Spot goods electric energy settlement task atlas construction method and related device

Abstract

The invention belongs to the field of electric power automation, and discloses a spot electric energy settlement task atlas construction method and related device.The method comprises the steps that a basic ontology model is constructed with settlement tasks of spot electric energy settlement as nodes and the logic relation between the settlement tasks as edges, data models of nodes and edges are established on the basic ontology model in an attribute graph modeling mode, and a spot goods electric energy settlement task graph ontology model is obtained; obtaining spot electric energy settlement text data, and based on a pre-trained named entity recognition model, performing extraction by adopting an entity and relationship combined extraction pipeline extraction method based on a relationship rule to obtain a spot electric energy settlement map triple; and based on the ontology model of the spot electric energy settlement task atlas, combining the triple of the spot electric energy settlement atlas to obtain the spot electric energy settlement task atlas. Unified arrangement of spot electric energy settlement data is realized, an automatic spot electric energy settlement system is effectively supported, and deep mining of the spot electric energy settlement data is facilitated.

Description

Technical Field

[0001] The present invention belongs to the field of electric power automation and relates to a method for constructing a spot electric energy settlement task map and a related device. Background Art

[0002] In the spot electricity settlement process, existing systems struggle to achieve efficient automated settlement calculations due to the extreme complexity and sheer volume of settlement data, as well as the high degree of coupling and interactivity between data. This data complexity leads to bottlenecks in system processing, low computational efficiency, and difficulty ensuring accuracy. Existing systems are particularly inadequate when faced with demands for more granular data processing. Therefore, to achieve efficient and accurate automated settlement of spot electricity, effective processing and optimization of settlement data are essential to improve the system's automated computing capabilities and adaptability. Summary of the Invention

[0003] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a method for constructing a spot electricity settlement task map and related devices.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] In a first aspect, the present invention provides a method for constructing a spot electricity settlement task graph, comprising: constructing a basic ontology model with settlement tasks of spot electricity settlement as nodes and logical relationships between settlement tasks as edges, and establishing data models of nodes and edges on the basic ontology model by using attribute graph modeling to obtain a spot electricity settlement task graph ontology model; obtaining spot electricity settlement text data, and extracting the spot electricity settlement text data based on a pre-trained named entity recognition model by using a pipeline extraction method for joint extraction of entities and relationships based on relational rules to obtain a spot electricity settlement graph triple; wherein the spot electricity settlement graph triple contains two entities and the relationship between the two entities; and obtaining a spot electricity settlement task graph based on the spot electricity settlement task graph ontology model and in combination with the spot electricity settlement graph triple.

[0006] Optionally, the logical relationship between the settlement tasks is the execution order of different types of settlement tasks; wherein, the types of settlement tasks include data extraction, data cleaning and verification, electricity calculation, settlement calculation and settlement result release; the execution order is: settlement tasks of data extraction type, settlement tasks of data cleaning and verification type, settlement tasks of electricity calculation type, settlement calculation type and settlement result release type.

[0007] Optionally, when the data model of nodes and edges is established by using attribute graph modeling, the attributes of the node include at least one of a unique identifier, a name, an execution subject, input data / output results, a precondition, and a postcondition; the attributes of the edge include at least one of an edge label, a source node, and a destination node.

[0008] Optionally, when the type of settlement task corresponding to the node is data extraction, the attributes of the node also include one or more of the following: data source, data format, extraction time range and data integrity; when the type of settlement task corresponding to the node is data cleaning and verification, the attributes of the node also include one or more of the following: data accuracy, data consistency, outlier processing and data quality assessment; when the type of settlement task corresponding to the node is electricity calculation, the attributes of the node also include one or more of the following: electricity calculation formula, meter reading, time interval and loss allocation; when the type of settlement task corresponding to the node is settlement calculation, the attributes of the node also include one or more of the following: electricity price policy, settlement items, billing cycle and additional fees; when the type of settlement task corresponding to the node is settlement result release, the attributes of the node also include one or more of the following: user information, settlement amount, billing cycle and payment status.

[0009] Optionally, the pre-trained named entity recognition model is obtained by pre-training an EBERT-BiLSTM-CRF model; wherein, the EBERT-BiLSTM-CRF model includes a pre-trained Transformer-based bidirectional encoder representation model, a bidirectional long short-term memory neural network model and a conditional random field model in the power field that are connected in sequence.

[0010] Optionally, the pre-trained named entity recognition model adopts a pipeline extraction method for joint extraction of entities and relationships based on relational rules to extract spot electricity settlement text data, including: obtaining event sentences of spot electricity settlement text data, and extracting two entities in the event sentences based on the pre-trained named entity recognition model; obtaining the types of the two entities and obtaining a preliminary identification relationship between the two entities based on the types of the two entities in combination with preset relational rules, and performing secondary identification through preset keywords of the preliminary identification relationship between the two entities in combination with the event sentences to obtain the relationship between the two entities; and combining the two entities and the relationship between the two entities to obtain a spot electricity settlement graph triple.

[0011] According to a second aspect of the present invention, a system for constructing a spot electricity settlement task graph is provided, comprising: an ontology model construction module for constructing a basic ontology model with settlement tasks of spot electricity settlement as nodes and logical relationships between settlement tasks as edges, and for establishing data models of nodes and edges on the basic ontology model using attribute graph modeling to obtain a spot electricity settlement task graph ontology model; a knowledge extraction module for acquiring spot electricity settlement text data, and extracting the spot electricity settlement text data based on a pre-trained named entity recognition model using a pipeline extraction method for entity and relationship joint extraction based on relational rules to obtain a spot electricity settlement graph triple; wherein the spot electricity settlement graph triple contains two entities and the relationship between the two entities; and a graph construction module for obtaining a spot electricity settlement task graph based on the spot electricity settlement task graph ontology model and in combination with the spot electricity settlement graph triple.

[0012] Optionally, the logical relationship between the settlement tasks is the execution order of different types of settlement tasks; wherein, the types of settlement tasks include data extraction, data cleaning and verification, electricity calculation, settlement calculation and settlement result release; the execution order is: settlement tasks of data extraction type, settlement tasks of data cleaning and verification type, settlement tasks of electricity calculation type, settlement calculation type and settlement result release type.

[0013] Optionally, when the data model of nodes and edges is established by using attribute graph modeling, the attributes of the node include at least one of a unique identifier, a name, an execution subject, input data / output results, a precondition, and a postcondition; the attributes of the edge include at least one of an edge label, a source node, and a destination node.

[0014] Optionally, when the type of settlement task corresponding to the node is data extraction, the attributes of the node also include one or more of the following: data source, data format, extraction time range and data integrity; when the type of settlement task corresponding to the node is data cleaning and verification, the attributes of the node also include one or more of the following: data accuracy, data consistency, outlier processing and data quality assessment; when the type of settlement task corresponding to the node is electricity calculation, the attributes of the node also include one or more of the following: electricity calculation formula, meter reading, time interval and loss allocation; when the type of settlement task corresponding to the node is settlement calculation, the attributes of the node also include one or more of the following: electricity price policy, settlement items, billing cycle and additional fees; when the type of settlement task corresponding to the node is settlement result release, the attributes of the node also include one or more of the following: user information, settlement amount, billing cycle and payment status.

[0015] Optionally, the pre-trained named entity recognition model is obtained by pre-training an EBERT-BiLSTM-CRF model; wherein, the EBERT-BiLSTM-CRF model includes a pre-trained Transformer-based bidirectional encoder representation model, a bidirectional long short-term memory neural network model and a conditional random field model in the power field that are connected in sequence.

[0016] Optionally, the knowledge extraction module is specifically used to: obtain event sentences of spot electricity settlement text data, and extract two entities in the event sentences based on a pre-trained named entity recognition model; obtain the types of the two entities and obtain a preliminary identification relationship between the two entities based on the types of the two entities in combination with preset relationship rules, and perform secondary identification through preset keywords of the preliminary identification relationship between the two entities in combination with the event sentences to obtain the relationship between the two entities; and combine the two entities and the relationship between the two entities to obtain a spot electricity settlement graph triple.

[0017] In a third aspect of the present invention, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for constructing a spot electricity settlement task map as described above are implemented.

[0018] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for constructing the spot electricity settlement task map are implemented.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] The present invention's method for constructing a spot electricity settlement task graph breaks down complex spot electricity settlement into multi-level, multi-instance settlement tasks. Based on the logical relationships between settlement tasks, it establishes node and edge data models in conjunction with attribute graph modeling, ultimately constructing a spot electricity settlement task graph ontology model. Then, based on a pre-trained named entity recognition model, a pipeline extraction method using entity and relationship joint extraction based on relationship rules is employed to extract triples of the spot electricity settlement graph. Next, the spot electricity settlement task graph ontology model and the spot electricity settlement graph triples are combined to obtain a spot electricity settlement task graph. This method achieves unified organization of spot electricity settlement data, more intuitively displays the data transfer relationships between settlement tasks, and facilitates deep mining of spot electricity settlement data. This method effectively addresses the current situation of spot electricity settlement, characterized by high data frequency, high volume of settlement data, and high coupling between settlement services. It provides a data foundation for automated spot electricity settlement under massive data volumes, effectively supporting an automated spot electricity settlement system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of a method for constructing a spot electricity settlement task map according to an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of the spot electricity settlement task graph ontology model framework of an embodiment of the present invention.

[0023] Figure 3 Schematic diagram of the named entity recognition model framework according to an embodiment of the present invention.

[0024] Figure 4 Schematic diagram of the encoding process of the BERT model according to an embodiment of the present invention.

[0025] Figure 5 Schematic diagram of the internal structure of the LSTM neural network according to an embodiment of the present invention.

[0026] Figure 6 This is a schematic diagram of the LSTM forget gate operation process of an embodiment of the present invention.

[0027] Figure 7 Schematic diagram of the LSTM input gate operation process of an embodiment of the present invention.

[0028] Figure 8 Schematic diagram of the LSTM memory state update process in an embodiment of the present invention.

[0029] Figure 9 Schematic diagram of the LSTM output gate operation process of an embodiment of the present invention.

[0030] Figure 10 Schematic diagram of the Bi-LSTM model structure of an embodiment of the present invention.

[0031] Figure 11 Schematic diagram of the CRF model principle of an embodiment of the present invention.

[0032] Figure 12 This is a relationship extraction flow chart of an embodiment of the present invention.

[0033] Figure 13 A system structure diagram for constructing a spot electricity settlement task map according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0035] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0036] The present invention is described in further detail below with reference to the accompanying drawings:

[0037] See also Figure 1 In one embodiment of the present invention, a method for constructing a spot electricity settlement task map is provided to realize the construction of a spot electricity settlement task map to support the automation of spot electricity settlement.

[0038] Specifically, the method for constructing a spot electricity settlement task map of the present invention includes the following steps:

[0039] S1: A basic ontology model is constructed with the settlement tasks of spot electricity settlement as nodes and the logical relationships between settlement tasks as edges. The data model of nodes and edges is established on the basic ontology model using attribute graph modeling to obtain the graph ontology model of spot electricity settlement tasks.

[0040] S2: Obtain spot electricity settlement text data, and based on a pre-trained named entity recognition model, use a pipeline extraction method for joint extraction of entities and relationships based on relational rules to extract the spot electricity settlement text data to obtain a spot electricity settlement graph triple; wherein the spot electricity settlement graph triple contains two entities and the relationship between the two entities.

[0041] S3: Based on the spot electricity settlement task graph ontology model and combined with the spot electricity settlement graph triples, the spot electricity settlement task graph is obtained.

[0042] The present invention's method for constructing a spot electricity settlement task graph breaks down complex spot electricity settlement into multi-level, multi-instance settlement tasks. Based on the logical relationships between settlement tasks, it establishes node and edge data models in conjunction with attribute graph modeling, ultimately constructing a spot electricity settlement task graph ontology model. Then, based on a pre-trained named entity recognition model, a pipeline extraction method using entity and relationship joint extraction based on relationship rules is employed to extract triples of the spot electricity settlement graph. Next, the spot electricity settlement task graph ontology model and the spot electricity settlement graph triples are combined to obtain a spot electricity settlement task graph. This method achieves unified organization of spot electricity settlement data, more intuitively displays the data transfer relationships between settlement tasks, and facilitates deep mining of spot electricity settlement data. This method effectively addresses the current situation of spot electricity settlement, characterized by high data frequency, high volume of settlement data, and high coupling between settlement services. It provides a data foundation for automated spot electricity settlement under massive data volumes, effectively supporting an automated spot electricity settlement system.

[0043] For clarification, the spot market specifically refers to the electricity market, which can be divided into the medium- and long-term market and the spot market based on the transaction cycle. The spot market is divided into the day-ahead and real-time markets, following a 96-point daily time curve with a minimum delivery period of 15 minutes. Electricity settlement specifically refers to the accounting of receivables and payables of operating entities based on their electricity trading results and market settlement rules, the issuance or acquisition of corresponding invoices, and the completion of fund collection and payment within the specified timeframe. The spot electricity settlement task map specifically refers to a tool that graphically displays the various settlement tasks and their interrelationships during the electricity market settlement process, providing an intuitive and clear visual representation of electricity market settlement. Ontology construction specifically refers to the ability to abstract settlement tasks into a clear, commonly recognized conceptual model for researchers in the field, forming a semantic foundation for information exchange and computer readability in this field. Knowledge extraction specifically refers to the task of identifying specific information from unstructured text and extracting it into structured form. Based on the extracted content, it can be categorized into three areas: entity extraction, relationship extraction, and attribute extraction.

[0044] In one possible implementation, the logical relationship between the settlement tasks is the execution order of different types of settlement tasks; wherein the types of settlement tasks include data extraction, data cleaning and verification, electricity calculation, settlement calculation and settlement result release; the execution order is: data extraction type settlement tasks, data cleaning and verification type settlement tasks, electricity calculation type settlement tasks, settlement calculation type settlement tasks and settlement result release type settlement tasks.

[0045] Explanatory, the settlement tasks of spot electricity settlement are divided according to functional types, specifically into data extraction type settlement tasks, data cleaning and verification type settlement tasks, electricity calculation type settlement tasks, settlement calculation type settlement tasks and settlement result publishing type settlement tasks.

[0046] Data extraction settlement tasks primarily extract, integrate, and store data related to electricity settlement from various data sources. Collected data includes electricity consumption and prices from power generation companies, power users, and power retail companies in medium- and long-term markets, spot markets, and ancillary service markets, as well as basic electricity data, market transaction data, and price and fee data from market operators. Data extraction consolidates disparate data into a unified database or table, providing data support for subsequent settlement calculations.

[0047] The settlement tasks of the data cleaning and verification type mainly realize the process of reviewing, verifying and correcting the extracted data to eliminate errors, inconsistencies and redundant information. It mainly includes: Data integrity and accuracy verification: ensuring that all necessary data fields are filled in and there are no missing, verifying the correctness of the data, and preventing incorrect data input. Data consistency verification: ensuring that the timestamps of all data are consistent and there are no time conflicts; all data units should be consistent, such as the electricity unit is megawatt-hour (MWh). Data logic verification: verifying whether the transaction records comply with market rules and logic; ensuring that the transaction price is within a reasonable range and there are no abnormal fluctuations. Data compliance verification: ensuring that all data complies with relevant rules and requirements. Data uniqueness verification: preventing duplicate data records; ensuring the uniqueness of each participant and transaction record. Data continuity verification: ensuring the continuity of the data sequence without interruption.

[0048] Electricity calculation is a core component of spot electricity settlement. It involves analyzing data on electricity trading, generation, consumption, and losses to determine the electricity settlement for each participant. This includes, but is not limited to: Electricity generation calculation: actual power generation, grid-connected power, and factory power consumption. Electricity consumption calculation: end-user power consumption. Inter-regional and inter-provincial transaction calculation: inter-regional and inter-provincial transaction power. Ancillary service calculation: frequency regulation, voltage regulation, black start, reserve capacity service power, and peak load shaving power.

[0049] Settlement calculation tasks, also a core component of spot electricity settlement, involve the settlement of electricity consumption, electricity prices, and electricity charges for each market participant. These tasks include, but are not limited to, day-ahead market settlement calculations for power generation, real-time market settlement calculations for power generation, day-ahead market settlement calculations for power consumption, real-time market settlement calculations for power consumption, and imbalance fund settlement.

[0050] Settlement results publishing tasks include result aggregation, consistency check, rule verification, bill generation, result publishing, result confirmation, and ultimately forming a detailed expense bill.

[0051] Regarding the execution order of different types of settlement tasks, the data extraction type of settlement task is the starting point of spot electricity settlement, providing basic data for subsequent settlement links; the data cleaning and verification type of settlement task is to clean and verify the extracted data, and depends on the data provided by the data extraction type of settlement task; the electricity calculation type of settlement task depends on the data provided by the data cleaning and verification type of settlement task, and outputs the electricity calculation results; the settlement calculation type of settlement task depends on the data provided by the electricity calculation type of settlement task to perform electricity bill settlement calculation; the settlement result publishing type of settlement task depends on the settlement data provided by the settlement calculation type of settlement task, and publishes the settlement results to relevant market entities and regulatory agencies.

[0052] Therefore, the execution order of different types of settlement tasks is as follows: data extraction type settlement tasks, data cleaning and verification type settlement tasks, power calculation type settlement tasks, settlement calculation type settlement tasks and settlement result publishing type settlement tasks.

[0053] Explanatory, see Figure 2 , showing a possible ontology model for a spot electricity settlement task graph. In the basic ontology model, settlement tasks for spot electricity settlement are nodes, and the logical relationships between settlement tasks are edges, thus linking all settlement tasks. The correspondence between the actual spot electricity settlement system and the spot electricity settlement task graph ontology model is shown in Table 1.

[0054] Table 1

[0055] Actual spot electricity settlement system Spot electricity settlement task graph ontology model Settlement tasks node Logical relationship between settlement tasks side Settlement task categories, parameters, etc. property

[0056] In one possible implementation, when the data model of nodes and edges is established using attribute graph modeling, the attributes of the node include at least one of a unique identifier, a name, an execution subject, input data / output results, a precondition, and a postcondition; the attributes of the edge include at least one of an edge label, a source node, and a destination node.

[0057] Explanatory, based on the sorted settlement tasks and their attributes, the settlement tasks are described using attribute graph modeling. Settlement tasks are used as node classifications, and attributes are used as description fields of nodes. It is stipulated that each node must have at least one non-null primary key, and the primary key must be non-repeatable to ensure the global uniqueness of the node. In this implementation, see Figure 2, the attributes of a node include a unique identifier (ID), name, execution subject, input data / output results, preconditions, and postconditions. Among them, the unique identifier specifically refers to the task identifier of each settlement task, which is used to distinguish different settlement tasks. The execution subject specifically refers to the subject that records the execution of the settlement task, such as power generation enterprises, power users, and power sales companies. Input data specifically refers to the input data required to describe the settlement task, such as meter readings, electricity price information, etc. Output results specifically refer to the output results generated by the settlement task, such as settlement bills and electricity statistics. Preconditions specifically refer to the preconditions that need to be met for the execution of the settlement task. Postconditions specifically refer to the postconditions that should be met after the settlement task is executed. Among them, the preconditions in the attributes of the node corresponding to the data extraction type settlement task can be empty, and the postconditions in the attributes of the node corresponding to the settlement result publishing type settlement task can be empty.

[0058] In one possible implementation, see Figure 2 When the type of settlement task corresponding to the node is data extraction, the attributes of the node also include one or more of the following: data source, data format, extraction time range and data integrity; when the type of settlement task corresponding to the node is data cleaning and verification, the attributes of the node also include one or more of the following: data accuracy, data consistency, outlier processing and data quality assessment; when the type of settlement task corresponding to the node is electricity calculation, the attributes of the node also include one or more of the following: electricity calculation formula, meter reading, time interval and loss allocation; when the type of settlement task corresponding to the node is settlement calculation, the attributes of the node also include one or more of the following: electricity price policy, settlement subject, billing cycle and additional fees; when the type of settlement task corresponding to the node is settlement result publishing, the attributes of the node also include one or more of the following: user information, settlement amount, billing cycle and payment status.

[0059] For example, see Figure 2 Table 2 shows a possible node model for a data extraction settlement task. Data source specifically refers to the source of the data, such as an electricity meter, sensor, or database. Data format specifically refers to the format and structure of the data to ensure proper extraction. Extraction time range specifically refers to the time period of the data to be extracted. Data integrity specifically refers to ensuring that the extracted data is complete and without missing data.

[0060] Table 2

[0061] Field Name Physical meaning Data Type Is it nullable? DEId Unique identifier INT False Name name STRING True Subject Execution entity STRING True Input data / Output resultsId Input data / output results INT True Data source Data Source STRING True Data format Data format STRING True Time range Extraction time range Date True Data integrity Data integrity STRING True Postcondition Postcondition STRING True

[0062] Among them, INT represents integer type, STRING represents string type, Date represents date type, False represents not empty, and True represents can be empty.

[0063] For example, see Figure 2 Table 3 shows a possible node model for nodes corresponding to data cleaning and verification type settlement tasks. Data accuracy specifically refers to checking data accuracy, such as whether numerical values ​​are reasonable and units are correct. Data consistency specifically refers to ensuring data consistency across different sources. Outlier handling specifically refers to identifying and handling outliers, such as missing values ​​and duplicate values. Data integrity specifically refers to reconfirming data integrity to ensure that nothing is missing.

[0064] Table 3

[0065] Field Name Physical meaning Data Type Is it nullable? DCV Unique identifier INT False Name name SRTING True Subject Execution entity SRTING True Input data / Output resultsId Input data / output results INT True Data accuracy Data accuracy SRTING True Data consistency Data consistency SRTING True Outlier processing Outlier handling SRTING True Data quality assessment Data quality assessment SRTING True Preconduction Prerequisites SRTING True Postcondition Postcondition SRTING True

[0066] For example, see Figure 2 Table 4 shows a possible node model for nodes corresponding to a settlement task of the power calculation type. The power calculation formula specifically refers to the formula used to calculate power. The meter reading specifically refers to the meter reading used as the basis for calculating power. The time interval specifically refers to the time interval for calculating power, such as hours, days, and months. The loss allocation specifically refers to the proportion of system loss allocated to the power and the amount of system loss.

[0067] Table 4

[0068]

[0069]

[0070] For example, see Figure 2 Table 5 shows a possible node model for nodes corresponding to settlement tasks of the settlement calculation type. The term "electricity price policy" specifically refers to understanding the applicable electricity price policy, including price structure and rates. The term "settlement account" specifically refers to setting settlement accounts based on the settlement order, including electricity clearing and market operating expenses. The term "billing cycle" specifically refers to determining the billing cycle for settlement, such as monthly or quarterly. The term "additional fees" specifically refers to considering possible additional fees, such as taxes and service fees.

[0071] Table 5

[0072]

[0073]

[0074] For example, see Figure 2Table 6 shows a possible node model for a settlement task with the settlement result publishing type. User information specifically refers to information about the associated user, such as name and account number. Settlement amount specifically refers to the settlement amount for each user. Billing cycle specifically refers to the cycle for specifying bills, which is consistent with the billing cycle for settlement calculations. Payment status specifically refers to the user's payment status, such as paid or unpaid.

[0075] Table 6

[0076] Field Name Physical meaning Data Type Is it nullable? SRRId Unique identifier INT False Name name SRTING True Subject Execution entity SRTING True Input data / Output resultsId Input data / output results INT True User User Information SRTING True Settlement amount Settlement amount INT True Billing cycle Billing cycle Date True Payment status Payment Status SRTING True Preconduction Prerequisites SRTING True

[0077] Explanatory, in the graph model, the edge is determined by the source node and the destination node. In the spot electricity settlement task graph ontology model, the edge represents the connection between two settlement tasks. According to the logical relationship between different settlement tasks, four types of edge relationships are established to represent the logical relationship between task nodes. For example, see Figure 2 Table 7 shows a possible edge data model of the spot electricity settlement task graph ontology model.

[0078] Table 7

[0079] Edge Label Source Node Destination Node D2CV Data extraction type Data cleaning and verification types CV2PC Data cleaning and verification types Power calculation type PC2SC Power calculation type Settlement calculation type SC2RR Settlement calculation type Settlement result publishing type

[0080] Based on the process and functions of spot energy settlement tasks, the graph ontology model for spot energy settlement tasks is simplified by simplifying the types and attributes of nodes and edges, making it more concise and efficient. Furthermore, if new nodes or attributes need to be added in the future, the graph ontology model for spot energy settlement tasks has flexible dynamic update and expansion capabilities.

[0081] In one possible embodiment, the pre-trained named entity recognition model is obtained based on the EBERT-BiLSTM-CRF model; wherein, the EBERT-BiLSTM-CRF model includes a pre-trained Transformer-based bidirectional encoder representation model, a bidirectional long short-term memory neural network model and a conditional random field model in the power field that are connected in sequence.

[0082] Explanatory, after completing the construction of the spot electricity settlement task graph ontology model, in order to realize the construction of the spot electricity settlement task graph, the next step is to carry out knowledge extraction, and extract entities and relationships of the spot electricity settlement text data according to the spot electricity settlement task graph ontology model.

[0083] In this embodiment, the extraction task primarily relies on a pretrained named entity recognition model, which is based on the EBERT-BiLSTM-CRF model. The EBERT-BiLSTM-CRF model consists of a pretrained Transformer-based bidirectional encoder representation model, a bidirectional long short-term memory neural network model, and a conditional random field model, all connected in sequence.

[0084] For example, a pre-training and fine-tuning model is used with a small training sample. The EBERT model (Electric Bidirectional Encoder Representation from Transformers) based on power sector data is selected as the text encoding representation layer to obtain word-level feature vectors, which are then transferred to the fine-tuning network for the downstream named entity recognition task. The downstream named entity recognition model network uses a combination of the BiLSTM model (Bi-directional Long Short-Term Memory) and the CRF model (Conditional Random Field) to finally complete the construction of the named entity recognition model. See [1]. Figure 3 In the named entity recognition model, the pre-trained EBERT model primarily uses its tokenizer and encoder layers to split and encode the input text data, outputting corresponding word vectors. The BiLSTM model uses a bidirectional LSTM (Long Short Term Memory) neural network trained on the corpus to collect and concatenate the forward and backward features of each word vector in the sentence, outputting the probability of the label corresponding to each word. The CRF model is used to obtain the optimal label annotation sequence at the sentence level and output the corresponding annotation label for the named entity.

[0085] Explanation: The EBERT model is based on the BERT model (Bidirectional Encoder Representation from Transformers), pre-trained with data from the power sector. The BERT model can encode and represent natural language text. After pre-training, the model only needs to add an output layer to complete fine-tuning for various downstream tasks, and can achieve good results. Figure 4 , showing the encoding process of the BERT model.

[0086] For example, during pre-training with power sector data, a word-level masking strategy was used when constructing the pre-training corpus for the power sector model. 15% of the words in the entire power sector corpus were randomly masked. Of the data to be masked, 80% were masked with a mask symbol ([MASK]), 10% with random characters, and 10% with the original characters remaining unchanged. This whole-word masking approach can improve the model's problem of ignoring the semantic information between words and entire sentences. The original model pre-training phase used character-level masking, which resulted in partial masking of words during processing.

[0087] Specifically, by adding the relative position parameters between words, introducing the attention matrix for encoding, and establishing the connection between words and sentences, the calculation formula of the attention matrix is ​​shown as follows:

[0088] Attention_rel(Q,K,V)=Attention(Q,K,V)+rel

[0089] Among them, rel is the relative position parameter between words, Attention(Q,K,V) is the formula for encoding the attention matrix, and the calculation method is shown in the following formula:

[0090]

[0091] Among them, V is the vector of input features, Q and K are the feature vectors for calculating weights, and d k is the dimension size of K, Q, K and V represent Query, Key and Value respectively, which is the initial input vector matrix X through three trainable parameter matrices W Q 、W K and W V The transformation is obtained, and the calculation formula is shown as follows:

[0092] Q=X·W Q , K=X·W K , V=X·W V

[0093] Other training strategies during training, such as residual connection and normalization, remain consistent with BERT. The residual connection process adds the unencoded word vectors to the vectors calculated by the Attention mechanism. The normalization layer normalizes the vectors during the encoding process to a standard normal distribution.

[0094] Explanation: The BiLSTM model consists of two independent, inversely coupled long short-term memory (LSTM) neural networks, combined with two unidirectional recurrent neural networks with opposite input sequences. This model has the ability to learn information in both the forward and backward directions, enabling it to better encode contextual semantic information. This makes it particularly effective in sequence labeling tasks such as named entity recognition. LSTM is a variation of the RNN (Recurrent Neural Network). By adding three gates to the hidden layer h, namely the forget gate, input gate, and output gate, as well as a new hidden state (cell state), LSTM addresses the exploding and vanishing gradient issues of RNNs and enhances their long-distance processing capabilities. (See [1] for more information.) Figure 5 , showing the internal structure of the LSTM neural network.

[0095] See also Figure 6 , shows the operation process of LSTM forget gate, forget gate f t By reading the last time output information h t-1 and the current input x t , using the σ(x) activation function to do nonlinear mapping, thereby outputting a vector f that discards irrelevant information t , each value of this vector is between 0 and 1, 1 means completely retained, and 0 means completely discarded. The specific calculation process is:

[0096]

[0097] Among them, σ(x) is the activation function sigmoid, which compresses the values ​​flowing through the network between 0 and 1, which helps to update or forget information. That is, the part of the information where the activation function is 0 will be completely eliminated, and the part of the information where the activation function is 1 will be completely retained, so as to achieve the purpose of remembering important information and forgetting irrelevant information; b f is the bias, which adds translation capability to the network classification; W f is a weight matrix, that is, the output information at the previous time and the input information at the current time have their own weight values.

[0098] See also Figure 7 , shows the operation process of LSTM input gate, input gate i t It is determined to store new information, including the input gate layer and the tanh layer. The input gate layer explicitly updates the value: i t =σ(W i ·[h t-1 ,x t ]+b i ). The tanh layer is based on the output information h of the previous moment t-1 and the current input xt Create a new candidate value vector Add to pending status:

[0099] See also Figure 8 , shows the LSTM memory state update process, multiplying the old memory state by the forget gate, indicating that irrelevant information has been discarded, and adding the information merged with the candidate value and the input gate layer to obtain the new memory state. The update formula is as follows:

[0100] See also Figure 9 , shows the operation process of the LSTM output gate, the output gate outputs information h according to the hidden layer at the previous moment t-1 and the current input x t Determine some of the information to be output: t =σ(W o ·[h t-1 ,x t ]+b o ). At the same time, the hidden layer output information at time t needs to be obtained based on the new memory state, and the new memory state C t The flow passes through the tanh(x) function to determine which parts of the new memory state need to be output, and then multiply it with the sigmoid layer in the output gate: h t =o t *tanh(C t ). Among them, C t 、h t are the output values ​​of this sequence, representing long-term memory information and short-term memory information respectively.

[0101] Since LSTM is a one-way transmission, it cannot represent the information features in the reverse process, which limits its performance in sequence labeling tasks. Bi-LSTM solves this problem. Figure 10 , shows the Bi-LSTM model structure. In the Bi-LSTM model, the word vector x t Input forward LSTM unit output Vector, input to LSTM unit output vector, and then and Connect the head and tail to get x t The vector representation h in the x sequence t From the perspective of text sequences, the model implements modeling of the semantics before and after the input sentence, and the obtained entity label sequence results are more accurate.

[0102] Explanation: CRF model is a classic model for solving sequence labeling problems. Let x = x1, x2, ..., xn ,y=y1,y2,...,y n If two random variable sequences are given, the CR model F can find the conditional probability distribution p(y|x) of sequence y given the sequence x. The CRF model forms an undirected graph model that can consider the global probability of the data. Therefore, it is very suitable for application scenarios where there is a dependency relationship between the prediction labels, such as named entity recognition tasks. Conditional random fields usually refer to chained conditional random fields, see Figure 11 , in the chained CRF model, the input sentence and label sequence x = x1, x2, ..., x n , the model outputs entity label sequence y=y1,y2,...,y n , completing the entity recognition labeling task. Among them, B is used as the beginning of the entity, I is used as the middle or end of the entity, and O represents the non-entity part. In the chain conditional random field, the conditional probability is defined as follows:

[0103]

[0104] Among them, t j (y i+1 ,y i ,x,i) is the transition feature function between adjacent tag sequences in the observed sequence, which is related to the current node and the previous node. k (y i ,x,i) is the state feature function defined at the corresponding sequence marker position, which is only related to the current node. j and u k are the weights corresponding to the transition feature function and the state feature function, respectively, and n is the sequence length. Z is a normalization factor used to ensure that the conditional probability is correctly defined, as defined below:

[0105]

[0106] When the characteristic function condition is met, t j (y i-1 ,y i ,x,i) and s k (y i ,x,i) is set to 1, and is set to 0 if it is not satisfied.

[0107] In a possible embodiment, the pre-trained named entity recognition model adopts a pipeline extraction method for joint extraction of entities and relationships based on relational rules to extract spot electricity settlement text data: obtain event sentences of spot electricity settlement text data, and extract two entities in the event sentences based on the pre-trained named entity recognition model; obtain the types of the two entities and obtain a preliminary identification relationship between the two entities based on the types of the two entities in combination with preset relational rules, and perform secondary identification through preset keywords of the preliminary identification relationship between the two entities in combination with the event sentences to obtain the relationship between the two entities; and combine the two entities and the relationship between the two entities to obtain a spot electricity settlement graph triple.

[0108] Explanatory,relationship extraction is an important task of knowledge extraction.,Generally, entities and relationships between entities are expressed in the form of,triples, for example,<e1,r,e2> Here, e1 and e2 represent entities, and r represents the relationship between the two entities. The goal of relation extraction is to extract the relationship information expressed by the triples from unstructured natural language text. Based on the attributes of each settlement task, a rule-based pipeline-based relation extraction is adopted to complete the knowledge extraction and automatic construction of the domain knowledge base. Based on the named entity recognition model, a rule-based pipeline extraction method is proposed for the joint extraction of entities and their relationships.

[0109] For example, see Figure 12 The relationship extraction process is divided into three parts: (1) First, a script is written according to the format characteristics of the spot electricity settlement text data to extract and classify event sentences, and a corpus for named entity recognition and relationship extraction is constructed through manual annotation; (2) Then, the manually annotated data is used to train and evaluate the named entity recognition model; (3) The trained named entity recognition model is used to predict the entity type, and the edge relationship is judged and the relationship is extracted according to the node type defined by the spot electricity settlement task graph ontology model. The relationship discriminator first discriminates the relationship based on the type of the source node and the destination node, and then performs a secondary recognition of the relationship through keywords to improve the accuracy of relationship extraction.

[0110] In general, the present invention aims at the current spot electricity settlement system based on the unified electricity market trading platform. With the frequent iteration of settlement rules, it is often difficult to carry out automated calculations continuously. In order to cope with the characteristics of high data frequency, tens of millions and hundreds of millions of settlement data volumes and high coupling and complexity between settlement businesses, a spot electricity settlement task map is established. Specifically, by decomposing the complex spot electricity settlement process into multi-level and multi-instance settlement tasks and clarifying the dependencies between the settlement tasks, an ontology model of the spot electricity settlement task map is established. At the same time, a named entity recognition model based on a deep neural network is designed and combined with a pipeline extraction method for entity and relationship joint extraction based on relational rules to realize knowledge extraction of spot electricity settlement text data. Finally, the spot electricity settlement task map is formed by combining the spot electricity settlement task map ontology model and the knowledge extraction results. It has important guidance and practical significance for the automated operation of the spot electricity settlement system under massive data.

[0111] The following are device embodiments of the present invention, which can be used to perform the method embodiments of the present invention. For details not disclosed in the device embodiments, please refer to the method embodiments of the present invention.

[0112] See also Figure 13 In another embodiment of the present invention, a spot electricity settlement task graph construction system is provided, which can be used to implement the above-mentioned spot electricity settlement task graph construction method. Specifically, the spot electricity settlement task graph construction system includes an ontology model construction module, a knowledge extraction module and a graph construction module.

[0113] Among them, the ontology model construction module is used to construct a basic ontology model with the settlement tasks of spot electricity settlement as nodes and the logical relationships between settlement tasks as edges, and to establish data models of nodes and edges on the basic ontology model by using attribute graph modeling to obtain a spot electricity settlement task graph ontology model; the knowledge extraction module is used to obtain spot electricity settlement text data, and based on a pre-trained named entity recognition model, adopts a pipeline extraction method of entity and relationship joint extraction based on relational rules to extract the spot electricity settlement text data to obtain a spot electricity settlement graph triple; wherein, the spot electricity settlement graph triple contains two entities and the relationship between the two entities; the graph construction module is used to obtain a spot electricity settlement task graph based on the spot electricity settlement task graph ontology model and in combination with the spot electricity settlement graph triple.

[0114] In one possible implementation, the logical relationship between the settlement tasks is the execution order of different types of settlement tasks; wherein the types of settlement tasks include data extraction, data cleaning and verification, electricity calculation, settlement calculation and settlement result release; the execution order is: data extraction type settlement tasks, data cleaning and verification type settlement tasks, electricity calculation type settlement tasks, settlement calculation type settlement tasks and settlement result release type settlement tasks.

[0115] In one possible implementation, when the data model of nodes and edges is established using attribute graph modeling, the attributes of the node include at least one of a unique identifier, a name, an execution subject, input data / output results, a precondition, and a postcondition; the attributes of the edge include at least one of an edge label, a source node, and a destination node.

[0116] In one possible implementation, when the type of settlement task corresponding to the node is data extraction, the attributes of the node also include one or more of the following: data source, data format, extraction time range and data integrity; when the type of settlement task corresponding to the node is data cleaning and verification, the attributes of the node also include one or more of the following: data accuracy, data consistency, outlier processing and data quality assessment; when the type of settlement task corresponding to the node is electricity calculation, the attributes of the node also include one or more of the following: electricity calculation formula, meter reading, time interval and loss allocation; when the type of settlement task corresponding to the node is settlement calculation, the attributes of the node also include one or more of the following: electricity price policy, settlement subject, billing cycle and additional fees; when the type of settlement task corresponding to the node is settlement result release, the attributes of the node also include one or more of the following: user information, settlement amount, billing cycle and payment status.

[0117] In one possible embodiment, the pre-trained named entity recognition model is obtained based on the EBERT-BiLSTM-CRF model; wherein, the EBERT-BiLSTM-CRF model includes a pre-trained Transformer-based bidirectional encoder representation model, a bidirectional long short-term memory neural network model and a conditional random field model in the power field that are connected in sequence.

[0118] In one possible embodiment, the knowledge extraction module is specifically used to: obtain event sentences of spot electricity settlement text data, and extract two entities in the event sentences based on a pre-trained named entity recognition model; obtain the types of the two entities and obtain a preliminary identification relationship between the two entities based on the types of the two entities in combination with preset relationship rules, and perform secondary identification through preset keywords of the preliminary identification relationship between the two entities in combination with the event sentences to obtain the relationship between the two entities; and combine the two entities and the relationship between the two entities to obtain a spot electricity settlement graph triple.

[0119] All relevant contents of each step involved in the embodiment of the aforementioned spot electricity settlement task map construction method can be referred to the functional description of the functional modules corresponding to the spot electricity settlement task map construction system in the embodiment of the present invention, and will not be repeated here.

[0120] The module division in the embodiments of the present invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in various embodiments of the present invention may be integrated into a single processor, exist physically as separate modules, or two or more modules may be integrated into a single module. The integrated modules may be implemented in either hardware or software functional modules.

[0121] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory, wherein the memory is configured to store a computer program, the computer program including program instructions, and the processor is configured to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions, specifically for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function. The processor described in the embodiment of the present invention can be used for the operation of the method for constructing a spot electricity settlement task map.

[0122] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a computer device, used to store programs and data. It is understood that the computer-readable storage medium herein may include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides storage space, which stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for being loaded and executed by a processor. These instructions may be one or more computer programs (including program code). It should be noted that the computer-readable storage medium herein may be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. The processor may load and execute the one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the method for constructing a spot electricity settlement task map in the above-mentioned embodiment.

[0123] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take 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.) containing computer-usable program code.

[0124] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, 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 processes in the flowcharts and / or block diagrams. 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.

[0125] 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.

[0126] 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 A step that specifies a function in one or more boxes.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for constructing a spot electricity settlement task map, characterized in that: include: A basic ontology model is constructed with the settlement tasks of spot electricity settlement as nodes and the logical relationships between settlement tasks as edges. A data model of nodes and edges is established on the basic ontology model using attribute graph modeling to obtain a graph ontology model of spot electricity settlement tasks. Obtain spot electricity settlement text data and, based on a pre-trained named entity recognition model, use a pipeline extraction method that extracts entities and relationships based on relational rules to extract the spot electricity settlement text data, obtaining a spot electricity settlement graph triple. The spot electricity settlement graph triple contains two entities and the relationship between the two entities. Based on the spot electricity settlement task graph ontology model and combined with the spot electricity settlement graph triples, the spot electricity settlement task graph is obtained.

2. The method for constructing a spot electricity settlement task map according to claim 1, characterized in that: The logical relationship between the settlement tasks is the execution order of different types of settlement tasks; Among them, the types of settlement tasks include data extraction, data cleaning and verification, electricity calculation, settlement calculation and settlement result release; the execution order is: settlement tasks of data extraction type, settlement tasks of data cleaning and verification type, settlement tasks of electricity calculation type, settlement calculation type and settlement result release type.

3. The method for constructing a spot electricity settlement task map according to claim 1, characterized in that: When establishing the data model of nodes and edges by adopting the attribute graph modeling method, the attributes of the nodes include at least one of a unique identifier, a name, an execution subject, input data / output results, a precondition, and a postcondition; The attributes of an edge include at least one of an edge label, a source node, and a destination node.

4. The method for constructing a spot electricity settlement task map according to claim 3, characterized in that: When the type of settlement task corresponding to the node is data extraction, the attributes of the node also include one or more of the following: data source, data format, extraction time range and data integrity; when the type of settlement task corresponding to the node is data cleaning and verification, the attributes of the node also include one or more of the following: data accuracy, data consistency, outlier processing and data quality assessment; when the type of settlement task corresponding to the node is electricity calculation, the attributes of the node also include one or more of the following: electricity calculation formula, meter reading, time interval and loss allocation; when the type of settlement task corresponding to the node is settlement calculation, the attributes of the node also include one or more of the following: electricity price policy, settlement items, billing cycle and additional fees; when the type of settlement task corresponding to the node is settlement result publishing, the attributes of the node also include one or more of the following: user information, settlement amount, billing cycle and payment status.

5. The method for constructing a spot electricity settlement task map according to claim 1, characterized in that: The pre-trained named entity recognition model is obtained based on the EBERT-BiLSTM-CRF model pre-training; wherein, the EBERT-BiLSTM-CRF model includes a pre-trained Transformer-based bidirectional encoder representation model, a bidirectional long short-term memory neural network model and a conditional random field model in the power field connected in sequence.

6. The method for constructing a spot electricity settlement task map according to claim 1, characterized in that: The pre-trained named entity recognition model uses a pipeline extraction method for entity and relationship joint extraction based on relationship rules to extract spot electricity settlement text data, including: The invention obtains event sentences of spot electricity settlement text data, and extracts two entities in the event sentences based on a pre-trained named entity recognition model; obtains the types of the two entities and obtains a preliminary identification relationship between the two entities according to the types of the two entities in combination with preset relationship rules, and performs secondary identification by combining the preset keywords of the preliminary identification relationship between the two entities with the event sentences to obtain the relationship between the two entities; and combines the two entities and the relationship between the two entities to obtain a spot electricity settlement graph triple.

7. A spot electricity settlement task map construction system, characterized in that: include: The ontology model construction module is used to construct a basic ontology model with the settlement tasks of spot electricity settlement as nodes and the logical relationships between settlement tasks as edges, and to establish a data model of nodes and edges using attribute graph modeling on the basic ontology model to obtain a graph ontology model of spot electricity settlement tasks; A knowledge extraction module is used to obtain spot electricity settlement text data and, based on a pre-trained named entity recognition model, uses a pipeline extraction method that extracts entities and relationships based on relational rules to extract the spot electricity settlement text data, thereby obtaining a spot electricity settlement graph triple. The spot electricity settlement graph triple contains two entities and a relationship between the two entities. The graph construction module is used to obtain the spot electricity settlement task graph based on the spot electricity settlement task graph ontology model and the spot electricity settlement graph triples.

8. The spot electricity settlement task map construction system according to claim 7, characterized in that: The logical relationship between the settlement tasks is the execution order of different types of settlement tasks; Among them, the types of settlement tasks include data extraction, data cleaning and verification, electricity calculation, settlement calculation and settlement result release; the execution order is: settlement tasks of data extraction type, settlement tasks of data cleaning and verification type, settlement tasks of electricity calculation type, settlement calculation type and settlement result release type.

9. The spot electricity settlement task map construction system according to claim 7, characterized in that: When establishing the data model of nodes and edges by adopting the attribute graph modeling method, the attributes of the nodes include at least one of a unique identifier, a name, an execution subject, input data / output results, a precondition, and a postcondition; The attributes of an edge include at least one of an edge label, a source node, and a destination node.

10. The spot electricity settlement task map construction system according to claim 9, characterized in that: When the type of settlement task corresponding to the node is data extraction, the attributes of the node also include one or more of the following: data source, data format, extraction time range and data integrity; when the type of settlement task corresponding to the node is data cleaning and verification, the attributes of the node also include one or more of the following: data accuracy, data consistency, outlier processing and data quality assessment; when the type of settlement task corresponding to the node is electricity calculation, the attributes of the node also include one or more of the following: electricity calculation formula, meter reading, time interval and loss allocation; when the type of settlement task corresponding to the node is settlement calculation, the attributes of the node also include one or more of the following: electricity price policy, settlement items, billing cycle and additional fees; when the type of settlement task corresponding to the node is settlement result publishing, the attributes of the node also include one or more of the following: user information, settlement amount, billing cycle and payment status.

11. The spot electricity settlement task map construction system according to claim 7, characterized in that: The pre-trained named entity recognition model is obtained based on the EBERT-BiLSTM-CRF model pre-training; wherein, the EBERT-BiLSTM-CRF model includes a pre-trained Transformer-based bidirectional encoder representation model, a bidirectional long short-term memory neural network model and a conditional random field model in the power field connected in sequence.

12. The spot electricity settlement task map construction system according to claim 7, characterized in that: The knowledge extraction module is specifically used to: The invention obtains event sentences of spot electricity settlement text data, and extracts two entities in the event sentences based on a pre-trained named entity recognition model; obtains the types of the two entities and obtains a preliminary identification relationship between the two entities according to the types of the two entities in combination with preset relationship rules, and performs secondary identification by combining the preset keywords of the preliminary identification relationship between the two entities with the event sentences to obtain the relationship between the two entities; and combines the two entities and the relationship between the two entities to obtain a spot electricity settlement graph triple.

13. A computer 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 computer program, the steps of the method for constructing a spot electricity settlement task map as described in any one of claims 1 to 6 are implemented.

14. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for constructing a spot electricity settlement task map as described in any one of claims 1 to 6 are implemented.

Images