Investment execution monitoring method and system adapting to multi-mode power grid project group characteristics

Abstract

An investment execution monitoring method and system adapted to multi-modal power grid project group characteristics are provided, and the method and the system aim at three aspects of specific operation and quality control in a project execution process, quantification of the performance of a unit in the execution process, and setting of a special monitoring index for a specific field or problem. A lean process management and control index, a unit execution monitoring index and a special item monitoring index are respectively constructed, the three indexes jointly form a project group index system, an intelligent monitoring model is constructed through a machine learning algorithm, and the intelligent monitoring model is used for carrying out intelligent monitoring on input power grid project group process basic data and the project group index system. Scientific decisions are made based on data analysis results through big data and artificial intelligence technologies. According to the design, scientific decision can be made through the data analysis result, the management efficiency is effectively improved, refined index monitoring and whole-process management of project group investment execution can be realized, and the project risk is effectively reduced.

Description

Technical Field

[0001] The present invention relates to an investment project monitoring method, and in particular to an investment execution monitoring method and system adapted to the characteristics of a multi-modal power grid project group. Background Art

[0002] Grid construction investment monitoring is a management activity that systematically tracks, analyzes, and evaluates key aspects of grid construction projects, including fund utilization, project progress, resource allocation, quality and safety, from planning to completion. Its core goal is to maximize investment returns while ensuring the compliance, safety, and sustainability of grid construction.

[0003] Power projects are characterized by large investment scale, long cycle and complex technology. In the current power industry investment project management process, the traditional investment monitoring method is for project managers to manually summarize data from various departments through Excel or paper reports, and generate macro indicators such as investment completion rate and production rate based on fixed cycles.

[0004] This method often relies on experience and static data, and is difficult to dynamically and real-timely reflect the investment execution status of project groups at different stages. It faces challenges in investment efficiency, risk management, resource allocation optimization, and other aspects.

[0005] As projects increase in scale and complexity, achieving refined management during the project investment process to ensure the scientific, effective, and efficient execution of a portfolio of investments has become a key issue in improving project management. To address these challenges, we are developing a comprehensive, multi-dimensional, and multi-layered monitoring indicator system to innovatively implement refined monitoring and analysis of the portfolio investment execution process, ensuring the achievement of project investment objectives.

[0006] Although this investment project monitoring method is feasible, it still has the following shortcomings:

[0007] 1. Manual detection methods often rely on experience and static data, making it difficult to dynamically and real-timely reflect the investment execution status of project groups at different stages, resulting in low management efficiency.

[0008] 2. The lack of dynamic real-time monitoring results in the inability to detect project risks in a timely manner, resulting in higher project risks.

[0009] The information disclosed in this background technology section is only intended to increase understanding of the overall background of the application and should not be considered as an admission or any form of suggestion that the information constitutes the prior art already known to ordinary technicians in this field. Summary of the Invention

[0010] The purpose of the present invention is to overcome the shortcomings of low management efficiency and high project risk in the prior art, and to provide an investment execution monitoring method and system that has high management efficiency and low project risk and is adapted to the characteristics of multimodal power grid project groups.

[0011] To achieve the above objectives, the technical solution of the present invention is:

[0012] An investment execution monitoring method adapted to the characteristics of a multimodal power grid project group, the monitoring method comprising:

[0013] S1. Data acquisition: Obtain basic data from the entire process of historical related project groups from early planning to settlement and final accounting as sample data. Use the RAG model to vectorize the basic project information, build an index, and store it in the database. Perform data cleaning and conversion to obtain project indicator construction data. Simultaneously, obtain real-time basic data of the project groups that need to be monitored daily, perform data cleaning and conversion, and obtain project operation monitoring data.

[0014] S2. Construct a project group indicator system. The project group indicators include lean process control indicators, unit execution monitoring indicators, and special monitoring indicators. The lean process control indicators are general indicators for monitoring project implementation. The unit execution monitoring indicators are indicators for evaluating the effectiveness of plan execution at the unit level. The special monitoring indicators are targeted monitoring indicators for monitoring special projects. The logical judgment statements in the project group indicators are converted into mathematical expressions or logical operation symbols to obtain indicator features that can be recognized by the model.

[0015] S3. Build a model: Use a machine learning algorithm to build an intelligent monitoring model. The input of the intelligent monitoring model is the basic data of the power grid project group process and various project group indicators. The output of the intelligent monitoring model is the project investment progress, project risks, and risk response decisions corresponding to the project risks.

[0016] S4. Project group monitoring: input the project operation monitoring data into the intelligent monitoring model to obtain the project investment progress, project risks and risk response decisions corresponding to the project risks, and interact with users through the RAG model.

[0017] The regulatory indicators include the completion index of construction commencement report, construction commencement compliance index, suspected untimely production report index, and production compliance index;

[0018] The progress indicators include suspected no progress indicators, suspected project delay indicators, and suspected multiple project closure indicators;

[0019] The fund management indicators include indicators suspected of inflating investment completion by service-related accounts, indicators suspected of inflating investment completion by material-related accounts, indicators suspected of delayed project settlement, indicators of reasonable deviations between guidance lines and financial expenditures, indicators suspected of unreasonable write-offs of large-scale service accounts, and indicators of mismatch between the timing of production and final settlement.

[0020] The material management indicators include indicators of suspected failure to collect materials in a timely manner after construction has commenced, indicators of suspected non-compliant collection and return of construction materials, indicators of suspected excess of estimated material demand for projects under construction, and indicators of failure to complete material transfer procedures in a timely manner.

[0021] The special monitoring indicators include the network source unlinked alarm indicator, the network source milestone plan mismatch alarm indicator, the file supplement untimely alarm indicator, the material collection and return procedures untimely alarm indicator, the construction results unachieved alarm indicator, the main equipment delivery untimely indicator, the suspected inability to be put into production on schedule indicator, the grid reinforcement and loss reduction effectiveness indicator, the grid reinforcement and load reduction effectiveness indicator, the milestone target completion delay indicator, the suspected inability to achieve the production target on schedule indicator, the peak summer and winter load reduction indicator, the unapproved start-up indicator, the construction progress mismatch alarm indicator, the procurement application timeliness indicator, the suspected railway and external power project construction progress mismatch indicator, and the suspected production sequence mismatch alarm indicator.

[0022] a. Risk assessment of regulatory indicators:

[0023] The conditions for determining the risk of the completion indicators of the commencement report are that:

[0024] ①Current time - earliest start report time>30 days;

[0025] ② The project has been reported to have started at the source end, but has not been reported to have started in the development professional investment statistics;

[0026] The risk of the commencement compliance indicator is determined by failure to meet one of the following conditions:

[0027] ① The project has uploaded the feasibility study approval document, and the feasibility study approval document number and time are not empty;

[0028] ② For projects of 35 kV and above, the approval documents have been uploaded, and the approval document number and approval time are not blank;

[0029] The conditions for determining the risk of the suspected untimely reporting of production indicators are that:

[0030] ① The (current time - latest time of completion report) of the project is greater than 30 days;

[0031] ② The project has been reported to be put into production at the source end, but has not been reported to be put into production in the development of professional investment statistics;

[0032] The risk determination condition for the production compliance indicator is: the amount of financial records ≤ total investment * 75%;

[0033] b. Risk assessment of progress indicators:

[0034] The conditions for determining the risk of suspected construction no progress indicators are that:

[0035] ① Construction has been underway for more than 6 months;

[0036] ②The cumulative construction progress is 0;

[0037] The conditions for determining the risk of the suspected project delay indicator are that:

[0038] ①The project has not been put into production;

[0039] ②The actual construction period (current time - actual start time) exceeds the following time limit;

[0040] The criteria for determining the risk of multiple closure indicators for suspected projects are: the project closure time has been adjusted three or more times since the project started;

[0041] c. Risk assessment of fund management indicators:

[0042] The judgment condition for the suspected service-related account to increase the investment completion index is to meet one of the following conditions:

[0043] ① Material budget / project budget>20%;

[0044] ② The cumulative service account / service budget estimate since the start is greater than 70%, and the cumulative material account / material budget estimate since the start is less than 20%;

[0045] The conditions for determining the risk of the suspected material-related entries driving up the investment completion indicator are: the cumulative material-related entries since the start / material-related budget > 30%, and the cumulative service-related entries since the start = 0;

[0046] The risk of the suspected untimely settlement indicator is determined to be a risk if one of the following conditions is met:

[0047] ① For 35 kV and 110 kV projects, the project settlement time is blank, and the difference between the current time and the completion time is between 30 and 60 days;

[0048] ② For 220-750 kV projects, the project settlement time is blank, and the difference between the current time and the completion time is between 70-100 days;

[0049] The risk of deviation between the guideline and the reasonable indicator of financial expenditure is determined when one of the following conditions is not met:

[0050] ① The physical quantity guideline of the project can be calculated;

[0051] ② The physical quantity guidance line of the project is available;

[0052] ③|Guideline value - financial expenditure|>0, and |Guideline value - financial expenditure|≤10%*total amount of approved budget estimate;

[0053] The criteria for determining the risk of unreasonable indicators of suspected large-value service write-offs are:

[0054] When used in projects of 35 kV and above, one of the following conditions must be met:

[0055] ① This month's service costs are less than -1 million;

[0056] ②|This month's service cost recorded / last month's cumulative service cost recorded since the beginning of the month|>10%;

[0057] When used in projects of 10 kV and below, one of the following conditions must be met:

[0058] ① This month's service cost is less than -100,000;

[0059] ②|This month's service cost recorded / last month's cumulative service cost recorded since the beginning of the month|>10%;

[0060] The risk of mismatch between the timing of production start-up and final accounting is determined when one of the following conditions is met:

[0061] ① The commissioning time of the online power grid scale monthly report is blank, and one of the project shutdown times exists;

[0062] ② The commissioning time of the online power grid scale monthly report is not empty and is later than one of the project shutdown times;

[0063] ③|Latest commissioning time of a single item of the PMS system - commissioning time of the completion acceptance report|>7 days;

[0064] d. Risk assessment of material management indicators:

[0065] The criteria for determining the risk of the indicator of suspected commencement of work but failure to timely receive materials are that:

[0066] ①The project has actually started;

[0067] ②Current time - project start time>=3 months;

[0068] ③The actual amount of materials received is 0;

[0069] The risk of suspected non-compliant indicators for the return of construction materials is determined if one of the following conditions is met:

[0070] ① The planned total investment is greater than RMB 500,000, the write-off amount / total project investment is greater than 20%, and the amount of returned materials / total project investment is greater than 20%;

[0071] ② The cumulative number of returns of the same material in a project is greater than the cumulative number of receipts for the same material*50%; and the amount of returns is greater than RMB 50,000, and the same material in the same project and the same receipt voucher are used for returns three or more times;

[0072] The conditions for determining the risk of exceeding the estimated material demand indicators for the suspected projects under construction are that:

[0073] ①The project has not actually been put into production;

[0074] ②Total investment>500,000 yuan;

[0075] ③ The total net amount of materials used in the project is greater than 90% of the total investment;

[0076] The criterion for determining the risk of failure to handle material transfer procedures in a timely manner is that the physical ID of the distribution transformer equipment associated with the commissioned project is inconsistent with the physical ID of the distribution transformer used for the project.

[0077] Risk assessment of the special monitoring indicators:

[0078] The conditions for determining the risk of the network source unassociated alarm indicator are: the power supply project is not associated with the power source;

[0079] The judgment condition for the network source milestone plan mismatch alarm indicator is: |Power transmission project planned commissioning time - power supply project's earliest planned commissioning time|> 6 months;

[0080] The judgment condition for the alarm indicator of untimely file supplementation is to meet one of the following conditions:

[0081] ① Associate user profiles;

[0082] ②The scale of related equipment files is not equal to the scale of project commissioning;

[0083] The criteria for determining the alarm indicator for untimely handling of materials return procedures is to meet one of the following conditions:

[0084] ①Current time - commissioning time>3 months;

[0085] ② The substation capacity used is not equal to the associated distribution transformer capacity;

[0086] The judgment condition for the construction effect not reaching the warning indicator is that one of the following conditions is met:

[0087] ① The project has associated operational equipment (distribution transformer, main transformer, and line);

[0088] ② Equipment utilization rate in the month following equipment operation = 0;

[0089] The risk of the main equipment delivery delay indicator is determined by the following conditions:

[0090] ① The M kV project has been under construction for more than N months;

[0091]

[0092] ② The main equipment arrival rate of the mkV project <n;

[0093]

[0094] The conditions for determining the risk of the suspected failure to put into production on schedule are that:

[0095] ①The project has been under construction for more than 6 months;

[0096] ② The construction progress of the X kV project is less than Y;

[0097]

[0098] ③ The progress of the x kV project in receiving funds is less than y;

[0099]

[0100] The judgment condition for the occurrence of risk of the effective index of grid reinforcement and loss reduction is: after the grid optimization and reinforcement project in the target area is put into operation, the line loss rate decreases by more than 0.2% compared with the level before the commissioning;

[0101] The criterion for determining the risk of the grid reinforcement and load reduction effectiveness index is that one of the following conditions is not met:

[0102] ① After the grid optimization and reinforcement project in the target area was put into operation, the number of overloaded units decreased by 30% compared with before the commissioning, and the number of heavy-loaded units decreased by 20%;

[0103] ② After the grid optimization and reinforcement project in the target area is put into operation, (the cumulative duration of heavy overload before the commissioning - the cumulative duration of heavy overload after the commissioning) / the cumulative duration of heavy overload before the commissioning > 20%;

[0104] The conditions for determining the risk of lagging behind in the milestone target completion indicator are that:

[0105] ①The project has started;

[0106] ② A single milestone is not completed for more than one month after the planned completion time;

[0107] The conditions for determining the risk of the suspected production target not being achieved on schedule are:

[0108] For the peak summer project, the following requirements must be met at the same time:

[0109] ①The project has not been put into production;

[0110] ②(Planned production time - current time) < 2 months;

[0111] ③Cumulative construction progress <60%;

[0112] ④Cumulative deposit progress <50%;

[0113] For the peak winter project, it also meets the following requirements:

[0114] ①The project has not been put into production;

[0115] ②(Planned production time - current time) < 2 months;

[0116] ③Cumulative construction progress <50%;

[0117] ④Cumulative deposit progress <40%;

[0118] The risk of the load shedding indicator during the peak summer and winter seasons is determined if one of the following conditions is not met:

[0119] ①Compared with the same period last year, the current data in the target area shows that the number of overloaded units has decreased by 40% and the number of heavy-loaded units has decreased by 30% since the start of production;

[0120] ②Compared with the current data of the target area in the same period of previous years, (the absolute value of the accumulated heavy overload time before production - the accumulated heavy overload time after production) / the accumulated heavy overload time before production > 20%;

[0121] The risk of starting construction without approval is determined by satisfying one of the following conditions:

[0122] ① The business expansion supporting project has started without user approval;

[0123] ②35kV business expansion supporting project start time - user approval time> 9 months;

[0124] ③110kV business expansion supporting project start time - user approval time> 16 months;

[0125] The risk of the construction progress mismatch warning indicator is determined by satisfying one of the following conditions:

[0126] ①10kV business expansion supporting project commissioning time - user demand power time> 6 months

[0127] ②35kV business expansion supporting project commissioning time - user demand power consumption time> 7 months;

[0128] ③ The commissioning time of 110kV business expansion supporting project minus the power demand time of users>6 months;

[0129] The risk of the procurement application timeliness indicator is determined to be at risk if one of the following conditions is met:

[0130] ① The first material procurement application time for the 35kV business expansion supporting project minus the user approval time is > 6 months;

[0131] ② The time from the first material procurement application to the user approval for the 110kV business expansion supporting project is greater than 10 months;

[0132] The conditions for determining the risk of mismatch between the construction progress of the railway and the external power project are that:

[0133] ①The project has not been put into production;

[0134] ② Traction station construction progress - external power construction progress > 20%

[0135] ③ The amount of materials received for external power projects / estimated equipment purchases is less than 30%;

[0136] The risk of the suspected production timing mismatch alarm indicator is determined to be a risk if one of the following conditions is met:

[0137] ① Wind power transmission: 35kV grid connection project start time - power source start time > 7 months (power source construction period - grid connection project construction period); 110kV grid connection project start time - power source start time > 3 months; 220kV grid connection project start time - power source start time > 12 months;

[0138] ② Photovoltaic power transmission: 35kV grid connection project start time - power source start time > 1 month, 110kV grid connection project start time - power source start time > 5 months, 220kV grid connection project start time - power source start time > 2 months.

[0139] In S3, the machine learning algorithm used is the random forest algorithm, which obtains multiple classification results by sampling the input data, and selects the classification result with the largest number of occurrences among all the classification results as the final result of the random forest algorithm;

[0140] Input the project indicator construction data and indicator features into the random forest algorithm to determine the number of decision trees n in the random forest. Initially, each decision tree in the random forest is in an untrained state;

[0141] 70% of the project indicator construction data is divided into a training set and 30% is divided into a test set. For each decision tree, samples are randomly selected from the training set with replacement to construct a sub-training set. The original training set includes m samples, and each tree sub-training set includes m samples.

[0142] When constructing the node splitting of each tree, some features are randomly selected from all features, and the said partial features are k features, k is less than the total number of features, and the best splitting features are selected according to the Gini index splitting criterion;

[0143] Repeat the above node splitting process until the stopping condition or the maximum depth limit of the tree is reached;

[0144] Repeat the above training process for n decision trees in the random forest, and draw a conclusion after the training is completed. The conclusion is the category prediction in the classification task or the numerical prediction in the regression task;

[0145] The trained random forest model is tested using a test set. If the accuracy of the trained random forest model is less than the set value, the sample data sampling range is expanded, and the data is re-acquired and the random forest model is trained. If the accuracy of the trained random forest model is greater than the set value, the trained random forest model is used as an intelligent monitoring model.

[0146] Said S3 also includes: said random forest model including the weight of each indicator in the project group indicator system, wherein the higher the weight of each indicator is, the higher the importance or risk of the indicator is, and the sum of the weights of each indicator is 1, and the total risk score of the project is calculated by weighted average method;

[0147] Setting a project risk monitoring and scoring system, wherein the project risk monitoring and scoring system includes assigning weights to project risks output by the intelligent monitoring model, setting a total project risk score for each project, and setting risk thresholds for power grid project groups, wherein the risk thresholds include high risk thresholds, medium risk thresholds, and low risk thresholds;

[0148] The risk score is obtained by combining the various project risks and corresponding weights output by the intelligent monitoring model. During the evaluation, the project risk score is obtained by subtracting the risk score from the total project risk score. The project risk score and the risk threshold are compared. When the risk monitoring score is lower than the low risk threshold, it means that the project operation is stable and continues to proceed according to the original plan. When the risk monitoring score is between the low risk threshold and the medium risk threshold, it means that the project risk is low and potential risks need to be paid attention to, but no major measures need to be taken immediately. When the risk monitoring score is between the medium risk threshold and the high risk threshold, it means that the project risk is large and immediate countermeasures need to be taken. When the risk monitoring score is higher than the high risk threshold, it means that the project risk is extremely high and the project needs to be suspended and the risk response decision corresponding to the project risk needs to be output.

[0149] In S1, the private domain data is vectorized using the RAG model, and then an index is constructed and stored in the database, including:

[0150] Process semi-structured data in various formats, convert them into the same processing paradigm, obtain data from different data sources, and perform standardization;

[0151] Filter, compress and format data to remove noise and redundant information and obtain metadata;

[0152] According to the token restrictions of the embedding model, the text is segmented into appropriate granularity, such as sentence segmentation or fixed-length segmentation;

[0153] Convert text into vector representation through embedding model;

[0154] Build an index for the vectorized data and write it to the database. Select Elasticsearch as the database for the business scenario.

[0155] In S4, the RAG model performs data retrieval by combining similarity retrieval with full-text retrieval. The similarity retrieval returns relevant records by calculating the similarity score between the query vector and the stored vector, and the full-text retrieval constructs an inverted index using keywords.

[0156] Integrate the retrieved relevant knowledge with the user's question to form a prompt, and then generate the answer through a large language model.

[0157] An investment execution monitoring system adapted to the characteristics of a multimodal power grid project group, the system is used to execute the aforementioned investment execution monitoring method adapted to the characteristics of a multimodal power grid project group, specifically comprising: a data acquisition module, an indicator system construction module, a model construction module, and a monitoring module;

[0158] The data acquisition module is used to obtain basic data from the entire process of historical related project groups from the early planning stage to the final settlement as sample data, vectorize the basic project information through the RAG model, construct an index and store it in the database, and perform data cleaning and conversion to obtain project indicator construction data. At the same time, it obtains real-time basic data of the project groups that need to be monitored on a daily basis, and performs data cleaning and conversion to obtain project operation monitoring data.

[0159] The indicator system construction module is used to: construct a project group indicator system, wherein the project group indicators include lean process control indicators, unit execution monitoring indicators, and special monitoring indicators. The lean process control indicators are general indicators for monitoring the implementation of projects, the unit execution monitoring indicators are indicators for evaluating the effectiveness of plan execution at the monitoring unit level, and the special monitoring indicators are targeted monitoring indicators for monitoring special projects. The logical judgment statements in the project group indicators are converted into mathematical expressions or logical operation symbols to obtain indicator features that can be recognized by the model.

[0160] The model building module is used to: build a model, and build an intelligent monitoring model through machine learning algorithms. The input of the intelligent monitoring model is the basic data of the power grid project group process and various project group indicators. The output of the intelligent monitoring model is the project investment progress, project risks and risk response decisions corresponding to the project risks;

[0161] The monitoring module is used to input project operation monitoring data into the intelligent monitoring model, obtain project investment progress, project risks and risk response decisions corresponding to project risks, and interact with users through the RAG model.

[0162] An investment execution monitoring device adapted to the characteristics of a multimodal power grid project group includes a memory and a processor, wherein the memory is used to store computer program code and transmit the computer program code to the processor;

[0163] The processor is configured to execute the aforementioned investment execution monitoring method adapted to the characteristics of a multimodal power grid project group according to the instructions in the computer program code.

[0164] A computer program product includes a computer program. The computer program is used by a processor to execute the aforementioned investment execution monitoring method adapted to the characteristics of a multi-modal power grid project group.

[0165] Compared with the prior art, the present invention has the following beneficial effects:

[0166] 1. In an investment execution monitoring method adapted to the characteristics of multimodal power grid project groups, the present invention constructs an intelligent monitoring model through a machine learning algorithm. The intelligent monitoring model uses input power grid project group process basic data and project group indicator system, and uses big data and artificial intelligence technology to make scientific decisions based on data analysis results. The model obtains project investment progress, project risks, and corresponding risk response decisions, effectively improving management level. Therefore, this design can make scientific decisions based on data analysis results and effectively improve management efficiency.

[0167] 2. The present invention's investment execution monitoring method, adapted to the characteristics of multimodal power grid project clusters, focuses on three aspects: specific operations and quality control during project execution; quantifying the performance of units during execution; and setting specialized monitoring indicators for specific areas or issues. These three types of indicators, collectively forming a project cluster indicator system, enable refined indicator monitoring and full-process management of project cluster investment execution. Therefore, this design enables refined indicator monitoring and full-process management of project cluster investment execution, effectively reducing project risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0168] Figure 1is a flow chart of the method of the present invention.

[0169] Figure 2 It is a schematic diagram of the model of the present invention.

[0170] Figure 3 It is a structural diagram of the system of the present invention.

[0171] Figure 4 This is a structural diagram of the device described in Example 4. DETAILED DESCRIPTION

[0172] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0173] Example 1:

[0174] See also Figure 1 , an investment execution monitoring method adapted to the characteristics of a multimodal power grid project group, the monitoring method comprising:

[0175] S1. Data acquisition: Obtain basic data from the entire process of historical related project groups from early planning to settlement and final accounting as sample data. Use the RAG model to vectorize the basic project information, build an index, and store it in the database. Perform data cleaning and conversion to obtain project indicator construction data. Simultaneously, obtain real-time basic data of the project groups that need to be monitored daily, perform data cleaning and conversion, and obtain project operation monitoring data.

[0176] S2. Construct a project group indicator system. The project group indicators include lean process control indicators, unit execution monitoring indicators, and special monitoring indicators. The lean process control indicators are general indicators for monitoring project implementation. The unit execution monitoring indicators are indicators for evaluating the effectiveness of plan execution at the unit level. The special monitoring indicators are targeted monitoring indicators for monitoring special projects. The logical judgment statements in the project group indicators are converted into mathematical expressions or logical operation symbols to obtain indicator features that can be recognized by the model.

[0177] S3. Build a model: Use a machine learning algorithm to build an intelligent monitoring model. The input of the intelligent monitoring model is the basic data of the power grid project group process and various project group indicators. The output of the intelligent monitoring model is the project investment progress, project risks, and risk response decisions corresponding to the project risks.

[0178] S4. Project group monitoring: input the project operation monitoring data into the intelligent monitoring model to obtain the project investment progress, project risks and risk response decisions corresponding to the project risks, and interact with users through the RAG model.

[0179] In S1, the basic data of the whole process includes structured data, semi-structured data and unstructured data. The structured data is the real-time data of the feasibility study approval time, construction start time, production start time, investment completion, and account amount. The semi-structured data is the historical data of the feasibility study approval time, construction start time, production start time, investment completion, and account amount. The unstructured data is data other than structured data and semi-structured data. The structured data is obtained through a relational database, the semi-structured data is obtained through Excel spreadsheets and JSON files, and the unstructured data is obtained through documents, reports and social media using natural language processing technology;

[0180] The data cleaning and conversion includes: when there are missing values ​​in the numerical data of the sample data, if the missing values ​​are more than the set value, the sample data will be deleted; if the missing values ​​are less than the set value and the data distribution is approximately normal, the mean will be used to fill in the missing values; when there is skewness in the numerical data, the median will be used to fill in the missing values; when there are missing values ​​and outliers in the categorical data, the mode will be used to fill in the missing values;

[0181] After all missing values ​​and outliers in the sample data are filled, the sample data is standardized or normalized;

[0182] The process of missing value processing is as follows:

[0183] 1. Data integrity check: Identify missing key and non-key fields;

[0184] 2. Data rationality check: verify whether the missing values ​​are consistent with business logic;

[0185] 3. Data cleaning and deduplication: Fill missing values ​​through similar data filling and semantic analysis;

[0186] 4. Multiple data source supplementation: Obtain supplementary data from other data sources;

[0187] 5. Terminology processing: infer missing terminology based on the term dictionary and context;

[0188] 6. Data verification: Verify the filled data to ensure its integrity and rationality;

[0189] When there are many missing values, the system handles the missing values ​​in the following ways:

[0190] 1. Mandatory completion: Key fields must be completed;

[0191] 2. Default value filling: Use default values ​​to fill non-key fields;

[0192] 3. Logical inference: infer missing values ​​based on business logic;

[0193] 4. Similar data filling: fill by finding data of similar items;

[0194] 5. Multiple data source supplementation: obtain supplementary data from other data sources;

[0195] The lean process control indicators include regulatory indicators, progress indicators, capital management indicators, material management indicators, and data management indicators;

[0196] The regulatory indicators include the completion index of construction commencement report, construction commencement compliance index, suspected untimely production report index, and production compliance index;

[0197] The progress indicators include suspected no progress indicators, suspected project delay indicators, and suspected multiple project closure indicators;

[0198] The fund management indicators include indicators suspected of inflating investment completion by service-related accounts, indicators suspected of inflating investment completion by material-related accounts, indicators suspected of delayed project settlement, indicators of reasonable deviations between guidance lines and financial expenditures, indicators suspected of unreasonable write-offs of large-scale service accounts, and indicators of mismatch between the timing of production and final settlement.

[0199] The material management indicators include indicators of suspected failure to collect materials in a timely manner after construction has commenced, indicators of suspected non-compliant collection and return of construction materials, indicators of suspected excess of estimated material demand for projects under construction, and indicators of failure to complete material transfer procedures in a timely manner.

[0200] The special monitoring indicators include the network source unlinked alarm indicator, the network source milestone plan mismatch alarm indicator, the file supplement untimely alarm indicator, the material collection and return procedures untimely alarm indicator, the construction results unachieved alarm indicator, the main equipment delivery untimely indicator, the suspected inability to be put into production on schedule indicator, the grid reinforcement and loss reduction effectiveness indicator, the grid reinforcement and load reduction effectiveness indicator, the milestone target completion delay indicator, the suspected inability to achieve the production target on schedule indicator, the peak summer and winter load reduction indicator, the unapproved start-up indicator, the construction progress mismatch alarm indicator, the procurement application timeliness indicator, the suspected railway and external power project construction progress mismatch indicator, and the suspected production sequence mismatch alarm indicator.

[0201] See also Figure 2 In S3, the machine learning algorithm used is the random forest algorithm, which obtains multiple classification results by sampling the input data, and selects the classification result with the largest number of occurrences among all the classification results as the final result of the random forest;

[0202] The improvements to the random forest algorithm in this embodiment include:

[0203] Feature selection optimization: The feature selection process has been optimized based on the characteristics of power grid infrastructure project data. For example, fields such as project name, voltage level, and WBS code are given higher weights because these fields have a significant impact on project risk analysis.

[0204] Data preprocessing: Special preprocessing was performed on the power grid infrastructure project data, including missing value filling and data standardization. For example, missing voltage level fields were filled based on historical data, and numeric fields (such as substation capacity) were standardized.

[0205] Objective function adjustment: Based on project requirements, the objective function was adjusted to more accurately reflect project risks. For example, the objective function not only considers the prediction error but also the severity of the risk event.

[0206] Input the project indicator construction data and indicator features into the random forest algorithm to determine the number of decision trees n in the random forest. Initially, each decision tree in the random forest is in an untrained state;

[0207] 70% of the project indicator construction data is divided into a training set and 30% is divided into a test set. For each decision tree, samples are randomly selected from the training set with replacement to construct a sub-training set. The original training set includes m samples, and each tree sub-training set includes m samples.

[0208] When constructing the node splitting of each tree, some features are randomly selected from all features, and the said partial features are k features, k is less than the total number of features, and the best splitting features are selected according to the Gini index splitting criterion;

[0209] Repeat the above node splitting process until the stopping condition or the maximum depth limit of the tree is reached.

[0210] Repeat the above training process for n decision trees in the random forest. After the training is completed, a conclusion is drawn, which is a category prediction in classification tasks or a numerical prediction in regression tasks.

[0211] The trained random forest model is tested using a test set. If the accuracy of the trained random forest model is less than the set value, the sample data sampling range is expanded, and the data is re-acquired and the random forest model is trained. If the accuracy of the trained random forest model is greater than the set value, the trained random forest model is used as an intelligent monitoring model.

[0212] Said S3 also includes: said random forest model including the weight of each indicator in the project group indicator system, wherein the higher the weight of each indicator is, the higher the importance or risk of the indicator is, and the sum of the weights of each indicator is 1, and the total risk score of the project is calculated by weighted average method;

[0213] Setting a project risk monitoring and scoring system, wherein the project risk monitoring and scoring system includes assigning weights to project risks output by the intelligent monitoring model, setting a total project risk score for each project, and setting risk thresholds for power grid project groups, wherein the risk thresholds include high risk thresholds, medium risk thresholds, and low risk thresholds;

[0214] The risk score is obtained by combining the various project risks and corresponding weights output by the intelligent monitoring model. During the evaluation, the project risk score is obtained by subtracting the risk score from the total project risk score. The project risk score and the risk threshold are compared. When the risk monitoring score is lower than the low risk threshold, it means that the project operation is stable and continues to proceed according to the original plan. When the risk monitoring score is between the low risk threshold and the medium risk threshold, it means that the project risk is low and potential risks need to be paid attention to, but no major measures need to be taken immediately. When the risk monitoring score is between the medium risk threshold and the high risk threshold, it means that the project risk is large and immediate countermeasures need to be taken. When the risk monitoring score is higher than the high risk threshold, it means that the project risk is extremely high and the project needs to be suspended and the risk response decision corresponding to the project risk needs to be output.

[0215] In S1, the private domain data is vectorized using the RAG model, and then an index is constructed and stored in the database, including:

[0216] Process semi-structured data in various formats, convert them into the same processing paradigm, obtain data from different data sources, and perform standardization;

[0217] Filter, compress and format data to remove noise and redundant information and obtain metadata;

[0218] According to the token restrictions of the embedding model, the text is segmented into appropriate granularity, such as sentence segmentation or fixed-length segmentation;

[0219] Convert text into vector representation through embedding model;

[0220] Build an index for the vectorized data and write it to the database. Select Elasticsearch as the database for the business scenario.

[0221] In S4, the RAG model performs data retrieval by combining similarity retrieval with full-text retrieval. The similarity retrieval returns relevant records by calculating the similarity score between the query vector and the stored vector, and the full-text retrieval constructs an inverted index using keywords.

[0222] Integrate the retrieved relevant knowledge with the user's question to form a prompt, and then generate the answer through a large language model.

[0223] Example 2:

[0224] a. Risk assessment of regulatory indicators:

[0225] The conditions for determining the risk of the completion indicators of the commencement report are that:

[0226] ①Current time - earliest start report time>30 days;

[0227] ② The project has been reported to have started at the source end, but has not been reported to have started in the development professional investment statistics;

[0228] The risk of the commencement compliance indicator is determined by failure to meet one of the following conditions:

[0229] ① The project has uploaded the feasibility study approval document, and the feasibility study approval document number and time are not empty;

[0230] ② For projects of 35 kV and above, the approval documents have been uploaded, and the approval document number and approval time are not blank;

[0231] The conditions for determining the risk of the suspected untimely reporting of production indicators are that:

[0232] ① The (current time - latest time of completion report) of the project is greater than 30 days;

[0233] ② The project has been reported to be put into production at the source end, but has not been reported to be put into production in the development of professional investment statistics;

[0234] The risk determination condition for the production compliance indicator is: the amount of financial records ≤ total investment * 75%;

[0235] b. Risk assessment of progress indicators:

[0236] The conditions for determining the risk of suspected construction no progress indicators are that:

[0237] ① Construction has been underway for more than 6 months;

[0238] ②The cumulative construction progress is 0;

[0239] The conditions for determining the risk of the suspected project delay indicator are that:

[0240] ①The project has not been put into production;

[0241] ②The actual construction period (current time - actual start time) exceeds the following time limit;

[0242] The criteria for determining the risk of multiple closure indicators for suspected projects are: the project closure time has been adjusted three or more times since the project started;

[0243] c. Risk assessment of fund management indicators:

[0244] The judgment condition for the suspected service-related account to increase the investment completion index is to meet one of the following conditions:

[0245] ① Material budget / project budget>20%;

[0246] ② The cumulative service account / service budget estimate since the start is greater than 70%, and the cumulative material account / material budget estimate since the start is less than 20%;

[0247] The conditions for determining the risk of the suspected material-related entries driving up the investment completion indicator are: the cumulative material-related entries since the start / material-related budget > 30%, and the cumulative service-related entries since the start = 0;

[0248] The risk of the suspected untimely settlement indicator is determined to be a risk if one of the following conditions is met:

[0249] ① For 35 kV and 110 kV projects, the project settlement time is blank, and the difference between the current time and the completion time is between 30 and 60 days;

[0250] ② For 220-750 kV projects, the project settlement time is blank, and the difference between the current time and the completion time is between 70-100 days;

[0251] The risk of deviation between the guideline and the reasonable indicator of financial expenditure is determined when one of the following conditions is not met:

[0252] ① The physical quantity guideline of the project can be calculated;

[0253] ② The physical quantity guidance line of the project is available;

[0254] ③|Guideline value - financial expenditure|>0, and |Guideline value - financial expenditure|≤10%*total amount of approved budget estimate;

[0255] The criteria for determining the risk of unreasonable indicators of suspected large-amount service write-offs are:

[0256] When used in projects of 35 kV and above, one of the following conditions must be met:

[0257] ① This month's service costs are less than -1 million;

[0258] ②|This month's service cost recorded / last month's cumulative service cost recorded since the beginning of the month|>10%;

[0259] When used in projects of 10 kV and below, one of the following conditions must be met:

[0260] ① This month's service cost is less than -100,000;

[0261] ②|This month's service cost recorded / last month's cumulative service cost recorded since the beginning of the month|>10%;

[0262] The risk of mismatch between the timing of production start-up and final accounting is determined when one of the following conditions is met:

[0263] ① The commissioning time of the online power grid scale monthly report is blank, and one of the project shutdown times exists;

[0264] ② The commissioning time of the online power grid scale monthly report is not empty and is later than one of the project shutdown times;

[0265] ③|Latest commissioning time of a single item of the PMS system - commissioning time of the completion acceptance report|>7 days;

[0266] d. Risk assessment of material management indicators:

[0267] The criteria for determining the risk of the indicator of suspected commencement of work but failure to timely receive materials are that:

[0268] ①The project has actually started;

[0269] ②Current time - project start time>=3 months;

[0270] ③The actual amount of materials received is 0;

[0271] The risk of suspected non-compliant indicators for the return of construction materials is determined if one of the following conditions is met:

[0272] ① The planned total investment is greater than RMB 500,000, the write-off amount / total project investment is greater than 20%, and the amount of returned materials / total project investment is greater than 20%;

[0273] ② The cumulative number of returns of the same material in a project is greater than the cumulative number of receipts for the same material*50%; and the amount of returns is greater than RMB 50,000, and the same material in the same project and the same receipt voucher are used for returns three or more times;

[0274] The conditions for determining the risk of exceeding the estimated material demand indicators for the suspected projects under construction are that:

[0275] ①The project has not actually been put into production;

[0276] ②Total investment>500,000 yuan;

[0277] ③ The total net amount of materials used in the project is greater than 90% of the total investment;

[0278] The criterion for determining the risk of failure to handle material transfer procedures in a timely manner is that the physical ID of the distribution transformer equipment associated with the commissioned project is inconsistent with the physical ID of the distribution transformer used for the project.

[0279] Risk assessment of the special monitoring indicators:

[0280] The conditions for determining the risk of the network source unassociated alarm indicator are: the power supply project is not associated with the power source;

[0281] The judgment condition for the network source milestone plan mismatch alarm indicator is: |Power transmission project planned commissioning time - power supply project's earliest planned commissioning time|> 6 months;

[0282] The judgment condition for the alarm indicator of untimely file supplementation is to meet one of the following conditions:

[0283] ① Associate user profiles;

[0284] ②The scale of related equipment files is not equal to the scale of project commissioning;

[0285] The criteria for determining the alarm indicator for untimely handling of materials return procedures is to meet one of the following conditions:

[0286] ①Current time - commissioning time>3 months;

[0287] ② The substation capacity used is not equal to the associated distribution transformer capacity;

[0288] The judgment condition for the construction effect not reaching the warning indicator is that one of the following conditions is met:

[0289] ① The project has associated operational equipment (distribution transformer, main transformer, and line);

[0290] ② Equipment utilization rate in the month following equipment commissioning = 0;

[0291] The risk of the main equipment delivery delay indicator is determined by the following conditions:

[0292] ① The M kV project has been under construction for more than N months;

[0293]

[0294] ② The main equipment arrival rate of the mkV project <n;

[0295]

[0296] The conditions for determining the risk of the suspected failure to put into production on schedule are that:

[0297] ①The project has been under construction for more than 6 months;

[0298] ② The construction progress of the X kV project is less than Y;

[0299]

[0300] ③ The progress of the x kV project in receiving funds is less than y;

[0301]

[0302] The judgment condition for the occurrence of risk of the effective index of grid reinforcement and loss reduction is: after the grid optimization and reinforcement project in the target area is put into operation, the line loss rate decreases by more than 0.2% compared with the level before the commissioning;

[0303] The criterion for determining the risk of the grid reinforcement and load reduction effectiveness index is that one of the following conditions is not met:

[0304] ① After the grid optimization and reinforcement project in the target area was put into operation, the number of overloaded units decreased by 30% compared with before the commissioning, and the number of heavy-loaded units decreased by 20%;

[0305] ② After the grid optimization and reinforcement project in the target area is put into operation, (the cumulative duration of heavy overload before the commissioning - the cumulative duration of heavy overload after the commissioning) / the cumulative duration of heavy overload before the commissioning > 20%;

[0306] The conditions for determining the risk of lagging behind in the milestone target completion indicator are that:

[0307] ①The project has started;

[0308] ② A single milestone is not completed for more than one month after the planned completion time;

[0309] The conditions for determining the risk of the suspected production target not being achieved on schedule are:

[0310] For the peak summer project, the following requirements must be met at the same time:

[0311] ①The project has not been put into production;

[0312] ②(Planned production time - current time) < 2 months;

[0313] ③Cumulative construction progress <60%;

[0314] ④Cumulative deposit progress <50%;

[0315] For the peak winter project, it also meets the following requirements:

[0316] ①The project has not been put into production;

[0317] ②(Planned production time - current time) < 2 months;

[0318] ③Cumulative construction progress <50%;

[0319] ④Cumulative deposit progress <40%;

[0320] The risk of the load shedding indicator during the peak summer and winter seasons is determined if one of the following conditions is not met:

[0321] ①Compared with the same period last year, the current data in the target area shows that the number of overloaded units has decreased by 40% and the number of heavy-loaded units has decreased by 30% since the start of production;

[0322] ②Compared with the current data of the target area in the same period of previous years, (the absolute value of the accumulated heavy overload time before production - the accumulated heavy overload time after production) / the accumulated heavy overload time before production > 20%;

[0323] The judgment condition for the risk of the unapproved start-up indicator is that one of the following conditions is met:

[0324] ① The business expansion supporting project has started without user approval;

[0325] ②35kV business expansion supporting project start time - user approval time> 9 months;

[0326] ③110kV business expansion supporting project start time - user approval time> 16 months;

[0327] The risk of the construction progress mismatch warning indicator is determined by satisfying one of the following conditions:

[0328] ①10kV business expansion supporting project commissioning time - user demand power time> 6 months

[0329] ②35kV business expansion supporting project commissioning time - user demand power consumption time> 7 months;

[0330] ③ The commissioning time of 110kV business expansion supporting project minus the power demand time of users>6 months;

[0331] The risk of the procurement application timeliness indicator is determined to be at risk if one of the following conditions is met:

[0332] ① The time from the first material purchase application to the user approval for the 35kV business expansion supporting project is greater than 6 months;

[0333] ② The time from the first material procurement application to the user approval for the 110kV business expansion supporting project is greater than 10 months;

[0334] The conditions for determining the risk of mismatch between the construction progress of the railway and the external power project are that:

[0335] ①The project has not been put into production;

[0336] ② Traction station construction progress - external power construction progress > 20%

[0337] ③ The amount of materials received for external power projects / estimated equipment purchases is less than 30%;

[0338] The risk of the suspected production timing mismatch alarm indicator is determined to be a risk if one of the following conditions is met:

[0339] ① Wind power transmission: 35kV grid connection project start time - power source start time > 7 months (power source construction period - grid connection project construction period); 110kV grid connection project start time - power source start time > 3 months; 220kV grid connection project start time - power source start time > 12 months;

[0340] ② Photovoltaic power transmission: 35kV grid connection project start time - power source start time > 1 month, 110kV grid connection project start time - power source start time > 5 months, 220kV grid connection project start time - power source start time > 2 months.

[0341] In S3, the machine learning algorithm used is the random forest algorithm, which obtains multiple classification results by sampling the input data, and selects the classification result with the largest number of occurrences among all the classification results as the final result of the random forest algorithm;

[0342] Input the project indicator construction data and indicator features into the random forest algorithm to determine the number of decision trees n in the random forest. Initially, each decision tree in the random forest is in an untrained state;

[0343] 70% of the project indicator construction data is divided into a training set and 30% is divided into a test set. For each decision tree, samples are randomly selected from the training set with replacement to construct a sub-training set. The original training set includes m samples, and each tree sub-training set includes m samples.

[0344] When constructing the node splitting of each tree, some features are randomly selected from all features, and the said partial features are k features, k is less than the total number of features, and the best splitting features are selected according to the Gini index splitting criterion;

[0345] Repeat the above node splitting process until the stopping condition or the maximum depth limit of the tree is reached;

[0346] Repeat the above training process for n decision trees in the random forest, and draw a conclusion after the training is completed. The conclusion is the category prediction in the classification task or the numerical prediction in the regression task;

[0347] The trained random forest model is tested using a test set. If the accuracy of the trained random forest model is less than the set value, the sample data sampling range is expanded, and the data is re-acquired and the random forest model is trained. If the accuracy of the trained random forest model is greater than the set value, the trained random forest model is used as an intelligent monitoring model.

[0348] Example 3:

[0349] The overall evaluation index rules for all projects in the project group include:

[0350] The completion indicators for commencement of construction include:

[0351] Suspected late start-up report = number of projects suspected of late start-up report / number of planned projects * 100%;

[0352] The commencement compliance indicators include:

[0353] Compliance of commencement of work = document completeness * 30% + compliance of commencement of work * 70%;

[0354] Document completeness = number of projects with complete documents / total number of projects under construction * 100%;

[0355] The suspected indicators of untimely reporting of commissioning include:

[0356] Suspected late reporting of commissioning = number of suspected late reporting of commissioning projects / number of monitoring projects * 100%;

[0357] The production compliance indicators include:

[0358] Compliance of commissioning = number of compliant projects commissioned / number of projects commissioned this year * 100%:

[0359] The suspected indicators of no progress in construction include:

[0360] Alarm rate of suspected construction progress = alarm rate of suspected construction start with no progress * 50% + alarm rate of suspected construction progress with no progress for three consecutive months * 50%;

[0361] Alarm rate of suspected projects reported to have started construction but with no progress = number of suspected projects reported to have started construction but with no progress / total number of projects * 100%;

[0362] The suspected project delay indicators include:

[0363] Suspected project delay alarm rate = number of suspected delayed projects / total number of projects * 100%;

[0364] Indicators of suspected multiple closures of projects include:

[0365] Alarm rate of suspected multiple closures of projects = number of suspected multiple closures / total number of ongoing projects * 100%;

[0366] The suspected service-related accounts that boost investment completion indicators include:

[0367] Alarm rate for suspected service-related investment completion = number of completed projects suspected of inflating investment by service-related investment / total number of projects under construction * 100%;

[0368] The suspected materials recorded to boost investment completion indicators include:

[0369] Alarm rate for suspected material-related investment inflated accounts = number of completed projects suspected of inflating investment by material-related investment / total number of projects under construction * 100%;

[0370] The indicators of suspected untimely project settlement include:

[0371] Rate of early warning of suspected untimely settlement of projects = Number of early warning projects suspected of untimely settlement / Number of 750-35 kV projects in operation with unsettled accounts

[0372] The indicators of reasonable rate of deviation between the guideline and financial expenditure include:

[0373] Reasonable deviation rate = number of reasonable deviation items / total number of items * 100%;

[0374] The suspected unreasonable indicators of large-value service write-offs include:

[0375] Alarm rate for suspected unreasonable write-offs of large-value service entries = number of suspected unreasonable write-offs of large-value service entries / total number of projects under construction * 100%;

[0376] The indicators of mismatch between the timing of production and final settlement include:

[0377] Alarm rate of suspected distribution network projects reporting inaccurate commissioning times = number of suspected distribution network projects with inaccurate commissioning times / projects under construction at 10 kV and below;

[0378] The indicators of suspected work having started but not receiving materials in time include:

[0379] Alarm rate of suspected projects that have started but materials have not been received in time = number of suspected projects that have started but materials have not been received in time / total number of projects * 100%;

[0380] The suspected non-compliant indicators for the return of construction materials include:

[0381] Alert rate for suspected non-compliance with engineering material withdrawal = number of projects with suspected non-compliance with engineering material withdrawal / projects under construction at 10 kV and below * 100%;

[0382] The suspected amount of materials used for projects under construction exceeding the estimated material demand indicators include:

[0383] Alarm rate for suspected excess material requirements for ongoing projects = number of projects suspected excess material requirements / total number of projects * 100%;

[0384] The indicators for failure to handle material transfer procedures in a timely manner include:

[0385] The alarm rate for failure to complete material transfer procedures in a timely manner is: alarm projects / projects put into production this year*100%.

[0386] Example 4:

[0387] See also Figure 3 , an investment execution monitoring system adapted to the characteristics of a multimodal power grid project group, the system is used to execute the investment execution monitoring method adapted to the characteristics of a multimodal power grid project group as described in Example 1, specifically comprising: a data acquisition module, an indicator system construction module, a model construction module, and a monitoring module;

[0388] The data acquisition module is used to obtain basic data from the entire process of historical related project groups from the early planning stage to the final settlement as sample data, vectorize the basic project information through the RAG model, construct an index and store it in the database, and perform data cleaning and conversion to obtain project indicator construction data. At the same time, it obtains real-time basic data of the project groups that need to be monitored on a daily basis, and performs data cleaning and conversion to obtain project operation monitoring data.

[0389] The indicator system construction module is used to: construct a project group indicator system, wherein the project group indicators include lean process control indicators, unit execution monitoring indicators, and special monitoring indicators. The lean process control indicators are general indicators for monitoring the implementation of projects, the unit execution monitoring indicators are indicators for evaluating the effectiveness of plan execution at the monitoring unit level, and the special monitoring indicators are targeted monitoring indicators for monitoring special projects. The logical judgment statements in the project group indicators are converted into mathematical expressions or logical operation symbols to obtain indicator features that can be recognized by the model.

[0390] The model building module is used to: build a model, and build an intelligent monitoring model through machine learning algorithms. The input of the intelligent monitoring model is the basic data of the power grid project group process and various project group indicators. The output of the intelligent monitoring model is the project investment progress, project risks and risk response decisions corresponding to the project risks;

[0391] The monitoring module is used to input project operation monitoring data into the intelligent monitoring model, obtain project investment progress, project risks and risk response decisions corresponding to project risks, and interact with users through the RAG model.

[0392] See also Figure 4 , an investment execution monitoring device adapted to the characteristics of a multimodal power grid project group, comprising a memory and a processor, wherein the memory is used to store computer program code and transmit the computer program code to the processor;

[0393] The processor is configured to execute the investment execution monitoring method adapted to the characteristics of a multimodal power grid project group as described in Example 1 according to the instructions in the computer program code.

[0394] A computer program product includes a computer program, wherein a processor executes the investment execution monitoring method adapted to the characteristics of a multimodal power grid project group as described in embodiment 1.

[0395] The above description is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiment. Any equivalent modifications or changes made by ordinary technicians in this field based on the contents disclosed in the present invention should be included in the protection scope recorded in the claims.

Claims

1. An investment execution monitoring method adapted to the characteristics of a multimodal power grid project group, characterized in that: The monitoring method comprises: S1. Data acquisition: Obtain basic data from the entire process of historical related project groups from early planning to settlement and final accounting as sample data. Use the RAG model to vectorize the basic project information, build an index, and store it in the database. Perform data cleaning and conversion to obtain project indicator construction data. Simultaneously, obtain real-time basic data of the project groups that need to be monitored daily, perform data cleaning and conversion, and obtain project operation monitoring data. S2. Construct a project group indicator system. The project group indicators include lean process control indicators and special monitoring indicators. The lean process control indicators are general indicators for monitoring project implementation. The unit execution monitoring indicators are indicators for evaluating the effectiveness of plan execution at the unit level. The special monitoring indicators are targeted monitoring indicators for monitoring special projects. The logical judgment statements in the project group indicators are converted into mathematical expressions or logical operation symbols to obtain indicator features that can be recognized by the model. S3. Build a model: Use a machine learning algorithm to build an intelligent monitoring model. The input of the intelligent monitoring model is the basic data of the power grid project group process and various project group indicators. The output of the intelligent monitoring model is the project investment progress, project risk, and the decision corresponding to the project risk. S4. Project group monitoring: input the project operation monitoring data into the intelligent monitoring model to obtain the project investment progress, project risks and risk response decisions corresponding to the project risks, and interact with users through the RAG model.

2. The investment execution monitoring method adapted to the characteristics of a multi-modal power grid project group according to claim 1, characterized in that: The lean process control indicators include regulatory indicators, progress indicators, capital management indicators, material management indicators, and data management indicators; The regulatory indicators include the completion index of construction commencement report, construction commencement compliance index, suspected untimely production report index, and production compliance index; The progress indicators include suspected no progress indicators, suspected project delay indicators, and suspected multiple project closure indicators; The fund management indicators include indicators suspected of inflating investment completion by service-related accounts, indicators suspected of inflating investment completion by material-related accounts, indicators suspected of delayed project settlement, indicators of reasonable deviations between guidance lines and financial expenditures, indicators suspected of unreasonable write-offs of large-scale service accounts, and indicators of mismatch between the timing of production and final settlement. The material management indicators include indicators of suspected failure to collect materials in a timely manner after construction has commenced, indicators of suspected non-compliant collection and return of construction materials, indicators of suspected excess of estimated material demand for projects under construction, and indicators of failure to complete material transfer procedures in a timely manner. The special monitoring indicators include the network source unlinked alarm indicator, the network source milestone plan mismatch alarm indicator, the file supplement untimely alarm indicator, the material collection and return procedures untimely alarm indicator, the construction results unachieved alarm indicator, the main equipment delivery untimely indicator, the suspected inability to be put into production on schedule indicator, the grid reinforcement and loss reduction effectiveness indicator, the grid reinforcement and load reduction effectiveness indicator, the milestone target completion delay indicator, the suspected inability to achieve the production target on schedule indicator, the peak summer and winter load reduction indicator, the unapproved start-up indicator, the construction progress mismatch alarm indicator, the procurement application timeliness indicator, the suspected railway and external power project construction progress mismatch indicator, and the suspected production sequence mismatch alarm indicator.

3. The investment execution monitoring method adapted to the characteristics of a multi-modal power grid project group according to claim 2, characterized in that: a. Risk assessment of regulatory indicators: The conditions for determining the risk of the completion indicators of the commencement report are that: ①Current time - earliest start report time>30 days; ② The project has been reported to have started at the source end, but has not been reported to have started in the development professional investment statistics; The risk of the commencement compliance indicator is determined by failure to meet one of the following conditions: ① The project has uploaded the feasibility study approval document, and the feasibility study approval document number and time are not empty; ② For projects of 35 kV and above, the approval documents have been uploaded, and the approval document number and approval time are not blank; The conditions for determining the risk of the suspected untimely reporting of production indicators are that: ① The (current time - latest time of completion report) of the project is greater than 30 days; ② The project has been reported to be put into production at the source end, but has not been reported to be put into production in the development of professional investment statistics; The risk determination condition for the production compliance indicator is: the amount of financial records ≤ total investment * 75%; b. Risk assessment of progress indicators: The conditions for determining the risk of suspected construction no progress indicators are that: ① Construction has been underway for more than 6 months; ②The cumulative construction progress is 0; The conditions for determining the risk of the suspected project delay indicator are that: ①The project has not been put into production; ②The actual construction period (current time - actual start time) exceeds the following time limit; The criteria for determining the risk of multiple closure indicators for suspected projects are: the project closure time has been adjusted three or more times since the project started; c. Risk assessment of fund management indicators: The judgment condition for the suspected service-related account to increase the investment completion index is to meet one of the following conditions: ① Material budget / project budget>20%; ② The cumulative service account / service budget estimate since the start is greater than 70%, and the cumulative material account / material budget estimate since the start is less than 20%; The conditions for determining the risk of the suspected material-related entries driving up the investment completion indicator are: the cumulative material-related entries since the start / material-related budget > 30%, and the cumulative service-related entries since the start = 0; The risk of the suspected untimely settlement indicator is determined to be a risk if one of the following conditions is met: ① For 35 kV and 110 kV projects, the project settlement time is blank, and the difference between the current time and the completion time is between 30 and 60 days; ② For 220-750 kV projects, the project settlement time is blank, and the difference between the current time and the completion time is between 70-100 days; The risk of deviation between the guideline and the reasonable indicator of financial expenditure is determined when one of the following conditions is not met: ① The physical quantity guideline of the project can be calculated; ② The physical quantity guidance line of the project is available; ③|Guideline value - financial expenditure|>0, and |Guideline value - financial expenditure|≤10%*total amount of approved budget estimate; The criteria for determining the risk of unreasonable indicators of suspected large-amount service write-offs are: When used in projects of 35 kV and above, one of the following conditions must be met: ① This month's service cost is less than -1 million; ②|This month's service cost recorded / last month's cumulative service cost recorded since the beginning of the month|>10%; When used in projects of 10 kV and below, one of the following conditions must be met: ① This month's service cost is less than -100,000; ②|This month's service cost recorded / last month's cumulative service cost recorded since the beginning of the month|>10%; The risk of mismatch between the timing of production start-up and final accounting is determined when one of the following conditions is met: ① The commissioning time of the online power grid scale monthly report is blank, and one of the project shutdown times exists; ② The commissioning time of the online power grid scale monthly report is not empty and is later than one of the project shutdown times; ③|Latest commissioning time of a single item of the PMS system - commissioning time of the completion acceptance report|>7 days; d. Risk assessment of material management indicators: The criteria for determining the risk of the indicator of suspected commencement of work but failure to timely receive materials are that: ①The project has actually started; ②Current time - project start time>=3 months; ③The actual amount of materials received is 0; The risk of suspected non-compliant indicators for the return of construction materials is determined if one of the following conditions is met: ① The planned total investment is greater than RMB 500,000, the write-off amount / total project investment is greater than 20%, and the amount of returned materials / total project investment is greater than 20%; ② The cumulative number of returns of the same material in a project is greater than the cumulative number of receipts for the same material*50%; and the amount of returns is greater than RMB 50,000, and the same material in the same project and the same receipt voucher are used for returns three or more times; The conditions for determining the risk of exceeding the estimated material demand indicators for the suspected projects under construction are that: ①The project has not actually been put into production; ②Total investment>500,000 yuan; ③ The total net amount of materials used in the project is greater than 90% of the total investment; The criterion for determining the risk of failure to handle material transfer procedures in a timely manner is that the physical ID of the distribution transformer equipment associated with the commissioned project is inconsistent with the physical ID of the distribution transformer used for the project.

4. The investment execution monitoring method adapted to the characteristics of a multi-modal power grid project group according to claim 2, characterized in that: Risk assessment of the special monitoring indicators: The conditions for determining the risk of the network source unassociated alarm indicator are: the power supply project is not associated with the power source; The judgment condition for the network source milestone plan mismatch alarm indicator is: |Power transmission project planned commissioning time - power supply project's earliest planned commissioning time|> 6 months; The judgment condition for the alarm indicator of untimely file supplementation is to meet one of the following conditions: ① Associate user profiles; ②The scale of related equipment files is not equal to the scale of project commissioning; The criteria for determining the alarm indicator for untimely handling of materials return procedures is to meet one of the following conditions: ①Current time - commissioning time>3 months; ② The substation capacity used is not equal to the associated distribution transformer capacity; The judgment condition for the construction effect not reaching the warning indicator is that one of the following conditions is met: ① The project has associated operational equipment (distribution transformer, main transformer, and line); ② Equipment utilization rate in the month following equipment operation = 0; The risk of the main equipment delivery delay indicator is determined by the following conditions: ① The M kV project has been under construction for more than N months; ② The main equipment arrival rate of the mkV project <n; The conditions for determining the risk of the suspected failure to put into production on schedule are that: ①The project has been under construction for more than 6 months; ② The construction progress of the X kV project is less than Y; ③ The progress of the x kV project in receiving funds is less than y; The judgment condition for the occurrence of risk of the effective index of grid reinforcement and loss reduction is: after the grid optimization and reinforcement project in the target area is put into operation, the line loss rate decreases by more than 0.2% compared with the level before the commissioning; The criterion for determining the risk of the grid reinforcement and load reduction effectiveness index is that one of the following conditions is not met: ① After the grid optimization and reinforcement project in the target area was put into operation, the number of overloaded units decreased by 30% compared with before the commissioning, and the number of heavy-loaded units decreased by 20%; ② After the grid optimization and reinforcement project in the target area is put into operation, (the cumulative duration of heavy overload before the commissioning - the cumulative duration of heavy overload after the commissioning) / the cumulative duration of heavy overload before the commissioning > 20%; The conditions for determining the risk of lagging behind in the milestone target completion indicator are that: ①The project has started; ② A single milestone is not completed for more than one month after the planned completion time; The conditions for determining the risk of the suspected production target not being achieved on schedule are: For the peak summer project, the following requirements must be met at the same time: ①The project has not been put into production; ②(Planned production time - current time) < 2 months; ③Cumulative construction progress <60%; ④Cumulative deposit progress <50%; For the peak winter project, it also meets the following requirements: ①The project has not been put into production; ②(Planned production time - current time) < 2 months; ③Cumulative construction progress <50%; ④Cumulative deposit progress <40%; The risk of the load shedding indicator during the peak summer and winter seasons is determined if one of the following conditions is not met: ①Compared with the same period last year, the current data in the target area shows that the number of overloaded units has decreased by 40% and the number of heavy-loaded units has decreased by 30% since the start of production; ②Compared with the current data of the target area in the same period of previous years, (the absolute value of the accumulated heavy overload time before production - the accumulated heavy overload time after production) / the accumulated heavy overload time before production > 20%; The risk of starting construction without approval is determined by satisfying one of the following conditions: ① The business expansion supporting project has started without user approval; ②35kV business expansion supporting project start time - user approval time> 9 months; ③110kV business expansion supporting project start time - user approval time> 16 months; The risk of the construction progress mismatch warning indicator is determined by satisfying one of the following conditions: ①10kV business expansion supporting project commissioning time - user demand power time> 6 months ②35kV business expansion supporting project commissioning time - user demand power consumption time> 7 months; ③ The commissioning time of 110kV business expansion supporting project minus the power demand time of users>6 months; The risk of the procurement application timeliness indicator is determined to be at risk if one of the following conditions is met: ① The first material procurement application time for the 35kV business expansion supporting project minus the user approval time is > 6 months; ② The time from the first material procurement application to the user approval for the 110kV business expansion supporting project is greater than 10 months; The conditions for determining the risk of mismatch between the construction progress of the railway and the external power project are that: ①The project has not been put into production; ② Traction station construction progress - external power construction progress > 20% ③ The amount of materials received for external power projects / estimated equipment purchases is less than 30%; The risk of the suspected production timing mismatch alarm indicator is determined to be a risk if one of the following conditions is met: ① Wind power transmission: 35kV grid connection project start time - power source start time > 7 months (power source construction period - grid connection project construction period); 110kV grid connection project start time - power source start time > 3 months; 220kV grid connection project start time - power source start time > 12 months; ② Photovoltaic power transmission: 35kV grid connection project start time - power source start time > 1 month, 110kV grid connection project start time - power source start time > 5 months, 220kV grid connection project start time - power source start time > 2 months.

5. The investment execution monitoring method adapted to the characteristics of a multi-modal power grid project group according to claim 1, characterized in that: In S3, the machine learning algorithm used is the random forest algorithm, which obtains multiple classification results by sampling the input data, and selects the classification result with the largest number of occurrences among all the classification results as the final result of the random forest algorithm; Input the project indicator construction data and indicator features into the random forest algorithm to determine the number of decision trees n in the random forest. Initially, each decision tree in the random forest is in an untrained state; 70% of the project indicator construction data is divided into a training set and 30% is divided into a test set. For each decision tree, samples are randomly selected from the training set with replacement to construct a sub-training set. The original training set includes m samples, and each tree sub-training set includes m samples. When constructing the node splitting of each tree, some features are randomly selected from all features, and the said partial features are k features, k is less than the total number of features, and the best splitting features are selected according to the Gini index splitting criterion; Repeat the above node splitting process until the stopping condition or the maximum depth limit of the tree is reached; Repeat the above training process for n decision trees in the random forest, and draw a conclusion after the training is completed. The conclusion is the category prediction in the classification task or the numerical prediction in the regression task; The trained random forest model is tested using a test set. If the accuracy of the trained random forest model is less than the set value, the sample data sampling range is expanded, and the data is re-acquired and the random forest model is trained. If the accuracy of the trained random forest model is greater than the set value, the trained random forest model is used as an intelligent monitoring model.

6. The investment execution monitoring method adapted to the characteristics of a multi-modal power grid project group according to claim 1, characterized in that: Said S3 also includes: said random forest model including the weight of each indicator in the project group indicator system, wherein the higher the weight of each indicator is, the higher the importance or risk of the indicator is, and the sum of the weights of each indicator is 1, and the total risk score of the project is calculated by weighted average method; Setting a project risk monitoring and scoring system, wherein the project risk monitoring and scoring system includes assigning weights to project risks output by the intelligent monitoring model, setting a total project risk score for each project, and setting risk thresholds for power grid project groups, wherein the risk thresholds include high risk thresholds, medium risk thresholds, and low risk thresholds; The risk score is obtained by combining the various project risks and corresponding weights output by the intelligent monitoring model. During the evaluation, the project risk score is obtained by subtracting the risk score from the total project risk score. The project risk score and the risk threshold are compared. When the risk monitoring score is lower than the low risk threshold, it means that the project operation is stable and continues to proceed according to the original plan. When the risk monitoring score is between the low risk threshold and the medium risk threshold, it means that the project risk is low and potential risks need to be paid attention to, but no major measures need to be taken immediately. When the risk monitoring score is between the medium risk threshold and the high risk threshold, it means that the project risk is large and immediate countermeasures need to be taken. When the risk monitoring score is higher than the high risk threshold, it means that the project risk is extremely high and the project needs to be suspended and the risk response decision corresponding to the project risk needs to be output.

7. The investment execution monitoring method adapted to the characteristics of a multi-modal power grid project group according to claim 1, characterized in that: In S1, the private domain data is vectorized using the RAG model, and then an index is constructed and stored in the database, including: Process semi-structured data in various formats, convert them into the same processing paradigm, obtain data from different data sources, and perform standardization; Filter, compress and format data to remove noise and redundant information and obtain metadata; According to the token restrictions of the embedding model, the text is segmented into appropriate granularity, such as sentence segmentation or fixed-length segmentation; Convert text into vector representation through embedding model; Build an index for the vectorized data and write it to the database. Select Elasticsearch as the database for the business scenario. In S4, the RAG model performs data retrieval by combining similarity retrieval with full-text retrieval. The similarity retrieval returns relevant records by calculating the similarity score between the query vector and the stored vector, and the full-text retrieval constructs an inverted index using keywords. Integrate the retrieved relevant knowledge with the user's question to form a prompt, and then generate the answer through a large language model.

8. An investment execution monitoring system adapted to the characteristics of multimodal power grid project groups, characterized in that: The system is used to execute the investment execution monitoring method adapted to the characteristics of a multimodal power grid project group according to any one of claims 1 to 7, specifically comprising: a data acquisition module, an indicator system construction module, a model construction module, and a monitoring module; The data acquisition module is used to obtain basic data from the entire process of historical related project groups from the early planning stage to the final settlement as sample data, vectorize the basic project information through the RAG model, construct an index and store it in the database, and perform data cleaning and conversion to obtain project indicator construction data. At the same time, it obtains real-time basic data of the project groups that need to be monitored on a daily basis, and performs data cleaning and conversion to obtain project operation monitoring data. The indicator system construction module is used to: construct a project group indicator system, wherein the project group indicators include lean process control indicators, unit execution monitoring indicators, and special monitoring indicators. The lean process control indicators are general indicators for monitoring the implementation of projects, the unit execution monitoring indicators are indicators for evaluating the effectiveness of plan execution at the monitoring unit level, and the special monitoring indicators are targeted monitoring indicators for monitoring special projects. The logical judgment statements in the project group indicators are converted into mathematical expressions or logical operation symbols to obtain indicator features that can be recognized by the model. The model building module is used to: build a model, and build an intelligent monitoring model through machine learning algorithms. The input of the intelligent monitoring model is the basic data of the power grid project group process and various project group indicators. The output of the intelligent monitoring model is the project investment progress, project risks and risk response decisions corresponding to the project risks; The monitoring module is used to input project operation monitoring data into the intelligent monitoring model, obtain project investment progress, project risks and risk response decisions corresponding to project risks, and interact with users through the RAG model.

9. An investment execution monitoring device adapted to the characteristics of a multi-modal power grid project group, characterized in that: comprising a memory and a processor, wherein the memory is configured to store computer program code and transmit the computer program code to the processor; The processor is configured to execute the investment execution monitoring method adapted to the characteristics of a multimodal power grid project group according to any one of claims 1 to 7 according to the instructions in the computer program code.

10. A computer program product comprising a computer program, characterized in that The computer program is executed by a processor to implement the investment execution monitoring method adapted to the characteristics of a multimodal power grid project group as described in any one of claims 1 to 7.

Images