A method and system for rapid response to power supply work orders based on similarity analysis

Abstract

This invention relates to the field of rapid response technology for power supply work orders, and particularly to a method and system for rapid response to power supply work orders based on similarity analysis. The method includes: collecting work order data; identifying the work order type; calling a work order template for response based on the work order type, and performing similarity analysis on the work order data to find responses to similar historical work orders; assisting in reviewing the response content; and reminding the user that the response content is incorrect. The response system includes a collection module for collecting work order data; an identification module for identifying the work order type; a calling module for calling a work order template for response based on the work order type, and performing similarity analysis on the work order data to find responses to similar historical work orders; an review module for assisting in reviewing the response content; and a reminder module for reminding the user that the response content is incorrect. This invention solves the problem of low efficiency in manual work order responses.

Description

Technical Field

[0001] This invention relates to the field of rapid response technology for power supply work orders, and in particular to a rapid response method and system for power supply work orders based on similarity analysis. Background Technology

[0002] Currently, all user electricity demand work orders in the power grid are responded to manually. Before responding manually, the work order type needs to be determined based on its content, and a corresponding response needs to be given according to the different work order types and contents. Manual response is inefficient, and the response content may contain errors or be non-standard, resulting in low work order response quality and a low pass rate. Low-quality response work orders are often required to be returned and re-responded, which seriously affects the service quality and efficiency of the power supply party. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for rapid response to power supply work orders based on similarity analysis, thereby solving the problem of low efficiency in manual work order response.

[0004] To solve the above technical problems, the technical solution of the present invention is as follows:

[0005] Firstly, a rapid response method for power supply work orders based on similarity analysis includes:

[0006] Step 1: Collect work order data;

[0007] Step 2: Identify the work order type in the work order data;

[0008] Step 3: Based on the work order type, call the work order template for replying, and perform similarity analysis on the work order data to find replies to similar historical work orders.

[0009] Specifically, the work order data collection is used to collect work order data from PMS, user acquisition, and marketing systems.

[0010] Specifically, the types of work orders include: meter reading and billing, business expansion and installation, power outage repair, power supply quality, service complaints, power supply safety, power grid construction, etc., totaling 28 categories.

[0011] Furthermore, step 2 includes:

[0012] Step 2.1: Construct a feature vocabulary;

[0013] Step 2.2: Identify the feature words of the current work order data through the feature word library.

[0014] Furthermore, the feature word library is obtained by organizing commonly used feature words from historical work order data, and combining each feature word into a tag group based on the frequency of its occurrence in various types of work orders; step 2.1 specifically involves:

[0015] Step 2.11: Extract the characteristic words A = [A0, A1, ..., A] from historical work orders. i A i Let A represent the i-th feature word, and let A represent the combined feature words.

[0016] Step 2.12: Calculate the number of times each feature word appears in each type of work order data, C = [C0, C1, ..., C]. 28 ], where C represents the combination of the frequency of occurrence of the feature words, and the work order types include 28 categories;

[0017] Step 2.13: Assign a label group to each feature word, in the following format:

[0018]

[0019] Among them, C j This indicates the number of times the feature word appears in the j-th type of work order.

[0020] Furthermore, step 2.2 involves classifying the work order data using a pre-trained deep neural network. Specifically, step 2.2 includes:

[0021] Step 2.21: Use a Chinese word segmentation tool to segment the text content in the work order data, and remove the numbers, punctuation, special symbols and stop words after segmentation;

[0022] Step 2.22: Match the obtained segmented content one by one with the feature words in the feature word library;

[0023] Step 2.23: If the same feature word is matched in the feature word library, obtain the label group of the current feature word in the feature word library; otherwise, discard the current feature word.

[0024] Step 2.24: Combine the obtained tag groups into a 28×28 matrix, padding any insufficient parts with 0 and discarding any excess parts;

[0025] Step 2.25: Input the matrix into the pre-trained deep convolutional neural network to perform work order type recognition.

[0026] Furthermore, the trained deep convolutional neural network was obtained by using transfer learning to change the final number of classifications in the LeNet-5 network to 28.

[0027] Furthermore, step 3 involves calling the corresponding reply template and reply example based on the work order type, and performing a similarity analysis of the work orders from aspects such as user, community, station area, line, time, and request content to find similar historical work orders; step 3 specifically includes:

[0028] Step 3.1: Find the work order data of this type from the historical work orders, and give the corresponding Type I similarity coefficient for these work order data according to the priority order of user, community, substation area and line;

[0029] Step 3.2: If the work order data belongs to the same user, the similarity coefficient is 1; if the work order data belongs to the same community but not the same user, the similarity coefficient is 0.9, and so on. The similarity coefficient for the same transformer area is 0.8, and the similarity coefficient for the same line is 0.7; if the data does not meet the requirements of the same user, community, transformer area, or line, the similarity coefficient is 0.6.

[0030] Step 3.3: After giving the Type I similarity coefficient, give the corresponding Type II similarity coefficient for the historical work order data based on time. If the time difference between the historical work order data and the current work order data is no more than 1 hour, the similarity coefficient is 1; if the time difference is greater than 1 hour but no more than 2 hours, the similarity coefficient is 0.9; and so on, no more than 3 hours, the similarity coefficient is 0.8; no more than 4 hours, the similarity coefficient is 0.7; no more than 5 hours, the similarity coefficient is 0.6; and the similarity coefficient for greater than 5 hours is uniformly 0.5.

[0031] Step 3.4: Based on the content of the request, generate the corresponding Type III similarity coefficient for the historical work order data. Use a Chinese word segmentation tool to obtain the word segmentation data M and N of the request content in the work order data and historical work order data respectively. Calculate the Type III similarity coefficient of the historical work order data using the following formula:

[0032]

[0033] Step 3.5: Calculate the overall similarity coefficient of historical work order data:

[0034] J = 0.3 * I + 0.2 * II + 0.5 * III

[0035] The historical work orders with the highest overall similarity coefficient are the most similar ones.

[0036] Furthermore, the response method also includes the following steps:

[0037] Step 4: Assist in reviewing the response content of the work order; specifically: review the logical errors in the response content, including: time logic errors, incomplete information, inconsistent response content, and non-standard response.

[0038] Furthermore, the response method also includes the following steps:

[0039] Step 5: Remind the user that there is an error in the work order reply; specifically: the reminder module sends reminders to the user based on the review results, including: reminders for time logic errors, incomplete information, non-corresponding reply content, and non-standard reply.

[0040] Secondly, a power supply work order rapid response system based on similarity analysis is provided, employing the aforementioned power supply work order rapid response method based on similarity analysis. The response system includes:

[0041] The data acquisition module is used to collect work order data;

[0042] The identification module is used to identify the work order type in the work order data;

[0043] The calling module is used to call the work order template for reply based on the work order type, and to perform similarity analysis on the work order data to find replies to similar historical work orders.

[0044] Furthermore, the response system also includes:

[0045] The review module is used to assist in reviewing the response content of work orders;

[0046] The reminder module is used to notify users that their work order responses are incorrect.

[0047] The present invention has the following beneficial effects:

[0048] I. This invention can solve the problem of low efficiency in manual work order responses; this invention enables the system to automatically classify work orders and provide corresponding work order response templates and excellent examples according to different work order types and requests. It also performs similarity analysis on work orders from aspects such as user, community, station area, line, time, and request content to quickly find similar historical work orders for user reference, thereby achieving rapid response;

[0049] Second, this invention enables the collection, identification, review, and notification of work orders. It allows the system to automatically review logical errors in the response content, such as time logic errors, incomplete information, inconsistent response content, and non-standard responses, and automatically reject them. This improves the quality of work order responses and the first-time pass rate, enabling efficient and rapid response to user requests and enhancing the service quality and efficiency of power grid work order information responses. Attached Figure Description

[0050] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;

[0051] Figure 2 This is a flowchart of a method for identifying work order types in an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] Please refer to Figure 1 and Figure 2 This invention is a rapid response system for power supply work orders based on similarity analysis, comprising:

[0054] The data acquisition module is used to collect work order data;

[0055] The identification module is used to identify the work order type in the work order data;

[0056] The calling module is used to call the work order template for reply based on the work order type, and to perform similarity analysis on the work order data to find replies to similar historical work orders;

[0057] The review module is used to assist in reviewing the response content of work orders;

[0058] The reminder module is used to notify users that their work order responses are incorrect.

[0059] This invention also provides a rapid response method for power supply work orders based on similarity analysis. Using the aforementioned response system, the method includes the following steps:

[0060] Step 1: The data acquisition module collects work order data; in this embodiment, the data acquisition module is responsible for collecting work order data from the PMS, user acquisition, and marketing systems;

[0061] Step 2: The identification module is used to identify the work order type in the work order data; in this embodiment, the work order types include: meter reading and billing, business expansion application, power outage emergency repair, power supply quality, service complaint, power supply safety, power grid construction, etc., totaling 28 categories; Step 2 specifically involves:

[0062] Step 2.1: Construct a feature word library; the feature word library is obtained by organizing commonly used feature words from historical work order data, and combining each feature word into a tag group based on the frequency of its occurrence in various types of work orders; in this embodiment, step 2.1 specifically involves:

[0063] Step 2.11: Extract the characteristic words A = [A0, A1, ..., A] from historical work orders. i A i Let A represent the i-th feature word, and let A represent the combined feature words.

[0064] Step 2.12: Calculate the number of times each feature word appears in each type of work order data, C = [C0, C1, ..., C]. 28 ], where C represents the combination of the frequency of occurrence of the feature words, and the work order types include 28 categories;

[0065] Step 2.13: Assign a label group to each feature word, in the following format:

[0066]

[0067] Among them, C j This indicates the number of times the feature word appears in the j-th type of work order.

[0068] Step 2.2: Identify the feature words of the current work order data using the feature word library; classify the work order data using the pre-trained deep neural network, see [link / reference]. Figure 2 Step 2.2 specifically involves:

[0069] Step 2.21: Use a Chinese word segmentation tool to segment the text content in the work order data, and remove the segmented numbers, punctuation, special symbols, and stop words;

[0070] Step 2.22: Match the obtained segmented content one by one with the feature words in the feature word library;

[0071] Step 2.23: If the same feature word is matched in the feature word library, obtain the label group of the current feature word in the feature word library; otherwise, discard the current feature word.

[0072] Step 2.24: Combine the obtained tag groups into a 28×28 matrix, padding any insufficient parts with 0 and discarding any excess parts;

[0073] Step 2.25: Input the matrix into the pre-trained deep convolutional neural network to perform work order type recognition.

[0074] In this embodiment, the trained deep convolutional neural network was obtained by using transfer learning to change the final number of classifications in the LeNet-5 network to 28.

[0075] Step 3: The calling module calls the work order template for replying based on the work order type, and performs similarity analysis on the work order data to find replies to similar historical work orders;

[0076] The calling module includes a template library and an example library. Based on the work order type, the calling module calls the corresponding reply template and reply example from the template library and the example library, and performs a similarity analysis of the work orders based on factors such as user, community, station area, line, time, and request content to quickly find similar historical work orders.

[0077] Step 3 specifically includes:

[0078] Step 3.1: Find the work order data of this type from the historical work orders, and give the corresponding Type I similarity coefficient for these work order data according to the priority order of user, community, substation area and line;

[0079] Step 3.2: If the work order data belongs to the same user, the similarity coefficient is 1; if the work order data belongs to the same community but not the same user, the similarity coefficient is 0.9, and so on. The similarity coefficient for the same transformer area is 0.8, and the similarity coefficient for the same line is 0.7; if the data does not meet the requirements of the same user, community, transformer area, or line, the similarity coefficient is 0.6.

[0080] Step 3.3: After giving the Type I similarity coefficient, give the corresponding Type II similarity coefficient for the historical work order data based on time. If the time difference between the historical work order data and the current work order data is no more than 1 hour, the similarity coefficient is 1; if the time difference is greater than 1 hour but no more than 2 hours, the similarity coefficient is 0.9; and so on, no more than 3 hours, the similarity coefficient is 0.8; no more than 4 hours, the similarity coefficient is 0.7; no more than 5 hours, the similarity coefficient is 0.6; and the similarity coefficient for greater than 5 hours is uniformly 0.5.

[0081] Step 3.4: Based on the content of the request, generate the corresponding Type III similarity coefficient for the historical work order data. Use a Chinese word segmentation tool to obtain the word segmentation data M and N of the request content in the work order data and historical work order data respectively. Calculate the Type III similarity coefficient of the historical work order data using the following formula:

[0082]

[0083] Step 3.5: Calculate the overall similarity coefficient of historical work order data:

[0084] J = 0.3 * I + 0.2 * II + 0.5 * III

[0085] The historical work orders with the highest overall similarity coefficient are the most similar ones.

[0086] Step 4: The review module assists in reviewing the response content of the work order; in this embodiment, the review module automatically reviews the logical errors in the response content, including: time logic errors, incomplete information, non-corresponding response content, and non-standard response, etc.

[0087] Step 5: The reminder module notifies the user that the work order reply content is incorrect; In this embodiment, the reminder module will send reminders to the user based on the review results, including: reminders for time logic errors, incomplete information, non-corresponding reply content, and non-standard reply, etc.

[0088] All parts not covered in this invention are the same as or implemented using existing technologies.

[0089] The above description, in conjunction with specific embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for rapid response to power supply work orders based on similarity analysis, characterized in that: include Step 1: Collect work order data; Step 2: Identify the work order type in the work order data, specifically including: Step 2.1: Construct a feature vocabulary; The feature word library is compiled from commonly used feature words in historical work order data, and each feature word is combined into a tag group based on its frequency of occurrence in various types of work orders. Specifically: Step 2.11: Extract the characteristic words A = [A0, A1, ..., A] from historical work orders. i A i Let A represent the i-th feature word, and let A represent the combined feature words. Step 2.12: Calculate the number of times each feature word appears in each type of work order data, C = [C0, C1, ..., C]. 28 ], where C represents the combination of the frequency of occurrence of the feature words, and the work order types include 28 categories; Step 2.13: Assign a label group to each feature word, in the following format: Among them, C j This indicates the number of times the feature word appears in the j-th type of work order; Step 2.2: Identify the feature words of the current work order data through the feature word library. Specifically, classify the work order data using a pre-trained deep neural network, including: Step 2.21: Use a Chinese word segmentation tool to segment the text content in the work order data, and remove the numbers, punctuation, special symbols and stop words after segmentation; Step 2.22: Match the obtained segmented content one by one with the feature words in the feature word library; Step 2.23: If the same feature word is matched in the feature word library, obtain the label group of the current feature word in the feature word library; otherwise, discard the current feature word. Step 2.24: Combine the obtained tag groups into a 28×28 matrix, padding any insufficient parts with 0 and discarding any excess parts; Step 2.25: Input the matrix into the pre-trained deep convolutional neural network to perform work order type recognition; The trained deep convolutional neural network was obtained by using transfer learning to change the final number of classifications in the LeNet-5 network to 28. Step 3: Based on the work order type, call the work order template for replying, and perform similarity analysis on the work order data to find replies to similar historical work orders.

2. The rapid response method for power supply work orders based on similarity analysis according to claim 1, characterized in that: Step 3 involves calling the corresponding reply template and reply example according to the work order type, and performing a similarity analysis of the work orders from the aspects of user, community, station area, line, time, and request content to find similar historical work orders; Step 3 specifically includes: Step 3.1: Find the work order data of this type from the historical work orders, and give the corresponding Type I similarity coefficient for these work order data according to the priority order of user, community, substation area and line; Step 3.2: If the work order data belongs to the same user, the similarity coefficient is 1; if the work order data belongs to the same community but not the same user, the similarity coefficient is 0.9, and so on. The similarity coefficient for the same transformer area is 0.8, and the similarity coefficient for the same line is 0.7; if the data does not meet the requirements of the same user, community, transformer area, or line, the similarity coefficient is 0.

6. Step 3.3: After providing the Type I similarity coefficient, provide the corresponding Type II similarity coefficient for the historical work order data based on time. If the time difference between the historical work order data and the current work order data is no more than 1 hour, the similarity coefficient is 1; if the time difference is greater than 1 hour but no more than 2 hours, the similarity coefficient is 0.9; and so on, no more than 3 hours, the similarity coefficient is 0.8; no more than 4 hours, the similarity coefficient is 0.7; no more than 5 hours, the similarity coefficient is 0.6; and the similarity coefficient for greater than 5 hours is uniformly 0.

5. Step 3.4: Provide the corresponding Type III similarity coefficient for the historical work order data based on the content of the request. Use a Chinese word segmentation tool to obtain the word segmentation data M and N of the request content in the work order data and historical work order data respectively, and calculate the Type III similarity coefficient of the historical work order data using the following formula: Step 3.5: Calculate the overall similarity coefficient of historical work order data: J＝0.3*Ⅰ+0.2*Ⅱ+0.5*Ⅲ The historical work orders with the highest overall similarity coefficient are the most similar ones.

3. The rapid response method for power supply work orders based on similarity analysis according to claim 1, characterized in that: The reply method also includes the following steps: Step 4: Assist in reviewing the response content of the work order; specifically: review the logical errors in the response content, including: time logic errors, incomplete information, inconsistent response content, and non-standard response.

4. The rapid response method for power supply work orders based on similarity analysis according to claim 3, characterized in that: The reply method also includes the following steps: Step 5: Remind the user that there is an error in the work order reply; specifically: the reminder module sends reminders to the user based on the review results, including: reminders for time logic errors, incomplete information, non-corresponding reply content, and non-standard reply.

5. A rapid response system for power supply work orders based on similarity analysis, characterized in that: The power supply work order rapid response method based on similarity analysis as described in any one of claims 1-4, wherein the response system comprises: The data acquisition module is used to collect work order data; The identification module is used to identify the work order type in the work order data; The calling module is used to call the work order template for reply based on the work order type, and to perform similarity analysis on the work order data to find replies to similar historical work orders.

6. The power supply work order rapid response system based on similarity analysis according to claim 5, characterized in that: The response system also includes: The review module is used to assist in reviewing the response content of work orders; The reminder module is used to notify users that their work order responses are incorrect.

Images