Scheduling operation ticket checking method based on LSTM deep learning algorithm

Abstract

The invention discloses a scheduling operation ticket checking method based on an LSTM deep learning algorithm, and relates to the technical field of operation ticket checking, and the method comprises the steps: collecting historical operation ticket data and equipment ledger data, simplifying an operation instruction in the historical operation ticket data into a general operation action, and storing the general operation action in the equipment ledger data; transformer substation and interval list arrangement is carried out on equipment ledger data; performing transverse formatting processing on the historical operation ticket data; a long short-term memory network model is constructed based on an operation statement dictionary, operation statements and general operation actions are in one-to-one correspondence, the long short-term memory network model is trained to generate an operation instruction classification model and an instruction logic verification model, and the models are verified through a self-prepared test operation ticket. According to the method, the face information and the upper and lower logic relation of the executed operation ticket are checked, the executed operation ticket is automatically checked and evaluated, experience is summarized, the operation ticket issuing integrity rate and accuracy are improved, the power grid operation risk is reduced, and meanwhile the ticket checking time of a dispatcher is shortened.

Description

Technical Field

[0001] The present invention relates to the technical field of operation ticket verification, and specifically to a scheduling operation ticket verification method based on LSTM deep learning algorithm. Background Art

[0002] Common ticket checking tools use syntax recognition and operation rules to build verification models, which can only verify the correctness of instruction text information, and cannot accurately verify the operation instruction information related to special operation modes and accident handling. However, with the development of power grid scale and the increase of power load, the situation of special operation mode adjustment and accident handling is on the rise, and the number of operation ticket verifications for these operations has also increased sharply. Existing ticket checking tools cannot meet the verification needs of operation tickets for complex operation modes.

[0003] Therefore, it is necessary to propose a method for checking dispatch operation tickets based on the LSTM deep learning algorithm, which uses machine learning to simulate the focus and key content of manual ticket review, and compares the fixed word segments with the dispatch automation system model library. The remaining word segments are extracted by machine learning long-term and short-term memory to achieve importance. Based on machine learning experience, the face information and logical relationships of historical dispatch operation tickets can be checked to achieve automatic verification and evaluation of historical dispatch operation tickets, form a verification experience library, and improve the efficiency of checking historical dispatch operation tickets; and it can further realize online operation ticket verification to assist in improving the correctness and completeness of dispatchers' ticket issuance, reduce the risk of wrong tickets, and thus reduce operational risks. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the present invention provides a scheduling operation ticket verification method based on the LSTM deep learning algorithm, which can solve the problems mentioned in the background technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a dispatching operation ticket verification method based on LSTM deep learning algorithm, comprising: collecting historical operation ticket data and equipment ledger data, simplifying the operation instructions in the historical operation ticket data into generalized operation actions, and organizing the equipment ledger data into substation and interval lists; horizontally formatting the historical operation ticket data; constructing a long short-term memory network model based on an operation statement dictionary, corresponding the operation statements to the generalized operation actions one by one, training the long short-term memory network model to generate an operation instruction classification model and an instruction logic verification model, and verifying the model through a self-prepared test operation ticket.

[0007] As a preferred solution of the dispatching operation ticket verification method based on the LSTM deep learning algorithm described in the present invention, historical operation ticket data and equipment ledger data are collected, the operation instructions in the operation ticket data are simplified into general operation actions, and the equipment ledger data is sorted into substation and interval lists, including the following steps: historical operation ticket data is exported from the dispatching command control system, and the operation instructions in the operation ticket data are manually cleaned by the controller; the substation name, voltage level, line name, equipment number and operation content contained in each operation instruction in the operation ticket data are extracted as a general operation action description; the substation name and its corresponding interval number are exported from the dispatching automation system to form equipment ledger data.

[0008] As a preferred solution of the dispatching operation ticket verification method based on the LSTM deep learning algorithm described in the present invention, wherein: the historical operation ticket data includes substation operation mode adjustment operation instructions, equipment maintenance operation instructions and accident handling operation instructions;

[0009] The controller performs manual cleaning on the historical operation ticket data; the manual cleaning includes the following processing: deleting repeated operation instructions, correcting irregular equipment naming, unifying the description terms of operation actions, correcting equipment number identification, and supplementing missing operation steps;

[0010] Extract the substation name, voltage level, line name, equipment number and operation content contained in each operation instruction in the operation ticket data as a generalized operation action description, specifically decompose each operation instruction in the historical operation ticket data into fixed elements and variable elements, and convert the operation instruction into a generalized operation action description according to a preset conversion rule, wherein the conversion rule is: convert the instruction in the format of "substation name + control and remote control + action type + voltage level + line name + equipment number" into the format of "region name + ground control + switch + control type", and convert the instruction in the format of "substation name + patrol and maintenance center + substation name + voltage level + line name + equipment number + standby status" into the format of "plant station + equipment type + standby status";

[0011] Among them, the fixed elements include substation name, voltage level, line name, and equipment number; the variable elements include operation action type and operation object status.

[0012] As a preferred solution of the scheduling operation ticket verification method based on the LSTM deep learning algorithm described in the present invention, the horizontal formatting processing includes comparing the operation statements with the equipment ledger data for tokenization, adding and marking human error data, and segmenting the processed operation statements to form an operation statement dictionary.

[0013] As a preferred solution of the scheduling operation ticket verification method based on the LSTM deep learning algorithm described in the present invention, wherein: the operation ticket data is horizontally formatted, including the following steps:

[0014] Compare each operation instruction in the operation ticket data with the equipment ledger database, and add marking symbols "[TAR][ / TAR]", "[STEP][ / STEP]", "[SUB]", "[BAY]" to the substation equipment name and number with matching items in the equipment ledger database to form a marked operation instruction;

[0015] Adding human error data to the marked operation instruction, specifically adding typos, adding extra characters, deleting necessary characters in the operation instruction, and adding an error type mark to the human error data;

[0016] The tokenized operation instructions and the operation instructions with error type tags are segmented according to the smallest word segmentation unit to form an operation statement dictionary;

[0017] If the word appearing in the word segmentation result is the equipment name or number in the equipment ledger database, the word is marked as a mandatory matching word.

[0018] As a preferred solution of the scheduling operation ticket verification method based on the LSTM deep learning algorithm described in the present invention, a long short-term memory network model is constructed based on the operation statement dictionary, the operation statements are matched one by one with the generalized operation actions, the long short-term memory network model is trained to generate the operation instruction classification model and the instruction logic verification model, and the model is verified by a self-made test operation ticket, including the following steps:

[0019] Classify the entries in the operation statement dictionary according to the word type, construct a feature vector including a device name sequence, an action description sequence, and a state description sequence, and establish a mapping relationship between the feature vector and the corresponding generalized operation action;

[0020] Constructing a long short-term memory network structure including an input layer, a hidden layer and an output layer, wherein the hidden layer includes three control units, namely a forget gate, an input gate and an output gate, the input layer receives a feature vector, and the output layer outputs a generalized operation action;

[0021] Dividing the feature vector into a training set and a validation set in a ratio of 8:2, using the training set to train a long short-term memory network to generate an operation instruction classification model;

[0022] Using the generalized operation action sequences in the training set to construct an instruction logic rule base, and training and generating an instruction logic verification model based on the rule base;

[0023] A test operation ticket data set is constructed, wherein the test operation ticket data set includes correct operation instructions, incorrect operation instructions and logically incorrect operation instructions, and the test operation ticket data set is input into an operation instruction classification model and an instruction logic verification model for test verification.

[0024] As a preferred solution of the scheduling operation ticket verification method based on the LSTM deep learning algorithm described in the present invention, the construction of the instruction logic rule library includes the following rules:

[0025] If a circuit breaker opening and closing operation instruction is detected, a corresponding ground wire removal instruction must exist before the circuit breaker opening and closing operation instruction;

[0026] If a device switching to maintenance state instruction is detected, there must be an associated power-off instruction before the device switching to maintenance state instruction;

[0027] If a parallel capacitor operation instruction is detected, the parallel capacitor operation instruction must be executed during the off-peak load period;

[0028] If a transformer parallel operation command is detected, the ratio difference of the relevant transformers must be within the specified range;

[0029] An instruction logic verification model is generated based on the rule base training, and the instruction logic verification model marks an operation instruction sequence that does not meet the rule requirements as a logic error.

[0030] To further solve the above technical problems, the present invention provides the following technical solutions: a dispatching operation ticket verification system based on the LSTM deep learning algorithm, comprising: a collection and sorting module, used to collect historical operation ticket data and equipment ledger data, simplify the operation instructions in the operation ticket data into generalized operation actions, and organize the equipment ledger data into substation and interval lists; a processing module, used to perform horizontal formatting on the operation ticket data; a verification module, used to construct a long short-term memory network model based on an operation statement dictionary, correspond the operation statements to the generalized operation actions one by one, train the long short-term memory network model to generate an operation instruction classification model and an instruction logic verification model, and verify the model through a self-prepared test operation ticket.

[0031] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the scheduling operation ticket verification method based on the LSTM deep learning algorithm as described above are implemented.

[0032] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the scheduling operation ticket verification method based on the LSTM deep learning algorithm as described above are implemented.

[0033] Beneficial effects of the present invention: The present invention uses machine learning to simulate the focus and key content of manual ticket review, and makes a strong comparison between fixed word segments and the dispatch automation system model library. The remaining word segments are extracted by machine learning long-term and short-term memory. Based on machine learning experience, the face information and logical relationships of historical dispatch operation tickets are verified, so as to realize automatic verification and evaluation of historical dispatch operation tickets, form a verification experience library, and improve the efficiency of checking historical dispatch operation tickets; and can further realize online operation ticket verification, assist in improving the accuracy and completeness of dispatchers' ticket issuance, reduce the risk of wrong tickets, and thus reduce operational risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0035] Figure 1 This is a schematic diagram of the overall process of a scheduling operation ticket verification method based on the LSTM deep learning algorithm proposed in the present invention;

[0036] Figure 2 A schematic diagram of a table that simplifies each operation instruction into a generalized operation action in a scheduling operation ticket verification method based on an LSTM deep learning algorithm proposed by the present invention;

[0037] Figure 3 This is a diagram of the formatting of operation statements in a method for checking a dispatch operation ticket based on an LSTM deep learning algorithm proposed by the present invention;

[0038] Figure 4 A noise graph is added to the scheduling operation ticket verification method based on the LSTM deep learning algorithm proposed by the present invention;

[0039] Figure 5 A word segmentation graph in a scheduling operation ticket verification method based on an LSTM deep learning algorithm proposed by the present invention;

[0040] Figure 6 A dictionary diagram in a scheduling operation ticket verification method based on LSTM deep learning algorithm proposed by the present invention;

[0041] Figure 7 This is a correlation diagram in a scheduling operation ticket verification method based on LSTM deep learning algorithm proposed by the present invention;

[0042] Figure 8A classification diagram of operation instructions in a scheduling operation ticket verification method based on LSTM deep learning algorithm proposed by the present invention;

[0043] Fig. 9 A logical training diagram for the model in a scheduling operation ticket verification method based on the LSTM deep learning algorithm proposed by the present invention;

[0044] Fig.10 This is a diagram for verifying model availability in a scheduling operation ticket verification method based on the LSTM deep learning algorithm proposed in the present invention. DETAILED DESCRIPTION

[0045] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0046] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0047] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a scheduling operation ticket verification method based on LSTM deep learning algorithm.

[0048] In the related technology, common ticket checking tools use syntax recognition and operation rules to build verification models, which can only verify the correctness of instruction text information, and cannot accurately verify operation instruction information related to special operating modes and accident handling.

[0049] The present application provides an effective solution to the above-mentioned problems. Next, multiple embodiments will be combined to explain in detail how to implement the scheduling operation ticket verification method based on the LSTM deep learning algorithm.

[0050] Figure 1 The overall flow chart of a scheduling operation ticket verification method based on LSTM deep learning algorithm is shown, including:

[0051] S1: Collect historical operation ticket data and equipment ledger data, simplify the operation instructions in the historical operation ticket data into general operation actions, and organize the substation and interval list of the equipment ledger data.

[0052] S1.1: Export 100 historical operation ticket data from the dispatching command control system, and the controller manually cleans the operation instructions in the operation ticket data.

[0053] Specifically, 100 representative historical operation ticket data (about 2,000 operation instructions) were selected from the dispatching command and control system DCCS as the training data set. The historical operation ticket data contains various types of operation instructions such as substation operation mode adjustment, equipment maintenance operation, and accident handling. These operation instructions are manually cleaned by controllers with rich dispatching operation experience. The cleaning content includes: deleting duplicate operation instructions, correcting irregular equipment naming, unifying operation action description terms, correcting incorrect equipment number identification, and supplementing missing key operation steps. Through manual cleaning, the accuracy and standardization of all operation instructions are ensured, and a standard operation instruction data set is established to provide high-quality basic data for subsequent model training.

[0054] S1.2: Extract the substation name, voltage level, line name, equipment number and operation content contained in each operation instruction in the operation ticket data into a general operation action description.

[0055] Specifically, each operation instruction in the historical operation ticket data is decomposed into fixed elements and variable elements. The fixed elements include substation name, voltage level, line name, and equipment number; the variable elements include operation action type and operation object status; the operation instruction is converted into a general operation action description according to the preset conversion rules. The conversion rules are as follows: the instruction in the format of "substation name + control and remote control + action type + voltage level + line name + equipment number" is converted into the format of "region name + ground control + switch + control type", and the instruction in the format of "substation name + patrol and maintenance center + substation name + voltage level + line name + equipment number + standby status" is converted into the format of "plant station + equipment type + standby status".

[0056] In an optional embodiment, the cleaned operation instructions are structured and converted in a generalized manner, and each complete operation instruction is divided into two parts: fixed elements and variable elements. The fixed elements include standardized information such as substation name, voltage level, line name, equipment number, etc.; the variable elements include descriptive information such as operation action type and operation object status. Through the preset conversion rules, complex operation instructions are simplified into a generalized operation action description in a unified format, such as "Zunyi Control Remotely Closes the 35kV Wuni Line 304 Circuit Breaker of the Nigao Substation" is converted to "Zunyi Local Control Switch Control Close", and "Wuchuan Patrol and Maintenance Center Changes the 35kV Wuzhen Line 305 Circuit Breaker of Zhennan Substation from Hot Standby to Cold Standby" is converted to "The Plant Station Changes the Circuit Breaker from Hot Standby to Cold Standby", forming a standardized operation action dictionary to provide a standardized corpus for subsequent model training.

[0057] S1.3: Export the substation name and its corresponding interval number from the dispatching automation system to form equipment ledger data.

[0058] Specifically, the equipment ledger data is exported from the dispatching automation system. The equipment ledger data includes substation information and interval information. The substation information includes standard substation naming, area code, voltage level classification, main transformer capacity, and bus number. The interval information includes interval number, interval type, and operating status. The equipment ledger data is indexed according to the hierarchical relationship between substations and intervals to form an equipment ledger database.

[0059] In an optional embodiment, for all substation equipment within the dispatching range, complete equipment ledger data is exported from the dispatching automation system, including basic information such as standard naming of substations, area codes, voltage level classification, main transformer capacity, number of buses, and detailed information such as number, type, and operating status of all intervals in each substation. This information is sorted in a unified format, a hierarchical equipment information index is established, and a standardized equipment ledger database is formed. This database serves as a benchmark for subsequent operation ticket formatting, is used to verify the accuracy of the equipment information involved in the operation instructions, and provides a standard reference for equipment information for model training.

[0060] Preferably, in the prior art, the data of the dispatch operation ticket often has problems such as mixed formats, inconsistent equipment naming, lengthy and non-standard operation descriptions, etc. At the same time, the equipment ledger data is scattered in different systems and lacks a unified indexing standard, which makes it difficult to establish accurate equipment information comparison when verifying the operation ticket. Through the implementation of step S1, the scattered historical operation ticket data is cleaned and standardized by professional controllers, and a standardized operation action dictionary is established to solve the problem of inconsistent operation instruction expressions; by converting complex operation instructions into generalized operation action expressions, the complexity of subsequent model processing is reduced; at the same time, a hierarchical equipment ledger index is established to provide a unified reference standard for equipment information verification of operation instructions. Compared with the simple text processing in the prior art, this data preprocessing method improves the standardization and availability of data while ensuring data quality, laying the foundation for subsequent deep learning model training.

[0061] S2: Format the historical operation ticket data horizontally.

[0062] Specifically, the horizontal formatting process includes comparing the operation statements with the equipment inventory data for tokenization, adding and marking human error data, and segmenting the processed operation statements to form an operation statement dictionary.

[0063] S2.1: Compare each operation instruction in the operation ticket data with the equipment ledger database, and add marking symbols "[TAR][ / TAR]", "[STEP][ / STEP]", "[SUB]", "[BAY]" to the substation equipment names and numbers that have matching items in the equipment ledger database to form marked operation instructions.

[0064] Specifically, each operation instruction in the operation ticket data is compared with the equipment ledger database one by one, and the comparison process includes the following steps: matching the substation name in the operation instruction with the standard substation name in the equipment ledger database, and adding the marking symbol "[SUB]" to the successfully matched substation name; matching the interval equipment information in the operation instruction with the interval number in the equipment ledger database, and adding the marking symbol "[BAY]" to the successfully matched interval equipment information; adding the marking symbol "[STEP][ / STEP]" to the operation step sequence number in the operation instruction; adding the marking symbol "[TAR][ / TAR]" to the core action information in the operation instruction; the addition of marking symbols is performed in a nested order from the outside to the inside to form a tokenized operation instruction text.

[0065] Preferably, since traditional operation ticket text processing methods often use simple string matching, it is difficult to accurately identify key information in the operation instructions and cannot process different expressions of the same device.

[0066] Therefore, S2.1 achieves precise positioning and structured marking of key information such as substation name, interval equipment, and operating procedures through equipment inventory database comparison and the addition of multi-level marking symbols, providing clear information boundaries for subsequent error identification.

[0067] S2.2: Add human error data to the tokenized operation instructions, including: adding typos, adding extra characters, deleting necessary characters in the operation instructions, and adding error type tags to the human error data.

[0068] Specifically, the tokenized operation instructions are processed for adding artificial error data, which includes the following contents: adding typos in the operation instructions, which include single-character replacement in the substation name, equipment name, and digital number; adding extra characters in the operation instructions, which include repeated punctuation marks, repeated equipment names, and extra space characters; deleting necessary characters in the operation instructions, which include numbers in the equipment number, keywords in the substation name, and core words in the operation action description; adding error type tags to the error data, which include "[ERR_CHAR]" for typos, "[ERR_ADD]" for extra characters, and "[ERR_DEL]" for missing characters.

[0069] The existing operation ticket verification system can only identify fixed-mode errors and lacks the ability to identify the diverse error types that occur in actual work. Therefore, S2.2 systematically adds human errors such as typos, extra characters, and missing characters, and marks the error types to construct training samples covering multiple error patterns, thereby improving the model's ability to identify different types of errors.

[0070] S2.3: Segment the tokenized operation instructions and the operation instructions with error type marks according to the smallest word segmentation unit to form an operation statement dictionary. If the word appearing in the word segmentation result is the equipment name or number in the equipment ledger database, the word is marked as a mandatory matching word.

[0071] Specifically, the tokenized operation instructions and the operation instructions with error type marks are imported into the word segmentation processing module, and the word segmentation processing is performed according to the minimum segmentation unit of the word. The word segmentation processing adopts the following rules: the text segment with the marking symbol is kept as a whole without segmentation; the equipment name and number existing in the equipment ledger database are used as forced matching words, and the forced matching mark "[FORCE]" is added; the operation action description is segmented according to the verb, object, and status word; the word segmentation result is stored in the operation statement dictionary, and each entry in the operation statement dictionary contains three fields: word body, word type, and tag information, where the word body is the actual text obtained by word segmentation, the word type includes the equipment name, number, action description, and status description, and the tag information contains all the tag symbols corresponding to the word.

[0072] The conventional general word segmentation algorithm cannot accurately handle the professional terms in the field of power dispatching, and it is easy to have the problem of professional word segmentation errors. S2.3 ensures the complete recognition of professional terms such as power equipment names and numbers by setting mandatory matching words and professional word segmentation rules. At the same time, by establishing an operation statement dictionary containing word ontology, type and tag information, it provides the model with training corpus with domain characteristics.

[0073] In general, step S2 solves key technical problems such as structured expression, error pattern construction, and professional terminology recognition in the text processing of power dispatching operation tickets through the organic combination of tokenization, error data construction, and professional word segmentation. This method not only improves the accuracy and comprehensiveness of operation instruction processing, but also enhances the model's error recognition ability through refined marking of error types, providing high-quality training data for subsequent deep learning-based verification models, and effectively supporting the intelligent verification of power dispatching operation tickets.

[0074] S3: Construct a long short-term memory network model based on the operation statement dictionary, match the operation statements with the general operation actions one by one, train the long short-term memory network model to generate the operation instruction classification model and instruction logic verification model, and verify the model through the self-prepared test operation ticket.

[0075] S3.1: Classify the entries in the operation statement dictionary according to the word type, construct a feature vector containing a device name sequence, an action description sequence, and a state description sequence, and establish a mapping relationship between the feature vector and the corresponding generalized operation action.

[0076] S3.2: Construct a long short-term memory network structure including an input layer, a hidden layer and an output layer. The hidden layer includes three control units: a forget gate, an input gate and an output gate. The input layer receives feature vectors and the output layer outputs generalized operation actions.

[0077] S3.3: Divide the feature vector into a training set and a validation set in a ratio of 8:2, use the training set to train the long short-term memory network, and generate an operation instruction classification model.

[0078] S3.4: Use the generalized operation action sequences in the training set to build an instruction logic rule base, and generate an instruction logic verification model based on the rule base training.

[0079] Specifically, the construction of the instruction logic rule base includes the following rules:

[0080] If a circuit breaker opening and closing operation instruction is detected, a corresponding ground wire removal instruction must exist before the circuit breaker opening and closing operation instruction;

[0081] If a device switching to maintenance state instruction is detected, there must be an associated power-off instruction before the device switching to maintenance state instruction;

[0082] If a parallel capacitor operation command is detected, the parallel capacitor operation command must be executed during the off-peak load period;

[0083] If a transformer parallel operation command is detected, the ratio difference of the relevant transformers must be within the specified range;

[0084] An instruction logic verification model is generated based on rule base training. The instruction logic verification model marks the operation instruction sequence that does not meet the rule requirements as a logic error.

[0085] S3.5: Construct a test operation ticket data set, which includes correct operation instructions, incorrect operation instructions, and logically incorrect operation instructions. Input the test operation ticket data set into the operation instruction classification model and the instruction logic verification model for test verification.

[0086] In summary, the present invention uses machine learning to simulate the focus and key content of manual ticket review, and makes a strong comparison between fixed word segments and the dispatch automation system model library. The remaining word segments are extracted by machine learning long-term and short-term memory to achieve importance. Based on machine learning experience, the face information and logical relationships of historical dispatch operation tickets are verified, so as to realize automatic verification and evaluation of historical dispatch operation tickets, form a verification experience library, and improve the efficiency of checking historical dispatch operation tickets; and it can further realize online operation ticket verification, assist in improving the accuracy and completeness of dispatchers' ticket issuance, reduce the risk of wrong tickets, and thus reduce operational risks.

[0087] Embodiment 2 is an embodiment of the present invention, which provides a dispatching operation ticket verification system based on the LSTM deep learning algorithm, including: a collection and sorting module, which is used to collect historical operation ticket data and equipment ledger data, simplify the operation instructions in the operation ticket data into generalized operation actions, and organize the equipment ledger data into substation and interval lists; a processing module, which is used to perform horizontal formatting on the operation ticket data; a verification module, which is used to construct a long short-term memory network model based on an operation statement dictionary, correspond the operation statements to the generalized operation actions one by one, train the long short-term memory network model to generate an operation instruction classification model and an instruction logic verification model, and verify the model through a self-prepared test operation ticket.

[0088] Embodiment 3 is an embodiment of the present invention, which is different from the previous embodiment in that: if the function is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0089] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0090] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0091] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0092] Example 4, reference Figure 2 to Figure 10 , which is an embodiment of the present invention, provides a scheduling operation ticket verification method based on the LSTM deep learning algorithm. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0093] This method is divided into three parts. The first part is to prepare the data set, including the operation ticket data set and the equipment ledger; the second part is to build a model and perform artificial intelligence verification of the operation ticket, including horizontal verification of the text description and vertical verification of the operation logic:

[0094] (1) Verification based on horizontal text description: By removing the power grid model information, the action description in the operation content is identified to verify the wrong words, multiple words, missing words, etc.:

[0095] (2) Based on the operation logic of the vertical operation ticket, the operation content is first intelligently classified, and then the logical correctness of the instruction sequence in the operation ticket is verified.

[0096] The third part is to use the trained model to check the operation ticket and verify the model availability.

[0097] The specific contents are as follows:

[0098] 1. Dataset preparation:

[0099] (1) Operation ticket data set preparation: Taking Zunyi power grid as an example, 100 historical operation tickets (about 2,000 operation instructions) were exported from the dispatching command and control system (DCCS). Experienced controllers manually cleaned and confirmed them to ensure the correctness of all instructions. At the same time, these more than 2,000 operation instructions were simplified into general operation actions (such as "Zunyi control remotely closes the 304 circuit breaker of the 35kV Wuni line of the Nigao transformer" is summarized as "Zunyi local control switch closes", "Wuchuan patrol and maintenance center changes the 305 circuit breaker of the 35kV Wuzhen line of the Zhennan transformer from hot standby to cold standby" is summarized as "the plant station changes the circuit breaker from hot standby to cold standby"). Figure 2 As shown in yellow.

[0100] (2) Equipment ledger preparation: Export the substation and its bay list from the dispatching automation system to form an equipment ledger;

[0101] 2. Format the operation ticket statement horizontally:

[0102] Take each operation statement as a unit and format it: compare the operation item instruction statement in the operation ticket with the substation and its interval list exported from the dispatching automation system, and format the substation equipment name number that can be found in the substation and interval list in the operation statement (such as the operation statement "Renhuai Patrol and Maintenance Center verifies that the 35kV Zhongtan Line 303 circuit breaker of Tanchang Transformer is in cold standby" is formatted as "[TAR] Renhuai Patrol and Maintenance Center[ / TAR][STEP] verifies that [SUB][BAY] is in cold standby[ / STEP]", "Zunyi Control remotely disconnects the central transformer 35kV Zhongtan Line 303 circuit breaker" is formatted as "[TAR] Zunyi Control[ / TAR][STEP] remotely disconnects the [SUB][BAY] circuit breaker[ / STEP]"). Figure 3 shown.

[0103] (2) Add noise to enhance data and improve generalization ability: The learning model needs to be well fitted on the training data and also needs to be able to perform well on unknown data. In order to make accurate predictions or classifications of new data (wrong votes faced in subsequent verification), experience errors (error points that regulators are prone to making) are artificially added and marked as 0; correct operation statements are marked as 1, such as Figure 4 shown.

[0104] (3) Word segmentation to form an operation statement dictionary: Word segmentation is performed on the horizontally formatted operation item instruction statements, such as Figure 5 As shown; and convert the statements into dictionaries to construct a dictionary of operation tickets for model training, as follows Figure 6 shown.

[0105] 3. Perform vertical logical association processing on the order of operation tickets:

[0106] (1) Take the operation ticket as the unit, and associate the operation statements with the general descriptions one by one: associate "[TAR] Renhuai Patrol and Maintenance Center[ / TAR][STEP] Verify that [SUB][BAY] is in cold standby[ / STEP]" with "Confirm with the plant that the circuit breaker is in cold standby", and associate "[TAR] Zunyi Control[ / TAR][STEP] Remotely disconnect [SUB][BAY] circuit breaker[ / STEP]" with "Zunyi Local Control Switch Control", such as Figure 7 As shown in the figure, associating all 100 operation tickets in this way will form the following logical relationship model: one:

[0108] 1. Verify with the plant and station that the circuit breaker is in cold standby;

[0109] 2. Zunyi local regulation switch control;

[0110] 3. The plant changes the circuit breaker from hot standby to cold standby;

[0111] 4. Install grounding wires at the plant and station;

[0112] 5. Install grounding wires at the plant and station;

[0113] 6. Notify the operation and maintenance unit that the line has been transferred from operation to maintenance. two:

[0115] 1. Open the grounding switch at the plant station;

[0116] 2. Pull open the grounding switch at the plant station;

[0117] 3. The plant changes the circuit breaker from cold standby to hot standby;

[0118] 4. The plant station changes the circuit breaker from cold standby to hot standby. three:

[0120] 1. Verify with the plant and station that the circuit breaker is in cold standby;

[0121] 2. Verify with the plant and station that the circuit breaker is in cold standby;

[0122] 3. Verify with the plant and station that the circuit breaker is in cold standby;

[0123] 4. Verify with the plant and station that the circuit breaker is in cold standby;

[0124] 5. Zunyi local regulation switch control;

[0125] 6. Zunyi ground regulation switch control;

[0126] 7. The plant station changes the busbar from hot standby to cold standby.

[0127] (2) Forming a classification model for operation instructions: Perform model training on the above data, and the training results are as follows: Figure 8 shown.

[0128] (3) By performing logical verification on the instruction description of an operation ticket, the instruction logic verification is formed, such as Fig. 9 shown.

[0129] 4. Verify the model usability: Prepare tickets by yourself (correct tickets, wrong tickets) to verify whether the model training is correct. Fig.10 shown.

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A dispatch operation ticket verification method based on LSTM deep learning algorithm, characterized in that: include: Collect historical operation ticket data and equipment ledger data, simplify the operation instructions in the historical operation ticket data into general operation actions, and sort out the substation and interval list of the equipment ledger data; Performing horizontal formatting processing on the historical operation ticket data; A long short-term memory network model is constructed based on the operation statement dictionary, and the operation statements are matched one by one with the general operation actions. The long short-term memory network model is trained to generate an operation instruction classification model and an instruction logic verification model, and the model is verified through a self-prepared test operation ticket.

2. The method for checking a dispatching operation ticket based on the LSTM deep learning algorithm according to claim 1, characterized in that: Collect historical operation ticket data and equipment ledger data, simplify the operation instructions in the operation ticket data into general operation actions, and sort out the substation and interval list of the equipment ledger data, including the following steps: Export historical operation ticket data from the dispatch command control system, and have the controller manually clean the operation instructions in the operation ticket data; Extracting the substation name, voltage level, line name, equipment number and operation content contained in each operation instruction in the operation ticket data into a generalized operation action description; The substation name and its corresponding interval number are exported from the dispatching automation system to form equipment ledger data.

3. The method for checking a dispatching operation ticket based on the LSTM deep learning algorithm according to claim 2, characterized in that: The historical operation ticket data includes substation operation mode adjustment operation instructions, equipment maintenance operation instructions and accident handling operation instructions; The controller performs manual cleaning on the historical operation ticket data; the manual cleaning includes the following processing: deleting repeated operation instructions, correcting irregular equipment naming, unifying the description terms of operation actions, correcting equipment number identification, and supplementing missing operation steps; Extract the substation name, voltage level, line name, equipment number and operation content contained in each operation instruction in the operation ticket data as a generalized operation action description, specifically decompose each operation instruction in the historical operation ticket data into fixed elements and variable elements, and convert the operation instruction into a generalized operation action description according to a preset conversion rule, wherein the conversion rule is: convert the instruction in the format of "substation name + control and remote control + action type + voltage level + line name + equipment number" into the format of "region name + ground control + switch + control type", and convert the instruction in the format of "substation name + patrol and maintenance center + substation name + voltage level + line name + equipment number + standby status" into the format of "plant station + equipment type + standby status"; Among them, the fixed elements include substation name, voltage level, line name, and equipment number; the variable elements include operation action type and operation object status.

4. The method for checking dispatch operation tickets based on LSTM deep learning algorithm as claimed in claim 3, characterized in that: The horizontal formatting process includes comparing the operation statements with the equipment ledger data to perform tokenization processing, adding and marking human error data, and segmenting the processed operation statements to form an operation statement dictionary.

5. The method for checking dispatch operation tickets based on LSTM deep learning algorithm according to claim 4, characterized in that: The operation ticket data is horizontally formatted, including the following steps: Compare each operation instruction in the operation ticket data with the equipment ledger database, and add marking symbols "[TAR][ / TAR]", "[STEP][ / STEP]", "[SUB]", "[BAY]" to the substation equipment name and number with matching items in the equipment ledger database to form a marked operation instruction; Adding human error data to the marked operation instruction, specifically adding typos, adding extra characters, deleting necessary characters in the operation instruction, and adding an error type mark to the human error data; The tokenized operation instructions and the operation instructions with error type tags are segmented according to the smallest word segmentation unit to form an operation statement dictionary; If the word appearing in the word segmentation result is the equipment name or number in the equipment ledger database, the word is marked as a mandatory matching word.

6. The method for checking dispatch operation tickets based on LSTM deep learning algorithm as claimed in claim 5, characterized in that: Based on the operation statement dictionary, a long short-term memory network model is constructed, and the operation statements are matched one by one with the generalized operation actions. The long short-term memory network model is trained to generate an operation instruction classification model and an instruction logic verification model. The model is verified by a self-prepared test operation ticket, including the following steps: Classify the entries in the operation statement dictionary according to the word type, construct a feature vector including a device name sequence, an action description sequence, and a state description sequence, and establish a mapping relationship between the feature vector and the corresponding generalized operation action; Constructing a long short-term memory network structure including an input layer, a hidden layer and an output layer, wherein the hidden layer includes three control units, namely a forget gate, an input gate and an output gate, the input layer receives a feature vector, and the output layer outputs a generalized operation action; Dividing the feature vector into a training set and a validation set in a ratio of 8:2, using the training set to train a long short-term memory network to generate an operation instruction classification model; Using the generalized operation action sequences in the training set to construct an instruction logic rule base, and training and generating an instruction logic verification model based on the rule base; A test operation ticket data set is constructed, wherein the test operation ticket data set includes correct operation instructions, incorrect operation instructions and logically incorrect operation instructions, and the test operation ticket data set is input into an operation instruction classification model and an instruction logic verification model for test verification.

7. The method for checking dispatch operation tickets based on LSTM deep learning algorithm according to claim 6, characterized in that: The construction of the instruction logic rule base includes the following rules: If a circuit breaker opening and closing operation instruction is detected, a corresponding ground wire removal instruction must exist before the circuit breaker opening and closing operation instruction; If a device switching to maintenance state instruction is detected, there must be an associated power-off instruction before the device switching to maintenance state instruction; If a parallel capacitor operation instruction is detected, the parallel capacitor operation instruction must be executed during the off-peak load period; If a transformer parallel operation command is detected, the ratio difference of the relevant transformers must be within the specified range; An instruction logic verification model is generated based on the rule base training, and the instruction logic verification model marks an operation instruction sequence that does not meet the rule requirements as a logic error.

8. A system using the method for checking dispatch operation tickets based on the LSTM deep learning algorithm as described in any one of claims 1 to 7, characterized in that: include: The collection and sorting module is used to collect historical operation ticket data and equipment ledger data, simplify the operation instructions in the operation ticket data into general operation actions, and sort out the substation and interval lists of the equipment ledger data; A processing module, used for horizontally formatting the operation ticket data; The verification module is used to construct a long short-term memory network model based on the operation statement dictionary, match the operation statements with the general operation actions one by one, train the long short-term memory network model to generate the operation instruction classification model and the instruction logic verification model, and verify the model through the self-prepared test operation ticket.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the scheduling operation ticket verification method based on the LSTM deep learning algorithm described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the scheduling operation ticket verification method based on the LSTM deep learning algorithm described in any one of claims 1 to 7 are implemented.

Images