction logic:

[0106] It mainly uses project investment plans and historical project material delivery data to analyze the correlation between project types, voltage levels and major equipment materials, and sorts out the correspondence between project types and material demands; and comprehensively considers factors such as annual growth rate and project approval rate to decompose the expected usage quantity and amount of each material subcategory.

[0107] b. "Bottom-up" prediction model:

[0108] The bottom-up forecasting model in this technical solution mainly uses historical material outbound data, bid packages, new and old material comparison tables, etc., and analyzes material characteristics such as material volatility, seasonality, importance, continuity, and project relevance. It uses multiple forecasting models that can handle both continuous and intermittent demand for combined forecasting to achieve automation and intelligence in forecasting model selection; it uses simulation "if-else" technology, combined with demand forecasting results, to establish a material intelligent allocation strategy library for different business scenarios, forming a "scenario-resource-strategy" operation model; it uses an automatic optimization algorithm to intelligently empower the forecast results of different forecasting models, effectively improving the accuracy of the demand forecasting model.

[0109] In addition, the "if-else technology" (also known as if-else structure) in this technical solution is briefly introduced as follows:

[0110] The if-else structure is a flow control statement in Python that can determine the next step of the program based on the truth or falsity of a condition. Its basic syntax is as follows:

[0111] ```

[0112] if condition:

[0113] Code Block 1

[0114] else:

[0115] Code Block 2

[0116] ```

[0117] The condition is an expression that evaluates to True or False. If the condition evaluates to True, code block 1 is executed; otherwise, code block 2 is executed.

[0118] In Python, indentation is used to indicate code blocks, so the indentation of code block 1 and code block 2 must be the same.

[0119] Here is a simple example that demonstrates how to use an if-else structure to determine the parity of a number entered by the user:

[0120] ```

[0121] num=int(input("Please enter a number:"))

[0122] if num%2==0:

[0123] print("This is an even number.")

[0124] else:

[0125] print("This is an odd number.")

[0126] ```

[0127] In this example, the user enters a number. If the result of the number modulo 2 is 0, code block 1 will be executed (that is, the input number is an even number), otherwise code block 2 will be executed (that is, the input number is an odd number).

[0128] Compared with the comparison structures such as the control expression and multi-branch expression in the case statement, the conditional expression in the "if_else" structure is more intuitive.

[0129] like Figure 3 As shown in , the following are some examples of using the simulation "if-else" technology to establish a material intelligent allocation strategy library for different business scenarios, forming a "scenario-resource-strategy" operation model, and intelligently empowering the prediction results of different prediction models through automatic optimization algorithms to effectively improve the accuracy of demand prediction models:

[0130] 1. If seasonality = medium, high; then model selection = linear and nonlinear seasonal combination model;

[0131] 2. If continuity = None; Volatility = Medium, High; Project Specificity = Low, Medium; then Model Selection = Discontinuous and Seasonal Combination Model;

[0132] 3. If seasonality = None; volatility = Medium; project expertise = High; importance = Low; then model selection = autoregressive combination model of trend and seasonality;

[0133] 4.…….

[0134] Since process control in Python programming belongs to the existing technology, its specific implementation steps and processes will not be described in detail here.

[0135] The advantages of this technical solution are:

[0136] 1) Use algorithm model library combination application to improve prediction accuracy:

[0137] Based on the research of existing material demand forecasting projects, the accuracy of demand forecasting has been greatly improved, mainly in terms of data modeling and algorithm engineering. It has broadened the traditional time series algorithm modeling and incorporated more efficient models with automatic parameter optimization, including Prophet, Auto-Arima, TBATS, etc., effectively improving the computing power and data applicability of the algorithm model library.

[0138] 2) Real-time monitoring of material supply and demand to improve front-end and back-end collaboration capabilities:

[0139] At the system operation level, comprehensive consideration is given to the remaining share of the agreed inventory, the remaining inventory in transit and in stock to form a "one-stop" online operation method for bidding and procurement plan submission and supply plan, and the decision-making and instructions are connected through "cloud" deployment to achieve online decision support.

[0140] 3) Simulate materials to dynamically improve resource optimization:

[0141] By using the simulation "if-else" technology and combining it with the results of demand forecasting, we establish an intelligent material allocation strategy library for different business scenarios, form a "scenario-resource-strategy" operation model, and adapt the optimal material strategy, thereby realizing intelligent coordination of material allocation and further improving the level of material services.

[0142] The present invention can be widely used in the fields of power grid supply chain distribution network protocol inventory material demand forecasting and decision-making management.

Claims

1. A method for predicting the demand for distribution network agreement inventory materials in a power supply chain based on a time series model, including the prediction and management of agreement inventory demand, characterized by: The forecasting and management of agreed inventory requirements is carried out in the following steps: 1) Basic data collection and data preprocessing: Basic data collection includes system connection data and protocol preparation data; The system connection data shall at least include: evaluation material master data, project unit master data, project master data, warehouse master data, outbound data, and framework agreement winning status; The agreement preparation data includes at least the bid package master data, procurement catalog, bid package materials, new and old material comparison, project investment plan, and rolling forecast new and old material comparison; After the basic data collection is completed, the data is cleaned and operations including removing negative values ​​and outliers are performed to complete the preprocessing of the basic data; 2) Determine the dimensions: Mining the correlation between projects and demand forecasts, using data analysis methods to form a profile of material demand characteristics, using the standard package dimension of outbound data after data cleaning and preprocessing, and monthly summary quantity as parameters, to conduct a profile analysis to evaluate material characteristics; 3) Build the model: Introduce time series models into the forecast of power supply chain demand. Based on the results of the portrait analysis of the material characteristics, use three types of model combinations: linear and nonlinear seasonal combination model, trend and seasonal autoregressive combination model, and discontinuous and seasonal combination model. According to the analysis and calculation of power data, make corresponding selections from the model library, and determine the most suitable model combination and related parameters for power data through parameter tuning to complete the intelligent selection of material forecasting models. 4) Model prediction: Analyze the characteristics of the assessed materials and locate the characteristics of the assessed materials from the perspective of seasonality, volatility, continuity, and importance; select a model based on the characteristics of the assessed materials; and finally output the predicted values ​​of the assessment material package dimensions through the model; 5) Model modification: By comparing and analyzing the material demand forecast results and actual delivery data, the forecast deviation is statistically calculated and the accuracy analysis is performed. Based on the correspondence between the forecast deviation index and the confidence level, the demand forecast results are put into practice to provide technical support for the procurement and supply decisions of power grid supply chain materials.

2. The method for predicting inventory material demand of power grid supply chain distribution network agreement based on time series model according to claim 1 is characterized by: The method for predicting the demand for inventory materials in the distribution network agreement of the power supply chain based on a time series model introduces the time series model into the prediction of the demand for the power supply chain, broadens the modeling of the traditional time series algorithm, and establishes a time series model suitable for the demand for inventory materials in the distribution network agreement of the power supply chain; by using modern big data analysis methods and artificial intelligence algorithms, the correlation between projects and demand forecasts is mined, and a material demand characteristic portrait is formed using data analysis methods. Through historical data cleaning, model building, and event management, the material characteristic portrait is comprehensively and multi-perspectively sorted out, analyzed, diagnosed, and evaluated, the computing power and data applicability of the algorithm model library are improved, the material demand prediction model is optimized and evaluated, and support is provided for the procurement and supply decision-making of materials in the power supply chain.

3. The method for predicting inventory material demand of power supply chain distribution network agreement based on time series model according to claim 1 is characterized by: The material property profile analysis described herein at least includes continuity analysis, volatility analysis, and seasonality analysis.

4. The method for predicting inventory material demand of power supply chain distribution network agreement based on time series model according to claim 1 is characterized in that The forecast deviation = (forecasted outbound amount - actual outbound amount) / actual outbound amount.

5. The method for predicting inventory material demand of power supply chain distribution network agreement based on time series model according to claim 4 is characterized by: According to the corresponding relationship between the forecast deviation index and the confidence level, the confidence levels of various materials are divided into four levels according to the forecast deviation results: a. Forecast deviation <= 20%, confidence level is I; b. 20% < prediction deviation <= 40%, the confidence level is II; c. 40% < prediction deviation <= 80%, the confidence level is III; d. Prediction deviation>80%, confidence level is IV.

6. The method for predicting inventory material demand of power grid supply chain distribution network agreement based on time series model according to claim 5 is characterized in that: For materials with confidence level I, combined with the batch demand coverage and the remaining framework agreement share, the demand forecasting model is used to calculate the results, and the agreement inventory procurement plan is directly formulated after quantity-price conversion; For materials with confidence levels II and III, combined with the batch demand coverage and the remaining framework agreement share, the demand forecast model calculation results are combined with the demand submission, and after quantity and price conversion, the agreement inventory procurement plan is jointly formulated; For materials with credibility level IV, manual submission is adopted directly, and after quantity and price conversion, an agreement inventory purchase plan is formulated.

7. The method for predicting inventory material demand of power supply chain distribution network agreement based on time series model according to claim 1 is characterized by: The method of determining the model combination and related parameters that best suit the power data through parameter tuning at least includes: 1) When selecting a model, consider the model's indicators including goodness of fit and white noise test of residuals; 2) Use methods including least squares and maximum likelihood estimation to estimate model parameters using historical data; 3) Using test methods including white noise test of residuals, cross validation, and goodness-of-fit test of the model, statistical tests were performed on the parameter estimation results to verify the accuracy and effectiveness of the model; 4) Adjust and optimize the model based on the test results to improve the prediction accuracy and stability of the model.

8. The method for predicting inventory material demand of power supply chain distribution network agreement based on time series model according to claim 1 is characterized by: The preprocessing of the basic data includes: 1) Data processing includes negative value removal and outlier processing; Remove negative values: Determine whether the current month is a negative value. If it is a negative value, set the current month to 0 and add it to the previous month to get the previous month's delivery value. If the sum is still a negative value, it is still set to 0, and the process is repeated until a non-negative value appears. Outlier processing: Based on the monthly summary quantity of the outbound data package dimension, the average value of the outbound volume of the month before and after the month + 3 times the standard deviation X is used as the boundary. Data with a quantity exceeding X is considered abnormal data, and an outlier reminder service is performed for verification; the outlier is judged, and if it is determined to be system abnormal data, it is fed back to the ERP system for adjustment and resynchronization of data; 2) Characteristic analysis and division; Use the monthly summary quantity of the outbound data package dimension after data cleaning and preprocessing as a parameter to conduct material characteristic profiling analysis; Material property profile analysis at least includes continuity analysis, volatility analysis, and seasonality analysis; Seasonality and importance are classified as high, medium and low respectively; continuity and volatility are classified as high, relatively high, medium, relatively low and low respectively; 3) Model construction and revision; For the demand forecasting model, the "top-down" forecasting logic is integrated with the "bottom-up" forecasting model to carry out model construction and correction; The "top-down" forecasting logic mentioned above includes using project investment plans and historical project material delivery data to analyze the correlation between project types, voltage levels and major equipment materials, sort out the corresponding rules between project types and material requirements; and comprehensively consider factors such as annual growth rate and project approval rate to decompose and derive the expected usage quantity and amount of each material subcategory; The "bottom-up" forecasting model uses data including historical material delivery data, bid packages, and old and new material comparison tables. On the basis of analyzing the volatility, seasonality, importance, continuity, and project relevance of materials, it uses multiple forecasting models that can handle both continuous and intermittent demand for combined forecasting to achieve automation and intelligence in forecasting model selection. It uses simulation "if-else" technology, combined with demand forecasting results, to establish a material intelligent allocation strategy library for different business scenarios, forming a "scenario-resource-strategy" operation model. It uses an automatic optimization algorithm to intelligently weight the forecast results of different forecasting models, effectively improving the accuracy of the demand forecasting model.

9. The method for predicting inventory material demand of power grid supply chain distribution network agreement based on time series model according to claim 1 is characterized by: The method for predicting inventory material demand of power grid supply chain distribution network agreement based on time series model adopts simulation "if-else" technology, establishes intelligent material allocation strategy library for different business scenarios, forms a "scenario-resource-strategy" operation model, and intelligently weights the prediction results of different prediction models through automatic optimization algorithm, so as to effectively improve the accuracy of demand prediction model.

10. The method for predicting inventory material demand of power supply chain distribution network agreement based on time series model according to claim 1 is characterized by: The method for predicting the demand for inventory materials in the power supply chain distribution network agreement based on the time series model introduces the time series model into the prediction of the demand for the power supply chain in terms of data modeling and algorithm engineering, broadens the traditional time series algorithm modeling, incorporates more efficient models that can automatically optimize parameters, and uses the combined application of the algorithm model library to improve the prediction accuracy, so as to improve the computing power and data applicability of the algorithm model library; The method for predicting the demand for distribution network agreement inventory materials in the power supply chain based on the time series model comprehensively considers the remaining share of the agreement inventory, the remaining in-transit inventory and the inventory in stock at the system operation level, forms a "one-stop" online operation mode for bidding and procurement plan submission and supply plan, connects the decision and instruction through "cloud" deployment, realizes online decision support, monitors the supply and demand of materials in real time, and improves the front-end and back-end coordination capabilities; The method for predicting the demand for inventory materials in the power grid supply chain distribution network agreement based on the time series model simulates the dynamics of materials to improve the level of resource optimization, uses the simulation "if-else" technology, combines the demand forecasting results, and establishes a material intelligent allocation strategy library for different business scenarios to form a "scenario-resource-strategy" operation model, adapt to the optimal material strategy, thereby realizing intelligent coordination of material allocation and further improving the material service level.

Images