Power marketing audit work order review method and device

Abstract

The present application provides a method and device for reviewing an electric power marketing audit work order, the method comprising: receiving an audit request, obtaining an audit work order and actual abnormality cause information corresponding to the audit request, the audit work order comprising: an audit subject and an abnormality cause description text; determining an abnormality cause identification model corresponding to the audit work order based on the audit subject, the abnormality cause identification model being obtained by pre-training a classification model based on a batch of historical audit work orders corresponding to the audit subject and their respective corresponding standard abnormality cause identification results; determining an abnormality cause identification result corresponding to the audit request based on the abnormality cause description text and the determined abnormality cause identification model; completing the audit of the electric power marketing audit work order corresponding to the audit request based on the actual abnormality cause information and the abnormality cause identification result. The present application can improve the efficiency and accuracy of the audit of electric power marketing audit work orders, thereby improving the reliability of the electric power marketing audit process.

Description

Technical Field

[0001] The present application relates to the technical field of power marketing auditing, and in particular to a method and device for reviewing power marketing auditing work orders. Background Art

[0002] During the power audit business process, the superior auditor generally issues a power marketing audit work order to the subordinate auditor. The subordinate auditor verifies the cause of the problem, fills out the power marketing audit work order and replies. The superior auditor reviews whether the replied power marketing audit work order is correct based on the actual situation. The content of the replied work order is a specific description of the abnormal situation generated by the business. The review of the power marketing audit work order is an effective means to reduce errors and risks in the power marketing audit process.

[0003] With the continuous adjustments of various business segments and the optimization of marketing auditing, the current method of reviewing work orders for power marketing auditing is no longer sufficient. This is particularly evident in the fact that a large number of audit work orders require manual review and response to complete the accuracy and compliance review and archiving of the work orders. This requires a high level of professional knowledge and experience from the reviewers. Due to the uneven experience of the current staff, this problem has hampered the overall efficiency of marketing auditing. Summary of the Invention

[0004] In response to the problems in the existing technology, this application proposes a method and device for reviewing power marketing audit work orders, which can improve the efficiency and accuracy of power marketing audit work order review, and thus improve the reliability of the power marketing audit process.

[0005] In order to solve the above technical problems, this application provides the following technical solutions:

[0006] In a first aspect, the present application provides a method for reviewing power marketing audit work orders, comprising:

[0007] Receive an audit request, and obtain an audit work order and actual abnormality reason information corresponding to the audit request, wherein the audit work order includes: an audit subject and an explanation text of the abnormality reason;

[0008] Determine, based on the audit theme, a preset abnormality cause identification model corresponding to the audit work order, wherein the preset abnormality cause identification model is obtained by pre-training a classification model based on a batch of historical audit work orders corresponding to the audit theme and their respective corresponding standard abnormality cause identification results;

[0009] Determine the abnormality cause identification result corresponding to the audit request based on the abnormality cause description text and the determined abnormality cause identification model;

[0010] According to the actual abnormal cause information and the abnormal cause identification result, the power marketing audit work order corresponding to the audit request is completed.

[0011] In one embodiment of the present application, before determining the preset abnormal cause identification model corresponding to the audit work order according to the audit subject, the method further includes:

[0012] Obtain sample data sets corresponding to various audit topics. Each sample data set includes: batch historical audit work orders for a unique audit topic and their corresponding standard abnormality cause identification results;

[0013] The classification model is trained using batch historical audit work orders of each audit topic and their corresponding standard abnormal cause identification results to obtain the abnormal cause identification model corresponding to the audit topic.

[0014] Furthermore, the preset abnormal cause identification model includes: a conditional random field layer, a word embedding layer, and a semantic analysis layer based on an attention mechanism.

[0015] Furthermore, completing the review of the power marketing audit work order corresponding to the audit request based on the actual abnormality cause information and the abnormality cause identification result includes:

[0016] Determine whether the actual abnormal cause information and the abnormal cause identification result are the same. If so, determine that the audit work order has passed the review; otherwise, determine that the audit work order has failed the review.

[0017] Furthermore, determining the abnormal cause identification result corresponding to the audit request based on the abnormal cause description text and the determined abnormal cause identification model includes:

[0018] Applying the word embedding layer in the preset abnormal cause identification model to obtain multiple word vectors in the abnormal cause description text;

[0019] Obtaining multiple annotation sequences through the semantic analysis layer and each word vector in the preset abnormal cause identification model;

[0020] According to the conditional random field layer and each annotation sequence in the preset abnormal cause identification model, a globally optimal annotation sequence is obtained and determined as the abnormal cause identification result corresponding to the audit request.

[0021] In a second aspect, the present application provides a power marketing audit work order review device, comprising:

[0022] A receiving module is used to receive an audit request and obtain an audit work order and actual abnormality reason information corresponding to the audit request. The audit work order includes: an audit subject and an explanation text of the abnormality reason;

[0023] A model determination module is used to determine a preset abnormality cause identification model corresponding to the audit work order based on the audit theme, wherein the preset abnormality cause identification model is obtained by pre-training a classification model based on a batch of historical audit work orders corresponding to the audit theme and their corresponding standard abnormality cause identification results;

[0024] a determination module, configured to determine an abnormality cause identification result corresponding to the audit request based on the abnormality cause description text and the determined abnormality cause identification model;

[0025] The audit module is used to complete the audit of the power marketing audit work order corresponding to the audit request based on the actual abnormal cause information and the abnormal cause identification result.

[0026] Furthermore, the power marketing audit work order review device further includes:

[0027] The historical data acquisition module is used to obtain sample data sets corresponding to various audit topics. Each sample data set includes: a batch of historical audit work orders for a unique audit topic and their corresponding standard abnormality cause identification results;

[0028] The training module is used to train the classification model using batch historical audit work orders of each audit topic and their corresponding standard abnormal cause identification results to obtain the abnormal cause identification model corresponding to the audit topic.

[0029] Furthermore, the preset abnormal cause identification model includes: a conditional random field layer, a word embedding layer, and a semantic analysis layer based on an attention mechanism.

[0030] Furthermore, the audit module includes:

[0031] The judgment unit is used to judge whether the actual abnormal cause information and the abnormal cause identification result are the same. If so, it is determined that the audit work order has passed the review; otherwise, it is determined that the audit work order has failed the review.

[0032] Furthermore, the determining module includes:

[0033] A word vector acquisition unit is used to apply the word embedding layer in the preset abnormality cause identification model to obtain multiple word vectors in the abnormality cause description text;

[0034] A standard sequence acquisition unit is used to obtain multiple annotation sequences through the semantic analysis layer and each word vector in the preset abnormal cause identification model;

[0035] The determination unit is used to obtain a globally optimal annotation sequence according to the conditional random field layer and each annotation sequence in the preset abnormal cause identification model and determine it as the abnormal cause identification result corresponding to the audit request.

[0036] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the power marketing audit work order review method when executing the program.

[0037] In a fourth aspect, the present application provides a computer-readable storage medium having computer instructions stored thereon, which, when executed, implement the power marketing audit work order review method.

[0038] As can be seen from the above technical solution, the present application provides a method and device for reviewing power marketing audit work orders. The method includes: receiving an audit request, obtaining an audit work order and actual abnormality cause information corresponding to the audit request, the audit work order including: an audit subject and an abnormality cause description text; determining a preset abnormality cause identification model corresponding to the audit work order based on the audit subject, the preset abnormality cause identification model being obtained by pre-training a classification model based on a batch of historical audit work orders corresponding to the audit subject and their respective corresponding standard abnormality cause identification results; determining an abnormality cause identification result corresponding to the audit request based on the abnormality cause description text and the determined abnormality cause identification model; completing the audit of the power marketing audit work order corresponding to the audit request based on the actual abnormality cause information and the abnormality cause identification result, which can improve the efficiency and accuracy of the audit of the power marketing audit work order, thereby improving the reliability of the power marketing audit process, realizing automatic identification of the cause of the work order problem, assisting the work order auditor to quickly complete the audit filing, saving labor costs, and promptly resolving risks and problems in the power marketing audit process. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0040] Figure 1 This is a flowchart of a method for reviewing a power marketing audit work order in an embodiment of the present application;

[0041] Figure 2 201 and 202 of the method for reviewing a power marketing audit work order in an embodiment of the present application;

[0042] Figure 3 This is a schematic diagram of the structure of the power marketing audit work order review device in an embodiment of the present application;

[0043] Figure 4 This is a structural diagram of a power marketing audit work order review device in another embodiment of the present application;

[0044] Figure 5 This is a schematic block diagram of the system structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0045] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0046] The existing technology also has the problem that the power marketing audit department handles a large number of audit topics every month, and each audit topic has a different number of audit work orders. Manually reviewing the work orders is costly and inefficient.

[0047] In order to solve the problems existing in the above-mentioned prior art, and considering that with the rapid development of parallel computing, deep learning technology has made great progress in recent years, this application proposes a method and device for reviewing power marketing audit work orders. Under the premise of known data classification results, a data model is generated according to the classification results of pre-data, and the pre-defined model classification results are obtained by adjusting the parameters and weight values ​​of the model. It is possible to realize the intelligent review of power marketing audit work orders based on deep learning technology. During the audit work order review stage, the abnormality recognition and classification model is called to automatically read the work order reply content, realize the automatic identification of the cause of the work order reply problem, assist the work order reviewer to quickly complete the review and archiving, realize the abnormal cause review and traceability analysis of the power marketing audit work order with intelligent assistance, and effectively resolve the contradiction between the low quality and efficiency of the audit and the continuous improvement of the audit requirements.

[0048] Based on this, in order to improve the efficiency and accuracy of the review of power marketing audit work orders, and thereby improve the reliability of the power marketing audit process, an embodiment of the present application provides a power marketing audit work order review device, which can be a server or a client device. The client device can include a smartphone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, and a smart wearable device. Among them, the smart wearable device can include smart glasses, smart watches, and smart bracelets.

[0049] In practical applications, the portion of the power marketing audit work order review process that is performed can be performed on the server side as described above, or all operations can be completed on the client device. The specific selection can be based on the processing capabilities of the client device and the limitations of the user's usage scenario. This application does not impose any restrictions on this. If all operations are completed on the client device, the client device may also include a processor.

[0050] The client device may include a communication module (i.e., a communication unit) that can establish a communication connection with a remote server to implement data transmission with the server. The server may include a server on the task scheduling center side, and in other implementation scenarios, may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a server structure of a distributed device.

[0051] The server and the client device may communicate using any suitable network protocol, including network protocols not yet developed on the date of filing this application. Examples of such network protocols include TCP / IP, UDP / IP, HTTP, and HTTPS. Furthermore, examples of such network protocols include RPC (Remote Procedure Call Protocol) and REST (Representational State Transfer) protocols, which are used on top of the aforementioned protocols.

[0052] It should be noted that the power marketing audit work order review method and device disclosed in this application can be used in the field of power marketing audit technology, and can also be used in any field other than the field of power marketing audit technology. The application field of the power marketing audit work order review method and device disclosed in this application is not limited.

[0053] The details are described in the following embodiments.

[0054] In order to improve the efficiency and accuracy of the power marketing audit work order review, and thus improve the reliability of the power marketing audit process, this embodiment provides a power marketing audit work order review method, the execution subject of which is a power marketing audit work order review device, and the power marketing audit work order review device includes but is not limited to a server, such as Figure 1 As shown, the method specifically includes the following contents:

[0055] Step 101: Receive an audit request, and obtain an audit work order and actual abnormality reason information corresponding to the audit request. The audit work order includes: an audit subject and an abnormality reason description text.

[0056] Specifically, an audit request sent by an auditor from the front end can be received, and the audit request can include: an audit work order unique identifier, which is used to distinguish different audit work orders, and can be an audit work order code or an abnormality record based on users; the audit work order, that is, the power marketing audit work order, can also include an audit work order unique identifier; the actual abnormal cause information can be the real abnormal cause information corresponding to the audit work order, which can be selected by the auditor from multiple standard abnormal cause identification results after on-site verification, and the corresponding relationship between the actual abnormal cause information and the audit work order unique identifier can be pre-stored locally in the power marketing audit work order audit device; the audit work order in this embodiment can be an audit work order with a filled-in abnormal cause description text, which can be used by the power marketing The auditor uploads it in advance from the client; for example, the audit topic can be: distributed photovoltaic power anomalies; the anomaly reason explanation text is a text that specifically describes the reason for the business anomaly, for example, "After verification, our company has implemented closed-loop management of meter reading since January 2020. Due to xxx system reasons, some photovoltaic power generation users cannot issue electricity bills normally. This problem has been reported to the project team many times. After the project team optimized the system, it gradually solved this problem. By June of this year, the problem has basically no longer occurred. Our company is now making up for the photovoltaic electricity charges that have not been issued normally, so there is a problem that the household's power generation in July is far greater than the normal theoretical value. Since this electricity is the cumulative electricity that has not been issued normally since January, the user does not have an overcapacity phenomenon."

[0057] Step 102: Based on the audit subject, determine the preset abnormal cause identification model corresponding to the audit work order. The preset abnormal cause identification model is obtained by pre-training the classification model based on the batch of historical audit work orders corresponding to the audit subject and their respective corresponding standard abnormal cause identification results.

[0058] Specifically, the audit subject and the abnormal cause identification model can be in a one-to-one correspondence.

[0059] Step 103: Determine the abnormal cause identification result corresponding to the audit request based on the abnormal cause description text and the determined abnormal cause identification model.

[0060] Specifically, the abnormal cause description text can be input into the preset abnormal cause identification model, and the output result of the abnormal cause identification model can be determined as the abnormal cause identification result corresponding to the audit request; for example, the abnormal cause identification result can be one of: system abnormality, meter error, combined electricity consumption for multiple months, and unauthorized electricity use.

[0061] Specifically, the classification model may be a Bert+BiLSTM+CRF model.

[0062] Step 104: Based on the actual abnormality cause information and the abnormality cause identification result, complete the power marketing audit work order review corresponding to the audit request.

[0063] Specifically, it can be determined whether the actual abnormal cause information and the abnormal cause identification result are the same. If so, the audit work order is determined to have passed the review; otherwise, the audit work order is determined to have failed the review. After the audit work order is determined to have passed the review, the recommended corrective measures corresponding to the abnormal cause identification result can be determined based on the pre-stored correspondence between the abnormal cause identification result and the corrective measures, and output and displayed to improve the efficiency of subsequent corrective actions. For example, if the abnormal cause identification result is "electricity consumption exceeds capacity", the corresponding recommended corrective measures may be: "Treat as a breach of contract, make up the basic electricity fee, and make up the electricity fee for the breach of contract."

[0064] It can be seen from the above description that the power marketing audit work order review method provided in this embodiment can improve the efficiency and accuracy of power marketing audit work order review, and thus can improve the reliability of the power marketing audit process; specifically, in the audit work order review stage, the abnormality recognition model is called, and by identifying the "cause description content" in the work order ("cause description content" can be equivalent to the above-mentioned abnormal cause description text), the abnormal cause identification result is automatically obtained, and the auditor is provided with intelligent assistance for work order analysis and judgment. The auditor can judge whether the audit work order is compliant based on the analysis and judgment results, saving working time for the audit work order auditors and reducing the work pressure of centralized audit. The return receipt of the audit work order can be equivalent to the audit work order with the abnormal cause description text filled in.

[0065] In one example, provincial-level auditors might generate and issue an audit work order, which is then dispatched to prefecture-level auditors. These auditors then dispatch further. District and county-level auditors, based on the expertise covered in the audit work order, will verify the cause of the issue (perhaps by contacting a specific individual, conducting an on-site inspection, or checking system data). They then complete the audit work order and respond accordingly. The completed audit work order typically goes first to the prefecture-level and then to the provincial level, where it undergoes a review process.

[0066] Specifically, in order to improve the visualization of the work order review process, in the process of dispatching audit work orders: all links except the provincial dispatch link can add work order template display and work order processing guidance buttons on the front end; by obtaining database viewing permissions, you can directly view the work order templates of the corresponding topics in the library; the client can obtain the template content fields with valid status under the audit topic in the table according to the audit topic identifier, and the correspondence between the audit topic and the template content fields needs to be determined in advance; when replying to the work order: add the result display of "reason explanation and intelligent identification reason" to the to-do list of the municipal / provincial work order review link; the server opens the work order review table query permission to obtain the information in the table The reason description and intelligent identification of the cause of the problem should be given under the number; when replying to the work order: the result display of "intelligent identification of problem causes and recommended corrective measures" is added to the work order review details interface at the municipal / provincial level: the server opens the work order review form and work order template table query permission; after the audit work order is processed by the district and county personnel, the client needs to send the "audit subject, work order number, and cause of the problem" to the server; the server will perform model analysis on the acquired data, and the analyzed data "audit subject, work order number, intelligent identification of problem causes, and recommended corrective measures" will be stored in the server database. The client queries the server database to ensure that provincial and municipal business personnel can see the relevant fields when they open the work order review page.

[0067] In order to further improve the reliability of the abnormal cause identification model, see Figure 2 In one embodiment of the present application, before step 102, the method further includes:

[0068] Step 201: Obtain sample data sets corresponding to various audit topics. Each sample data set includes: a batch of historical audit work orders for a unique audit topic and their corresponding standard abnormality cause identification results.

[0069] Specifically, the standard abnormality cause identification result may represent the actual monitoring result of manually marked historical audit work orders.

[0070] Step 202: Apply the batch historical audit work orders of each audit topic and their corresponding standard abnormal cause identification results to train the classification model to obtain the abnormal cause identification model corresponding to the audit topic.

[0071] Specifically, the abnormal cause identification model may include: a conditional random field layer, a word embedding layer, and a semantic analysis layer based on an attention mechanism.

[0072] In order to further improve the accuracy of the power marketing audit work order review, the abnormal cause identification result corresponding to the audit request is determined based on the abnormal cause description text and the determined abnormal cause identification model, including:

[0073] Apply the word embedding layer in the preset abnormal cause identification model to obtain multiple word vectors in the abnormal cause description text; obtain multiple annotation sequences through the semantic analysis layer and each word vector in the preset abnormal cause identification model; obtain the globally optimal annotation sequence based on the conditional random field layer and each annotation sequence in the preset abnormal cause identification model and determine it as the abnormal cause identification result corresponding to the audit request.

[0074] Specifically, the BERT layer in the preset abnormal cause identification model can be applied to obtain the word vectors in the abnormal cause description text; through the BiILSTM layer, the context feature information between the word vectors can be obtained to obtain multiple annotation sequences; finally, the CRF layer and multiple annotation sequences can be applied to obtain the globally optimal annotation sequence (which can be equivalent to the abnormal cause identification result).

[0075] To further illustrate this solution, this application provides an application example of a method for reviewing a power marketing audit work order, which is described in detail as follows:

[0076] (1) Constructing an abnormality cause identification model

[0077] The power marketing audit department can handle hundreds of audit topics each month, and each audit topic has a different number of audit work orders. Manually reviewing the work orders is costly and inefficient.

[0078] Since audit work orders are generated according to the "audit topic" dimension, the business and rules corresponding to each audit topic are inconsistent. Therefore, during the model training process, "abnormal cause identification models" are established according to different "audit topics".

[0079] For example, taking the audit theme corresponding to distributed photovoltaic business as the starting point, we sorted out 799 audit work orders and constructed an abnormality cause identification model using the Bert+BiLSTM+CRF combination algorithm. This allows for automatic classification of audit work orders through artificial intelligence technology. The specific steps are as follows:

[0080] 1) Label sorting of work order abnormality reasons

[0081] Using the BRAT text annotation tool and expert experience, we conducted a preliminary review of 799 audit work orders for distributed photovoltaic services, annotating those with clear abnormalities. Examples are as follows:

[0082] Audit Work Order 1: "After verification, the associated customer number is: xxxxx, power generation customer number: xxxx, power generation address: No. xxxxx, Qinhuangdao City, Hebei Province, meter reading section number: xxxxx, power generation: 0, online power: 0, error power: 0. Specific reason: After verification, the circuit breaker at the customer-side inverter is damaged and cannot be closed, resulting in a power generation of 0 kilowatt-hours." Based on expert experience, the abnormal cause for this audit order is [Inverter Fault]. The annotated data is [Inverter, Circuit Breaker Damaged].

[0083] Audit work order 2: The content is "On July 16, 2020, after verification by relevant personnel of the Maolanggou Power Supply Station of the Pingquan Power Supply Company, the distributed distribution box in the user's home tripped. Relevant personnel have been asked to repair it, but it has not been repaired yet. Therefore, no electricity was generated in June, so no rectification is required and there is no abnormality." Based on expert experience, the abnormal cause corresponding to this audit work order is [leakage protection switch tripped], and the marked data is [distributed distribution box tripped].

[0084] Through the above methods, a total of 763 audit work orders were marked, and the abnormal reasons for 36 work orders could not be clearly identified. A total of 20 abnormal reason labels were sorted out, and the specific sorting results are shown in Table 1 below:

[0085] Table 1

[0086] Serial number Abnormal reason label Amount of labeled data (i.e. number of audit work orders) 1 The leakage protection switch trips 149 2 Inverter failure 101 3 User equipment removal 81 4 User equipment is damaged 54 5 The excess capacity is caused by the sunshine 47 6 Wrong wiring causes reverse current 44 7 Electricity was not distributed in time to new users 41 8 Collection anomalies lead to multi-month merging 39 9 Private electricity connection 34 10 File capacity maintenance error 32 11 Wiring error 29 12 Multiple months of electricity consumption combined and issued 22 13 Meter error 19 14 Metering device failure 19 15 User side plate failure 14 16 186 System Abnormality 13 17 Staff operational error 12 18 Energy meter clock problem 8 19 Construction issues 3 20 Switch burnt out 2

[0087] 2) Work order abnormality cause identification model training

[0088] During the training process, the labeled dataset needs to be split into a training dataset and a validation dataset in an 8:2 ratio. The training dataset is used to train the classification model to obtain a work order abnormality cause identification model. The classification model includes: a BERT word embedding layer, a BiLSTM semantic analysis layer, and a CRF layer.

[0089] a) BERT word embedding layer

[0090] BERT is a bidirectional Transformer multi-layer stacked encoder. When pre-training the BERT word embedding layer, two methods can be used: Masked LM and Next Sentence Prediction to capture word-level and sentence-level representations, respectively.

[0091] Considering BERT's powerful information memory and extraction capabilities, this application example can directly set the BERT word embedding layer at the bottom layer to extract contextual text information of the audit work order content. The word embedding can be obtained through BERT pre-training + fine-tuning. The 100-dimensional word vector obtained by the BERT word embedding layer and the 20-dimensional word length feature vector are concatenated to form a 120-dimensional vector and sent to the BiLSTM layer.

[0092] b) BiLSTM semantic analysis layer

[0093] The Bi-directional Long Short-Term Memory (BiLSTM) model is a classic bidirectional recurrent neural network model in the field of natural language processing. It is used in many scenarios, including sequence tagging, named entity recognition, and seq2seq models. BiLSTM consists of a forward LSTM and a backward LSTM, both of which are commonly used to model contextual information in natural language processing tasks.

[0094] The BiLSTM semantic analysis layer can be used to implement the construction of named entity recognition application scenarios. When using the BiLSTM algorithm for model training, the number of hidden layer neurons can be defined as 128; each time, the label with the highest output score is selected as the label of the current word and output in the fully connected layer softmax.

[0095] The attention mechanism can be applied during RNN model training. It is a "short-circuit" mechanism commonly used in natural language processing to model long-range semantic dependencies. It can intuitively indicate the "attention" contribution of each word to the result prediction. The attention mechanism model not only preserves the content structure of the work order text with a hierarchical structure, but also increases contextual understanding at both the word and sentence levels.

[0096] c) CRF layer

[0097] The Conditional Random Field (CRF) layer models the conditional probability distribution of a set of output random variables given a set of input random variables. It is a discriminant probabilistic undirected graphical model that models conditional probability distributions. According to the definition of CRF, a relative sequence is a given observation sequence X and output sequence Y, and the model is described by defining the conditional probability P(Y|X).

[0098] The CRF layer can add some constraints to the final predicted labels to ensure that they are valid. These constraints can be automatically learned by the CRF layer from the training dataset during training. The CRF layer is trained to find the transition matrix. During the training of the CRF layer, the CRF loss function consists of the true path score and the total score of all possible paths. Among all possible paths, the score of the true path should be the highest. The loss function of CRF can be shown as follows:

[0099]

[0100] Among them, P real represents the true path score, P n Indicates the score of the nth path.

[0101] In the training process of this abnormal cause identification model, the goal is to minimize the loss function, so the loss function of CRF becomes the following form:

[0102]

[0103] 3) Evaluation of abnormal cause identification model

[0104] The evaluation results of the abnormal cause identification model training for the distributed photovoltaic audit topic are as follows:

[0105] Accuracy: 92.11%, Precision: 88.82%, Recall: 90.00%, FB1: 89.40.

[0106] 4) Abnormal cause identification model verification

[0107] The validation set was used to validate the abnormal cause identification model for distributed photovoltaic inspection work orders. By analyzing the validation data set, which included 154 work orders with 12 abnormal cause labels, the model validation results met the preset standards. The prediction results for each abnormal cause label are shown in Table 2 below:

[0108] Table 2

[0109] Serial number Abnormal reason label Sample size Accuracy Recall F1 1 Meter error 9 100.00％ 88.89％ 94.12 2 Wrong wiring causes reverse current 10 72.73％ 80.00％ 76.19 3 File capacity maintenance error 8 100.00％ 87.50％ 93.33 4 Multiple months of electricity consumption combined and issued 20 95.24％ 100.00％ 97.56 5 Metering device failure 4 66.67％ 50.00％ 57.14 6 The leakage protection switch trips 30 100.00％ 93.33％ 96.55 7 Inverter failure 23 84.00％ 100.00％ 87.50 8 The excess capacity is caused by the sunshine 9 90.00％ 100.00％ 94.74 9 Private electricity connection 8 66.67％ 75.00％ 70.59 10 Electricity was not distributed in time to new users 6 100.00％ 83.33％ 90.91 11 User equipment removal 14 76.92％ 83.33％ 80.00 12 User equipment is damaged 11 91.67％ 100.00％ 95.65

[0110] (2) Intelligent review of audit work orders

[0111] Based on the abnormal cause identification model, intelligent analysis and judgment of the reply work order is realized. The work order reviewer determines the review and handling opinions of the work order based on the problem cause of the audit work order judged by the abnormal cause identification model, thereby improving the efficiency of audit work order review.

[0112] During the audit work order review phase, the exception cause identification model can be invoked to automatically extract the cause of the work order issue by identifying the "reason explanation" content in the work order, providing intelligent assistance to audit work order reviewers. Auditors can then determine whether the audit work order is compliant based on the analysis results. This feature can save audit work order reviewers time and reduce the workload of centralized work order review.

[0113] In one example, using the existing method to review power marketing audit work orders, more than 3,000 power marketing audit work orders require 4 people to review and complete in 7 days. However, using the power marketing audit work order review method provided by this solution, more than 3,000 power marketing audit work orders only require 4 people to review and complete in 2 days, significantly improving the efficiency of audit work order review.

[0114] From the software level, in order to improve the efficiency and accuracy of the power marketing audit work order review, and thus to improve the reliability of the power marketing audit process, this application provides an embodiment of a power marketing audit work order review device for implementing all or part of the content of the power marketing audit work order review method, see Figure 3 The power marketing audit work order review device specifically includes the following contents:

[0115] The receiving module 31 is used to receive an audit request and obtain an audit work order and actual abnormality reason information corresponding to the audit request. The audit work order includes: an audit subject and an abnormality reason description text;

[0116] A model determination module 32 is configured to determine, based on the audit theme, a preset abnormality cause identification model corresponding to the audit work order, wherein the preset abnormality cause identification model is obtained by pre-training a classification model based on a batch of historical audit work orders corresponding to the audit theme and their corresponding standard abnormality cause identification results;

[0117] A determination module 33 is configured to determine an abnormality cause identification result corresponding to the audit request based on the abnormality cause description text and the determined abnormality cause identification model;

[0118] The audit module 34 is used to complete the audit of the power marketing audit work order corresponding to the audit request based on the actual abnormality cause information and the abnormality cause identification result.

[0119] See also Figure 4 In one embodiment of the present application, the power marketing audit work order review device further includes:

[0120] The historical data acquisition module 41 is used to acquire sample data sets corresponding to various audit topics. Each sample data set includes: a batch of historical audit work orders corresponding to a unique audit topic and their corresponding standard abnormality cause identification results;

[0121] The training module 42 is used to train the classification model using the batch historical audit work orders of each audit topic and their corresponding standard abnormal cause identification results to obtain the abnormal cause identification model corresponding to the audit topic.

[0122] Among them, the abnormal cause identification model may include: a conditional random field layer, a word embedding layer, and a semantic analysis layer based on an attention mechanism.

[0123] In one embodiment of the present application, the audit module includes:

[0124] The judgment unit is used to judge whether the actual abnormal cause information and the abnormal cause identification result are the same. If so, it is determined that the audit work order has passed the review; otherwise, it is determined that the audit work order has failed the review.

[0125] In one embodiment of the present application, the determining module includes:

[0126] A word vector acquisition unit is used to apply the word embedding layer in the preset abnormality cause identification model to obtain multiple word vectors in the abnormality cause description text;

[0127] A standard sequence acquisition unit is used to obtain multiple annotation sequences through the semantic analysis layer and each word vector in the preset abnormal cause identification model;

[0128] The determination unit is used to obtain a globally optimal annotation sequence according to the conditional random field layer and each annotation sequence in the preset abnormal cause identification model and determine it as the abnormal cause identification result corresponding to the audit request.

[0129] The embodiment of the power marketing audit work order review device provided in this specification can be specifically used to execute the processing flow of the embodiment of the above-mentioned power marketing audit work order review method. Its functions will not be repeated here, and you can refer to the detailed description of the embodiment of the above-mentioned power marketing audit work order review method.

[0130] From the above description, it can be seen that the power marketing audit work order review method and device provided in this application can improve the efficiency and accuracy of power marketing audit work order review, thereby improving the reliability of the power marketing audit process, realizing automatic identification of the causes of work order problems, assisting work order reviewers to quickly complete the review and archiving, saving labor costs, and promptly solving the risks and problems existing in the power marketing audit process.

[0131] From a hardware perspective, in order to improve the efficiency and accuracy of the power marketing audit work order review, and thereby improve the reliability of the power marketing audit process, the present application provides an embodiment of an electronic device for implementing all or part of the content of the power marketing audit work order review method. The electronic device specifically includes the following content:

[0132] Processor (processor), memory (memory), communication interface (Communications Interface) and bus; wherein, the processor, memory, and communication interface communicate with each other through the bus; the communication interface is used to realize information transmission between the power marketing audit work order review device and related devices such as user terminals; the electronic device can be a desktop computer, a tablet computer and a mobile terminal, etc., and this embodiment is not limited to this. In this embodiment, the electronic device can be implemented with reference to the embodiment for implementing the power marketing audit work order review method and the embodiment for implementing the power marketing audit work order review device, and their contents are merged here, and repeated parts are not repeated.

[0133] Figure 5 Schematic block diagram of the system structure of the electronic device 9600 according to an embodiment of the present application. Figure 5 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that the Figure 5 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0134] In one or more embodiments of the present application, the power marketing audit work order review function may be integrated into the central processing unit 9100. The central processing unit 9100 may be configured to perform the following control:

[0135] Step 101: Receive an audit request, obtain an audit work order and actual abnormality reason information corresponding to the audit request, and the audit work order includes: audit subject and abnormality reason description text;

[0136] Step 102: Determine a preset abnormality cause identification model corresponding to the audit work order based on the audit theme, wherein the preset abnormality cause identification model is obtained by pre-training a classification model based on a batch of historical audit work orders corresponding to the audit theme and their corresponding standard abnormality cause identification results;

[0137] Step 103: Determine an abnormality cause identification result corresponding to the audit request based on the abnormality cause description text and the determined abnormality cause identification model;

[0138] Step 104: Based on the actual abnormality cause information and the abnormality cause identification result, complete the power marketing audit work order review corresponding to the audit request.

[0139] From the above description, it can be seen that the electronic device provided by the embodiment of the present application can improve the efficiency and accuracy of the review of power marketing audit work orders, thereby improving the reliability of the power marketing audit process.

[0140] In another embodiment, the power marketing audit work order review device can be configured separately from the central processor 9100. For example, the power marketing audit work order review device can be configured as a chip connected to the central processor 9100, and the power marketing audit work order review function is realized through the control of the central processor.

[0141] like Figure 5 As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily have to include Figure 5 In addition, the electronic device 9600 may also include all components shown in Figure 5 For components not shown, reference may be made to the prior art.

[0142] like Figure 5 As shown, the central processing unit 9100 is sometimes also referred to as a controller or operation control, and may include a microprocessor or other processor device and / or logic device. The central processing unit 9100 receives input and controls the operation of various components of the electronic device 9600.

[0143] Memory 9140 can be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It can store the aforementioned failure-related information and also store programs that execute the relevant information. The CPU 9100 can execute the programs stored in memory 9140 to implement information storage or processing.

[0144] The input unit 9120 provides input to the central processing unit 9100. The input unit 9120 may be, for example, a keypad or touch input device. The power supply 9170 is used to provide power to the electronic device 9600. The display 9160 is used to display objects such as images and text. The display may be, for example, an LCD display, but is not limited thereto.

[0145] The memory 9140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), or a SIM card. Alternatively, it may be a memory that retains information even when power is off, can be selectively erased, and is provided with more data. Examples of such memory are sometimes referred to as EPROMs. The memory 9140 may also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 by the central processing unit 9100.

[0146] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various driver programs for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0147] The communication module 9110 is a transmitter / receiver 9110 that transmits and receives signals via an antenna 9111. The communication module (transmitter / receiver) 9110 is coupled to the central processor 9100 to provide input signals and receive output signals, which may be the same as in a conventional mobile communication terminal.

[0148] Based on different communication technologies, multiple communication modules 9110 can be provided in the same electronic device, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby implementing common telecommunication functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Furthermore, the audio processor 9130 is also coupled to the central processing unit 9100, enabling local recording via the microphone 9132 and playback of stored audio via the speaker 9131.

[0149] As can be seen from the above description, the electronic device provided by the embodiments of the present application can improve the efficiency and accuracy of the review of power marketing audit work orders, thereby improving the reliability of the power marketing audit process.

[0150] The embodiments of the present application also provide a computer-readable storage medium capable of implementing all steps of the power marketing audit work order review method in the above-mentioned embodiment. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the computer program implements all steps of the power marketing audit work order review method in the above-mentioned embodiment. For example, when the processor executes the computer program, the following steps are implemented:

[0151] Step 101: Receive an audit request, obtain an audit work order and actual abnormality reason information corresponding to the audit request, and the audit work order includes: audit subject and abnormality reason description text;

[0152] Step 102: Determine a preset abnormality cause identification model corresponding to the audit work order based on the audit theme, wherein the preset abnormality cause identification model is obtained by pre-training a classification model based on a batch of historical audit work orders corresponding to the audit theme and their corresponding standard abnormality cause identification results;

[0153] Step 103: Determine an abnormality cause identification result corresponding to the audit request based on the abnormality cause description text and the determined abnormality cause identification model;

[0154] Step 104: Based on the actual abnormality cause information and the abnormality cause identification result, complete the power marketing audit work order review corresponding to the audit request.

[0155] From the above description, it can be seen that the computer-readable storage medium provided in the embodiment of the present application can improve the efficiency and accuracy of the review of power marketing audit work orders, thereby improving the reliability of the power marketing audit process.

[0156] In this application, the various embodiments of the above method are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. For related parts, please refer to the partial description of the method embodiment.

[0157] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0158] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0159] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0160] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0161] Specific embodiments are used in this application to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for reviewing power marketing audit work orders, characterized in that: include: Receive an audit request, and obtain an audit work order and actual abnormality reason information corresponding to the audit request, wherein the audit work order includes: an audit subject and an explanation text of the abnormality reason; Determine, based on the audit theme, a preset abnormality cause identification model corresponding to the audit work order, wherein the preset abnormality cause identification model is obtained by pre-training a classification model based on a batch of historical audit work orders corresponding to the audit theme and their respective corresponding standard abnormality cause identification results; Determine the abnormality cause identification result corresponding to the audit request based on the abnormality cause description text and the determined abnormality cause identification model; According to the actual abnormality cause information and the abnormality cause identification result, complete the power marketing audit work order review corresponding to the audit request; Determining the abnormal cause identification result corresponding to the audit request based on the abnormal cause description text and the determined abnormal cause identification model includes: Applying the word embedding layer in the preset abnormal cause identification model to obtain multiple word vectors in the abnormal cause description text; Obtaining multiple annotation sequences through the semantic analysis layer and each word vector in the preset abnormal cause identification model; According to the conditional random field layer and each annotation sequence in the preset abnormal cause identification model, a globally optimal annotation sequence is obtained and determined as the abnormal cause identification result corresponding to the audit request.

2. The power marketing audit work order review method according to claim 1, characterized in that: Before determining the preset abnormal cause identification model corresponding to the audit work order according to the audit subject, the method further includes: Obtain sample data sets corresponding to various audit topics. Each sample data set includes: batch historical audit work orders for a unique audit topic and their corresponding standard abnormality cause identification results; The classification model is trained using batch historical audit work orders of each audit topic and their corresponding standard abnormal cause identification results to obtain the abnormal cause identification model corresponding to the audit topic.

3. The power marketing audit work order review method according to claim 1 or 2, characterized in that: The preset abnormal cause identification model includes: a conditional random field layer, a word embedding layer, and a semantic analysis layer based on an attention mechanism.

4. The power marketing audit work order review method according to claim 1, characterized in that: The step of completing the power marketing audit work order review corresponding to the audit request based on the actual abnormality cause information and the abnormality cause identification result includes: Determine whether the actual abnormal cause information and the abnormal cause identification result are the same. If so, determine that the audit work order has passed the review; otherwise, determine that the audit work order has failed the review.

5. A power marketing audit work order review device, characterized in that: include: A receiving module is used to receive an audit request and obtain an audit work order and actual abnormality reason information corresponding to the audit request. The audit work order includes: an audit subject and an explanation text of the abnormality reason; A model determination module is used to determine a preset abnormality cause identification model corresponding to the audit work order based on the audit theme, wherein the preset abnormality cause identification model is obtained by pre-training a classification model based on a batch of historical audit work orders corresponding to the audit theme and their corresponding standard abnormality cause identification results; a determination module, configured to determine an abnormality cause identification result corresponding to the audit request based on the abnormality cause description text and the determined abnormality cause identification model; An audit module, configured to complete the audit of the power marketing audit work order corresponding to the audit request based on the actual abnormality cause information and the abnormality cause identification result; The determining module includes: A word vector acquisition unit is used to apply the word embedding layer in the preset abnormality cause identification model to obtain multiple word vectors in the abnormality cause description text; A standard sequence acquisition unit is used to obtain multiple annotation sequences through the semantic analysis layer and each word vector in the preset abnormal cause identification model; The determination unit is used to obtain a globally optimal annotation sequence according to the conditional random field layer and each annotation sequence in the preset abnormal cause identification model and determine it as the abnormal cause identification result corresponding to the audit request.

6. The power marketing audit work order review device according to claim 5, characterized in that: Also includes: The historical data acquisition module is used to obtain sample data sets corresponding to various audit topics. Each sample data set includes: a batch of historical audit work orders for a unique audit topic and their corresponding standard abnormality cause identification results; The training module is used to train the classification model using batch historical audit work orders of each audit topic and their corresponding standard abnormal cause identification results to obtain the abnormal cause identification model corresponding to the audit topic.

7. The power marketing audit work order review device according to claim 5 or 6, characterized in that: The preset abnormal cause identification model includes: a conditional random field layer, a word embedding layer, and a semantic analysis layer based on an attention mechanism.

8. The power marketing audit work order review device according to claim 5, characterized in that: The audit module includes: The judgment unit is used to judge whether the actual abnormal cause information and the abnormal cause identification result are the same. If so, it is determined that the audit work order has passed the review; otherwise, it is determined that the audit work order has failed the review.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the power marketing audit work order review method described in any one of claims 1 to 4 is implemented.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed, the power marketing audit work order review method described in any one of claims 1 to 4 is implemented.

Images