Electric power marketing and selling power data collection method fusing RPA and AI technologies

Abstract

The invention relates to the technical field of big data, in particular to an electric power marketing and selling power data collection method fusing RPA and AI technologies. The electricity sale data collection method comprises the following steps: the RPA collects structured electricity sale quantity data; the RPA collects unstructured national economy data and performs comprehensive data cleaning on the extracted webpage text; annotating the text data; the marked data set is input into a BERT model; performing information extraction on the labeled data set by utilizing the trained BERT model; and the data are combined with structured power sale quantity data extracted from a marketing system to generate a unified structured data format, so that high-quality data support is provided for subsequent data analysis and decision-making. The method has the following remarkable advantages that efficient data extraction and cleaning are achieved through RPA automation, manual intervention is greatly reduced, and the overall data processing speed is remarkably increased.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a method for collecting power marketing and sales data by integrating RPA and AI technologies. Background Art

[0002] In the field of power marketing and electricity sales analysis, traditional data collection and processing methods mainly rely on manual operations. This method usually requires operators to manually log in to the marketing system to export structured electricity sales data. At the same time, operators also need to manually log in to the official website of the National Bureau of Statistics, locate the page containing national economic data, manually extract relevant data from it, and enter it into a pre-set structured table. The whole process is not only cumbersome, but also has a high operational risk, which can easily lead to data omissions or processing errors. Especially when the amount of data is huge and the updates are frequent, manual processing methods are difficult to respond quickly, which can easily lead to data processing delays, thereby affecting subsequent power marketing analysis and decision-making. At present, the implementation schemes that are most similar to the present invention mainly include the following:

[0003] (I) Manual data collection and processing: This is the most traditional method. Although it is flexible, it is inefficient and prone to errors. Operators must process large amounts of data step by step, and when faced with complex web page content, data omissions or misinterpretations are likely to occur.

[0004] (II) Single RPA solution: Some companies use RPA technology to automate some data collection tasks, such as simulating manual operations to automatically log in to the website and download data. This method improves efficiency to a certain extent, but there is still the problem of data processing complexity. RPA scripts can only process structured data downloads and cannot automatically process data cleaning and conversion of unstructured text.

[0005] Explanation of relevant terms:

[0006] RPA: Robotic Process Automation, RPA is a technology that uses software robots to simulate human operations on computers to automatically perform rule-based, repetitive tasks. RPA can automatically interact with various business systems during the data collection process, reducing manual intervention and improving work efficiency.

[0007] AI: Artificial intelligence is a technology that enables computers to perform tasks that usually require human intelligence, such as understanding language, recognizing images, making decisions, and solving problems. By using algorithms and large amounts of data, AI systems can learn, adapt to new information, and perform a variety of complex tasks, thereby improving efficiency and innovation capabilities, and promoting the development of automation and intelligence in multiple industries.

[0008] Power marketing and sales analysis: This refers to the process of collecting, organizing, analyzing and evaluating data related to power sales. The purpose is to understand the current situation and trends of power sales and provide support for power companies to formulate business strategies.

[0009] In the analysis of power sales, the accuracy and timeliness of data are crucial to the effectiveness of the analysis results. However, the main problems in the data collection and processing of existing technologies include: data collection work mostly relies on manual operations, and operators need to frequently log in to relevant systems and manually extract data. This method is not only time-consuming and inefficient, but also prone to data omissions or errors due to human negligence.

[0010] The electricity sales data exported from the marketing system comes from reports from 27 power grid provinces. After each power grid province submits the data, the system will automatically verify the data according to the preset verification rules. If a problem is found, the relevant power grid province will be notified to make corrections and resubmit. Since the correction and submission time of each power grid province is inconsistent, operators need to log in to the system frequently and export data repeatedly to ensure the accuracy and completeness of the final data. This process not only increases the workload, but also reduces the efficiency of data collection.

[0011] Data extracted from websites such as the National Bureau of Statistics are usually unstructured text data, which must go through a series of cleaning and processing steps, including removing irrelevant information, extracting useful content, etc. This further increases the reliance on manual intervention, which can easily lead to inconsistent data processing or information loss.

[0012] These problems seriously affect the efficiency and accuracy of data processing, and thus affect subsequent electricity sales analysis and decision-making. Summary of the invention

[0013] The purpose of this invention is to provide an automated data collection and processing method that integrates RPA and AI technologies. RPA scripts are used to automatically complete tasks such as logging in, navigating, and extracting data from data sources such as marketing systems and the official website of the National Bureau of Statistics, thereby reducing manual intervention and improving the efficiency of data collection. The specific technical solution is as follows:

[0014] A method for collecting electricity sales data in power marketing integrating RPA and AI technologies includes the following processes:

[0015] S100: RPA collects structured electricity sales data;

[0016] S200: RPA collects unstructured national economic data and performs comprehensive data cleaning on the extracted web page text;

[0017] S300: Text data annotation; the BIOES (Beginning, Inside, Outside, End, Single) annotation strategy is used in the annotation process, where:

[0018] Tag B: indicates the beginning of the entity;

[0019] Label I: represents the middle part of the entity;

[0020] Tag O: indicates a non-entity part;

[0021] Tag E: indicates the end of the entity;

[0022] Label S: represents a separate word entity;

[0023] Tag VALUE: indicates that the content marked by this tag is the attribute value of the subject;

[0024] Tag SUB: indicates that the content marked by this tag is the main body;

[0025] Tag OBJ: indicates that the content marked by this tag is an object;

[0026] Tag REL: indicates that the content marked by this tag is a relationship;

[0027] S400: Input the labeled data set into the BERT model; by training the labeled data, the BERT model can learn to identify entities and relations from texts and generate corresponding entity, relation, and attribute triples;

[0028] S500: Use the trained BERT model to extract information from the annotated data set, extract (subject-relationship-object) and (subject-relationship-attribute) triples from unstructured text; combine it with the structured electricity sales data extracted from the marketing system to generate a unified structured data format, providing high-quality data support for subsequent data analysis and decision-making.

[0029] Preferably, the S100 includes the following process:

[0030] S110: initialization process;

[0031] S120: system login process;

[0032] S130: Sales table data export process;

[0033] S140: Electricity sales data export process.

[0034] Preferably, the specific process of S110 is as follows:

[0035] S111: Determine whether the configuration file exists. If yes, go to S112; otherwise, go to S116;

[0036] S112: Open the configuration file and go to S113;

[0037] S113: Read the worksheet name, and go to S114;

[0038] S114: read each row of data in a loop and record it in the dictionary variable, then go to S115;

[0039] S115: close the configuration table;

[0040] S116: An exception is thrown.

[0041] Preferably, the specific process of S120 is as follows:

[0042] S121: close the browser and go to S122;

[0043] S122: Open the energy internet marketing service system and go to S123;

[0044] S123: Determine whether the maximum number of login times is exceeded, if yes, go to S126, otherwise go to S124;

[0045] S124: Select a management unit, enter a user name and password, and go to S125;

[0046] S125: Determine whether the login is successful, if yes, go to S127, otherwise return to S123;

[0047] S126: record failure information, and go to S127;

[0048] S127: Complete the system login process.

[0049] Preferably, the specific process of S130 is as follows:

[0050] S131: Navigate to the page of “Sales Table 5_Electricity Sales Status of the Whole Industry”, and go to S132;

[0051] S132: Delete the sales table 5_all industry sales electricity data file, and go to S133;

[0052] S133: Determine whether the maximum number of scans is exceeded, if yes, go to S135, otherwise go to S134;

[0053] S134: simulate click, export sales table 5_all industry sales electricity data in pages, and go to S135;

[0054] S135: Determine whether the export sales table 5_all industry sales electricity file exists, if yes, go to S137, otherwise return to S133;

[0055] S136: Record the sales table 5_industry-wide sales electricity export failure information, and go to S137;

[0056] S137: Complete the sales table data export process.

[0057] Preferably, the specific process of S140 is as follows:

[0058] S141: Navigate to the "Industry Express" page and go to S142;

[0059] S142: Delete Express Report 1_Industry Express Report data file, and go to S143;

[0060] S143: Determine whether the maximum number of scans is exceeded, if yes, go to S146, otherwise go to S144;

[0061] S144: simulate a click, page the electricity sales data, and go to S145;

[0062] S145: Determine whether the export express report 1_industry express report file exists, if yes, go to S147, otherwise return to S143;

[0063] S146: Record the failure information of Quick Report 1_Industry Quick Report, and go to S147;

[0064] S147: The electricity sales data export process is completed.

[0065] Preferably, the specific process of S200 is as follows:

[0066] S210: Log in to the website of the National Bureau of Statistics;

[0067] S220: Extract national economic data.

[0068] Preferably, the specific process of S210 is as follows:

[0069] S211: Close the browser and go to S212;

[0070] S212: Determine whether the maximum number of login times is exceeded, if yes, go to S215, otherwise go to S213;

[0071] S213: Log in to the National Bureau of Statistics website and go to S214;

[0072] S214: Determine whether the login is successful, if yes, go to S216, otherwise return to S212;

[0073] S215: record failure information, and go to S216;

[0074] S216: Complete the login process.

[0075] Preferably, the specific process of S220 is as follows:

[0076] S221: Click the data navigation menu and go to S222;

[0077] S222: Click "Publish Submenu" of "Data Details Page" to jump to the data publishing details page and go to S223;

[0078] S223: Obtain the list of dynamically published data in the list view in the "data publishing details page", and go to S224;

[0079] S224: Set the target page found flag to False, and go to S225;

[0080] S225: loop through the dynamic data release list, filter out the data list containing "national economy" on the specified date, modify the flag; go to S226;

[0081] S226: Determine whether the target page is found, if yes, go to S228, otherwise go to S227;

[0082] S227: Click the next page button to jump to the next page and return to S223;

[0083] S228: open the target page and go to S229;

[0084] S229: Get the text content of all paragraph elements;

[0085] S230: Merge the texts and clean the text contents, then go to S240;

[0086] S240: Saving the cleaned text to a PDF file.

[0087] By combining RPA with AI technology, the present invention achieves a high degree of automation and intelligence in data collection and analysis. First, the RPA tool automates the process of collecting structured electricity sales data from the energy internet marketing service system and extracting unstructured national economic data from the National Bureau of Statistics. This process reduces the reliance on manual operations and greatly improves the efficiency and accuracy of data collection.

[0088] The data cleaning step implemented by RPA automatically removes redundant information and invalid content to ensure the integrity and consistency of the data.

[0089] The cleaned data is annotated and fed into the BERT model for training, which automatically extracts and generates structured triple information. This process not only improves the data structure processing capabilities, but also enables the entities, relationships, and attributes in complex texts to be accurately extracted and formatted.

[0090] Therefore, the present invention has the following significant advantages:

[0091] Improved efficiency: RPA automation enables efficient data extraction and cleaning, greatly reducing manual intervention and significantly improving overall data processing speed.

[0092] Accuracy and consistency: RPA ensures consistency in data extraction and cleaning, reducing errors and inconsistencies common in manual processing.

[0093] Automated structured processing: Through the training of the BERT model, it is possible to automatically extract entities, relationships, and attributes in the text, generate structured triple information, and provide high-quality input for subsequent data analysis and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 It is a schematic diagram of the process flow of RPA collecting electricity sales data in an embodiment of the present invention.

[0095] Figure 2 It is a schematic diagram of the initialization process of the RPA collecting electricity sales data in an embodiment of the present invention.

[0096] Figure 3 This is a schematic diagram of the system login process of the RPA collecting electricity sales data in an embodiment of the present invention.

[0097] Figure 4 It is a schematic diagram of the sales table data export process of the RPA collecting electricity sales data in an embodiment of the present invention.

[0098] Figure 5 It is a schematic diagram of the process of exporting electricity sales data during the process of RPA collecting electricity sales data in an embodiment of the present invention.

[0099] Figure 6 It is a schematic diagram of the National Bureau of Statistics official website login process in the process of RPA collecting unstructured national economic data in an embodiment of the present invention.

[0100] Figure 7 It is a schematic diagram of the national economic data extraction process in the process of RPA collecting unstructured national economic data in an embodiment of the present invention. DETAILED DESCRIPTION

[0101] A method for collecting electricity sales data in power marketing integrating RPA and AI technologies includes the following processes:

[0102] S100: RPA collects structured electricity sales data; S100 includes the following process:

[0103] S110: Initialization process; the main task of this process is to load and prepare the configuration information required for the RPA task to run. By reading the pre-prepared configuration file (usually an Excel file), key information such as the URL, login username and password of the energy Internet marketing service system is obtained to prepare for subsequent system login and data operations. The specific process is as follows:

[0104] S111: Determine whether the configuration file exists. If yes, go to S112; otherwise, go to S116;

[0105] S112: Open the configuration file and go to S113;

[0106] S113: Read the worksheet name, and go to S114;

[0107] S114: read each row of data in a loop and record it in the dictionary variable, then go to S115;

[0108] S115: close the configuration table;

[0109] S116: An exception is thrown.

[0110] S120: System login process; based on the configuration information read in the initialization process, automatically log in to the energy Internet marketing service system; the RPA script will enter the user name and password and perform identity authentication; if the login fails, the system will automatically retry multiple times until the login is successful or the number of retries is exceeded; once the maximum number of retries is exceeded, the system will output an error message. The specific process is as follows:

[0111] S121: close the browser and go to S122;

[0112] S122: Open the energy internet marketing service system and go to S123;

[0113] S123: Determine whether the maximum number of login times is exceeded, if yes, go to S126, otherwise go to S124;

[0114] S124: Select a management unit, enter a user name and password, and go to S125;

[0115] S125: Determine whether the login is successful, if yes, go to S127, otherwise return to S123;

[0116] S126: record failure information, and go to S127;

[0117] S127: Complete the system login process.

[0118] S130: Sales table data export process; after successfully logging into the system, RPA will automatically navigate to the "Sales Table 5_All Industry Sales Electricity Situation" page, and export the actual electricity sales data of 27 power grids and provinces last month in pages; during the export process, RPA will verify each exported file to ensure the integrity of the file; if the export fails, the system will automatically retry until the export is successful or the number of retries is exhausted. The specific process is as follows:

[0119] S131: Navigate to the page of “Sales Table 5_Electricity Sales Status of the Whole Industry”, and go to S132;

[0120] S132: Delete the sales table 5_all industry sales electricity data file, and go to S133;

[0121] S133: Determine whether the maximum number of scans is exceeded, if yes, go to S135, otherwise go to S134;

[0122] S134: simulate click, export sales table 5_all industry sales electricity data in pages, and go to S135;

[0123] S135: Determine whether the export sales table 5_all industry sales electricity file exists, if yes, go to S137, otherwise return to S133;

[0124] S136: Record the sales table 5_industry-wide sales electricity export failure information, and go to S137;

[0125] S137: Complete the sales table data export process.

[0126] S140: Electricity sales data export process; RPA navigates to the Quick Report 1_Industry Quick Report page and continues to export electricity sales data of 27 power grids and provinces in pages; Similar to the sales table data export process, this process will also verify the exported file to ensure that the data export is successful; if it fails, the system will retry multiple times until the file is successfully exported or the maximum number of retries is triggered. The specific process is as follows:

[0127] S141: Navigate to the "Industry Express" page and go to S142;

[0128] S142: Delete Express Report 1_Industry Express Report data file, and go to S143;

[0129] S143: Determine whether the maximum number of scans is exceeded, if yes, go to S146, otherwise go to S144;

[0130] S144: simulate a click, page the electricity sales data, and go to S145;

[0131] S145: Determine whether the export express report 1_industry express report file exists, if yes, go to S147, otherwise return to S143;

[0132] S146: Record the failure information of Quick Report 1_Industry Quick Report, and go to S147;

[0133] S147: The electricity sales data export process is completed.

[0134] S200: RPA collects unstructured national economic data and performs comprehensive data cleaning on the extracted web page text; the specific process is as follows:

[0135] S210: Log in to the website of the National Bureau of Statistics; the specific process of S210 is as follows:

[0136] S211: Close the browser and go to S212;

[0137] S212: Determine whether the maximum number of login times is exceeded, if yes, go to S215, otherwise go to S213;

[0138] S213: Log in to the National Bureau of Statistics website and go to S214;

[0139] S214: Determine whether the login is successful, if yes, go to S216, otherwise return to S212;

[0140] S215: record failure information, and go to S216;

[0141] S216: Complete the login process;

[0142] S220: Extract national economic data; simulate user operations by clicking on components to accurately navigate to the page containing target data. When the page is loaded, the system will use the "Get Similar Elements List" component to extract all relevant paragraph data on the page. This component is set to the "Get Element Text Content" operation type, which can automatically identify and capture text on the page to ensure data integrity; all extracted text data will be merged into a complete data set through the "Aggregate Text" component, providing a basis for subsequent processing and ensuring centralized storage of data. After data collection is completed, the system embeds a Python script for data cleaning, removing redundant information through regular expressions and text processing libraries; finally, the cleaned data will be saved as a PDF file, providing a high-quality input data source for model training and further analysis.

[0143] The specific process is as follows:

[0144] S221: Click the data navigation menu and go to S222;

[0145] S222: Click "Publish Submenu" of "Data Details Page" to jump to the data publishing details page and go to S223;

[0146] S223: Obtain the list of dynamically published data in the list view in the "data publishing details page", and go to S224;

[0147] S224: Set the target page found flag to False, and go to S225;

[0148] S225: loop through the dynamic data release list, filter out the data list containing "national economy" on the specified date, modify the flag; go to S226;

[0149] S226: Determine whether the target page is found, if yes, go to S228, otherwise go to S227;

[0150] S227: Click the next page button to jump to the next page and return to S223;

[0151] S228: open the target page and go to S229;

[0152] S229: Get the text content of all paragraph elements;

[0153] S230: Merge the texts and clean the text contents, then go to S240;

[0154] S240: Saving the cleaned text to a PDF file.

[0155] S300: Text data annotation; the BIOES (Beginning, Inside, Outside, End, Single) annotation strategy is used in the annotation process, where:

[0156] Tag B: indicates the beginning of the entity;

[0157] Label I: represents the middle part of the entity;

[0158] Tag O: indicates a non-entity part;

[0159] Tag E: indicates the end of the entity;

[0160] Label S: represents a separate word entity;

[0161] Tag VALUE: indicates that the content marked by this tag is the attribute value of the subject;

[0162] Tag SUB: indicates that the content marked by this tag is the main body;

[0163] Tag OBJ: indicates that the content marked by this tag is an object;

[0164] Tag REL: indicates that the content marked by this tag is a relationship;

[0165] For example, for the text "The manufacturing purchasing managers' index (PMI) is 49.1%, down 0.3 percentage points from the previous month", the data annotation results are shown in Table 1 below:

[0166] Table 1 Data annotation results

[0167]

[0168] S400: Input the labeled data set into the BERT model; by training on these labeled data, the BERT model can learn to identify entities and relations from text and generate corresponding entity, relationship, and attribute triples; during the training process, the model continuously optimizes parameters to improve the accuracy of extracting triples from text.

[0169] For example, after the above labeled data is trained by the model, the following triple information can be generated:

[0170] ("Manufacturing Purchasing Managers Index (PMI)", "value", "49.1%").

[0171] (“Manufacturing Purchasing Managers’ Index (PMI)”, “down from the previous month”, “0.3 percentage points”).

[0172] After processing, these triple information will be converted into a structured data format. The converted structured data is shown in Table 2.

[0173] Table 2 National Economic Data

[0174] Serial number Industry Name Purchasing Managers Index (PMI) Compared with the previous month 1 manufacturing 49.1％ -0.3％

[0175] S500: Use the trained BERT model to extract information from the annotated data set, extract "subject-relationship-object" and "subject-relationship-attribute" triples from unstructured text; combine it with the structured electricity sales data extracted from the marketing system to generate a unified structured data format, providing high-quality data support for subsequent data analysis and decision-making.

Claims

1. A method for collecting electricity sales data in power marketing integrating RPA and AI technologies, characterized in that: The process includes the following: S100: RPA collects structured electricity sales data; S200: RPA collects unstructured national economic data and performs comprehensive data cleaning on the extracted web page text; S300: Text data annotation; the BIOES (Beginning, Inside, Outside, End, Single) annotation strategy is used in the annotation process, where: Tag B: indicates the beginning of the entity; Label I: represents the middle part of the entity; Tag O: indicates a non-entity part; Tag E: indicates the end of the entity; Label S: represents a separate word entity; Tag VALUE: indicates that the content marked by this tag is the attribute value of the subject; Tag SUB: indicates that the content marked by this tag is the main body; Tag OBJ: indicates that the content marked by this tag is an object; Tag REL: indicates that the content marked by this tag is a relationship; S400: Input the labeled data set into the BERT model; by training the labeled data, the BERT model can learn to identify entities and relations from texts and generate corresponding entity, relation, and attribute triples; S500: Use the trained BERT model to extract information from the annotated data set, extract (subject-relationship-object) and (subject-relationship-attribute) triples from unstructured text; combine it with the structured electricity sales data extracted from the marketing system to generate a unified structured data format, providing high-quality data support for subsequent data analysis and decision-making.

2. According to claim 1, a method for collecting electricity sales data by integrating RPA and AI technology, characterized in that: S100 includes the following processes: S110: initialization process; S120: system login process; S130: Sales table data export process; S140: Electricity sales data export process.

3. According to claim 2, a method for collecting electricity sales data by integrating RPA and AI technology, characterized in that: The specific process of S110 is as follows: S111: Determine whether the configuration file exists. If yes, go to S112; otherwise, go to S116; S112: Open the configuration file and go to S113; S113: Read the worksheet name, and go to S114; S114: read each row of data in a loop and record it in the dictionary variable, then go to S115; S115: close the configuration table; S116: An exception is thrown.

4. According to claim 2, a method for collecting electricity sales data by integrating RPA and AI technology, characterized in that: The specific process of S120 is as follows: S121: close the browser and go to S122; S122: Open the energy internet marketing service system and go to S123; S123: Determine whether the maximum number of login times is exceeded, if yes, go to S126, otherwise go to S124; S124: Select a management unit, enter a user name and password, and go to S125; S125: Determine whether the login is successful, if yes, go to S127, otherwise return to S123; S126: record failure information, and go to S127; S127: Complete the system login process.

5. According to claim 2, a method for collecting electricity sales data by integrating RPA and AI technology, characterized in that: The specific process of S130 is as follows: S131: Navigate to the page of "Sales Table 5 Electricity Sales Status of All Industries", and go to S132; S132: Delete the sales table 5_all industry sales electricity data file, and go to S133; S133: Determine whether the maximum number of scans is exceeded, if yes, go to S135, otherwise go to S134; S134: simulate click, export sales table 5_all industry sales electricity data in pages, and go to S135; S135: Determine whether the export sales table 5_all industry sales electricity file exists, if yes, go to S137, otherwise return to S133; S136: Record the sales table 5_industry-wide sales electricity export failure information, and go to S137; S137: Complete the sales table data export process.

6. According to claim 2, a method for collecting electricity sales data by integrating RPA and AI technology, characterized in that: The specific process of S140 is as follows: S141: Navigate to the "Industry Express" page and go to S142; S142: Delete Express Report 1_Industry Express Report data file, and go to S143; S143: Determine whether the maximum number of scans is exceeded, if yes, go to S146, otherwise go to S144; S144: simulate a click, page the electricity sales data, and go to S145; S145: Determine whether the export express report 1_industry express report file exists, if yes, go to S147, otherwise return to S143; S146: Record the failure information of Quick Report 1_Industry Quick Report, and go to S147; S147: The electricity sales data export process is completed.

7. According to claim 1, a method for collecting electricity sales data by integrating RPA and AI technology, characterized in that: The specific process of S200 is as follows: S210: Log in to the website of the National Bureau of Statistics; S220: Extract national economic data.

8. According to claim 7, a method for collecting electricity sales data by integrating RPA and AI technology, characterized in that: The specific process of S210 is as follows: S211: Close the browser and go to S212; S212: Determine whether the maximum number of login times is exceeded, if yes, go to S215, otherwise go to S213; S213: Log in to the National Bureau of Statistics website and go to S214; S214: Determine whether the login is successful, if yes, go to S216, otherwise return to S212; S215: record failure information, and go to S216; S226: Complete the login process.

9. According to claim 7, a method for collecting electricity sales data by integrating RPA and AI technology, characterized in that: The specific process of S220 is as follows: S221: Click the data navigation menu and go to S222; S222: Click "Publish Submenu" of "Data Details Page" to jump to the data publishing details page and go to S223; S223: Get the list of dynamically published data in the list view in the "data publishing details page", and go to S224; S224: Set the target page found flag to False, and go to S225; S225: loop through the dynamic data release list, filter out the data list containing "national economy" on the specified date, and modify the flag; Go to S226; S226: Determine whether the target page is found, if yes, go to S228, otherwise go to S227; S227: Click the next page button to jump to the next page and return to S223; S228: open the target page and go to S229; S229: Get the text content of all paragraph elements; S230: Merge the texts and clean the text contents, then go to S240; S240: Saving the cleaned text to a PDF file.

10. According to claim 9, a method for collecting electricity sales data by integrating RPA and AI technology, characterized in that: The text content cleaning in S230 includes: embedding a Python script, removing HTML tags, JavaScript scripts, CSS styles, and redundant blank characters and hyperlinks through regular expressions and text processing libraries to ensure the quality of subsequent AI model training data.

Images