CN114281878A - Multimode data fusion method, device and medium for power market - Google Patents
Multimode data fusion method, device and medium for power market Download PDFInfo
- Publication number
- CN114281878A CN114281878A CN202111612678.1A CN202111612678A CN114281878A CN 114281878 A CN114281878 A CN 114281878A CN 202111612678 A CN202111612678 A CN 202111612678A CN 114281878 A CN114281878 A CN 114281878A
- Authority
- CN
- China
- Prior art keywords
- entity
- graph
- anchor point
- anchor
- knowledge graph
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The application discloses a multimodal data fusion method, equipment and medium for an electric power market, which comprise the following steps: constructing a knowledge graph facing the electric power market; entities of the knowledge graph represent various indicators in the electricity market; the edges of the knowledge graph represent the relationship between the indexes; preprocessing entity features in the knowledge graph to convert various data into uniform multidimensional vectors; identifying an anchor entity from the knowledge graph, and fusing anchor entity information; and converting the knowledge graph into an anchor point graph formed by the fused anchor point entities. Therefore, the knowledge graph is adopted to realize multi-mode fusion in the electric power market, various indexes are expressed in a graph structure and are represented as uniform multi-dimensional vectors, data in multiple modes and multiple fields are effectively integrated, a large knowledge graph is converted into a small anchor point graph formed by anchor points, the number of entities is reduced, the efficiency is improved, and rapid and cross-platform heterogeneous data retrieval and integrated management can be realized.
Description
Technical Field
The invention relates to the field of data processing, in particular to a multimodal data fusion method, equipment and medium for an electric power market.
Background
At present, the normal operation of the power system depends on data transmission and mutual cooperation among various service systems, and these service systems are built successively in different ages and on different platforms, and databases, operation platforms and specific data structures used by them may be different, thereby causing a large amount of heterogeneous structured and unstructured data to appear in the automation system, such as power grid topology and operation data of different formats, power equipment information, geographic environment data, meteorological data, audio and video, and a large amount of text data of different formats.
In order to realize communication interaction and information integration among the heterogeneous data, a large number of data conversion interfaces and intermediate links need to be added among different platforms in the power system. However, these data platforms are relatively independent, and there is no connection between data, so that it is difficult to implement fast, cross-platform data retrieval and integrated management.
Therefore, how to establish a multimodal data fusion method oriented to the power market is a technical problem to be solved urgently by those skilled in the art.
Disclosure of Invention
In view of this, the present invention provides a multimodal data fusion method, device and medium for an electric power market, which can promote effective integration of multimodal and multi-domain data, and implement fast and cross-platform heterogeneous data retrieval and integrated management. The specific scheme is as follows:
a multimodal data fusion method oriented to a power market comprises the following steps:
constructing a knowledge graph facing the electric power market; entities of the knowledge graph represent a plurality of indicators in the electricity market; the edges of the knowledge graph represent the relationship between the indexes;
preprocessing the entity characteristics in the knowledge graph to convert various data into uniform multidimensional vectors;
identifying an anchor entity from the knowledge graph, and fusing anchor entity information;
and converting the knowledge graph into an anchor point graph formed by the fused anchor point entities.
Preferably, in the above multimodal data fusion method provided by the embodiment of the present invention, the constructing a knowledge graph oriented to the power market includes:
collecting key data in the electric power market; the format of the key data comprises structured data, semi-structured data and unstructured data;
extracting information including power market index names, power market index relationships and power market index attributes from the key data;
and constructing a knowledge graph facing the electric power market according to the extracted information.
Preferably, in the above multimodal data fusion method provided by the embodiment of the present invention, the preprocessing the entity features in the knowledge graph includes:
converting the picture into a feature vector by adopting pre-trained fast-RCNN;
and converting the text into the feature vector by adopting a pre-trained BERT model.
Preferably, in the above multi-modal data fusion method provided by the embodiment of the present invention, the identifying an anchor entity from the knowledge-graph includes:
and extracting key entities from the pictures and the question texts which are converted into the feature vectors to serve as anchor entities.
Preferably, in the above multi-modal data fusion method provided in the embodiment of the present invention, the fusing the anchor entity information includes:
acquiring the importance degree of neighbors and relations corresponding to the anchor entity;
and taking the importance degree as weight, and fusing the information of the multi-order neighbors in a graph neural network mode to serve as a fused anchor point entity.
Preferably, in the above multi-modal data fusion method provided by the embodiment of the present invention, while converting the knowledge-graph into an anchor point graph composed of fused anchor point entities, the method further includes:
when the end points of the knowledge graph are edges on the anchor point entity, keeping the edges in the anchor point graph;
when the endpoints of the two sides in the knowledge graph are not on the edge of the anchor point entity, deleting the edge;
when one end point in the knowledge graph is on the anchor point entity and the other end point is on the side of the non-anchor point entity, searching another anchor point entity closest to the non-anchor point entity on the knowledge graph, and calculating the distance between the non-anchor point entity and the other anchor point entity; if the distance is smaller than a set value, adding an anchor point entity to one edge of another anchor point entity in the anchor point graph; if the distance is not less than the set value, the edge is deleted.
Preferably, in the above multi-modal data fusion method provided by the embodiment of the present invention, after constructing the knowledge graph oriented to the power market, the method further includes:
performing coreference resolution and entity disambiguation on the knowledge graph to fuse knowledge.
Preferably, in the above multi-modal data fusion method provided by the embodiment of the present invention, after converting the knowledge-graph into an anchor point graph composed of fused anchor point entities, the method further includes:
and quantizing the confidence degrees of the anchor point entities in the anchor point graph, and discarding the anchor point entities corresponding to the confidence degrees smaller than a preset threshold value.
The embodiment of the invention also provides a multi-modal data fusion device facing the power market, which comprises a processor and a memory, wherein when the processor executes a computer program stored in the memory, the multi-modal data fusion method provided by the embodiment of the invention is realized.
The embodiment of the present invention further provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the above-mentioned multi-modal data fusion method provided by the embodiment of the present invention.
According to the technical scheme, the multimodal data fusion method for the power market comprises the following steps: constructing a knowledge graph facing the electric power market; entities of the knowledge graph represent various indicators in the electricity market; the edges of the knowledge graph represent the relationship between the indexes; preprocessing entity features in the knowledge graph to convert various data into uniform multidimensional vectors; identifying an anchor entity from the knowledge graph, and fusing anchor entity information; and converting the knowledge graph into an anchor point graph formed by the fused anchor point entities.
The invention adopts the knowledge graph to realize multi-mode fusion in the electric power market, expresses various indexes into a graph structure, and represents the indexes as uniform multi-dimensional vectors, thereby promoting the effective integration of multi-mode multi-field data, converts the large knowledge graph into a smaller anchor point graph consisting of anchor points, reduces the number of entities, improves the efficiency, and can realize quick and cross-platform heterogeneous data retrieval and integrated management.
In addition, the invention also provides corresponding equipment and a computer readable storage medium aiming at the multi-mode data fusion method, so that the method has higher practicability, and the equipment and the computer readable storage medium have corresponding advantages.
Drawings
In order to more clearly illustrate the embodiments of the present invention or technical solutions in related arts, the drawings used in the description of the embodiments or related arts will be briefly introduced below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of a multimodal data fusion method for a power market according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of the relationship between the electric power market indicator systems provided by the embodiment of the present invention;
FIG. 3 is a schematic diagram of power market knowledge extraction provided by an embodiment of the invention;
FIG. 4 is a schematic diagram of an electric power market data index ontology according to an embodiment of the present invention;
fig. 5 is a schematic view of an anchor point frame corresponding to the power market indicator fusion relationship according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides a multimodal data fusion method for a power market, which comprises the following steps as shown in figure 1:
s101, constructing a knowledge graph facing to the electric power market; entities of the knowledge graph represent various indicators in the electricity market; the edges of the knowledge graph represent the relationship between the indexes;
specifically, multiple indexes in the power market are represented as an entity e in the graph, and the relationship between the indexes is represented as an edge in the graph, so that one edge can be represented as the graphTriple of (e)h,r,et) Indicates a head entity ehAnd tail entity etThe construction process of the knowledge graph with a certain relation r is a process of extracting knowledge elements (namely facts) from original data by adopting a series of automatic or semi-automatic technical means and storing the knowledge elements into a knowledge base from the original data, and the process is an iterative updating process.
In practical application, a common index system of each power market and a unique index system of each power market are extracted, and index items of different levels of each power market index system are used for reference, as shown in fig. 2, the index system which covers the whole process of the power market is summarized and extracted, and the index system comprises: energy economy indexes, power system operation indexes, power market transaction indexes and power market management indexes. Each index system has a strong association relationship, taking the index of electric power market trading as an example, the change of economic situation can influence the expectation of electric power users to the market, and further influence the trading scale of the electric power market trading, the electric power trading scale is restrained by the operation of the electric power system, the cost of power generation enterprises can be influenced by the price of energy, and further influence the quotation of the power generation enterprises during electric power trading, the fluctuation of the electric power trading price can be influenced by market-oriented market main bodies, and the management needs to be carried out aiming at the market so as to ensure the stable operation of the electric power market.
S102, preprocessing entity features in the knowledge graph to convert various data into uniform multi-dimensional vectors;
it should be noted that, due to the multi-modal characteristics of the features of the power market, some features are pictures, and some features are characters, so that the present invention adopts a preprocessing mode to characterize different modal features as feature vectors which are convenient to process.
S103, identifying an anchor entity from the knowledge graph, and fusing anchor entity information;
and S104, converting the knowledge graph into an anchor point graph formed by the fused anchor point entities.
In practical application, after the anchor entity characteristics are obtained, non-anchor entities in the graph can be deleted in the next step, and particularly, key anchor point entities identified in the original knowledge graph are extracted to be used as points of a final anchor point graph, so that the original large knowledge graph is converted into a smaller anchor point knowledge graph formed by anchor points, and the performance can be greatly improved.
In the multi-mode data fusion method provided by the embodiment of the invention, the knowledge graph is adopted to realize multi-mode fusion in the power market, various indexes are expressed in a graph structure and are represented as uniform multi-dimensional vectors, so that the multi-mode multi-field data is effectively integrated, a large knowledge graph is converted into a smaller anchor point graph formed by anchor points, the number of entities is reduced, the efficiency is improved, and quick and cross-platform heterogeneous data retrieval and integrated management can be realized.
In specific implementation, in the multi-modal data fusion method provided in the embodiment of the present invention, the step S101 of constructing a knowledge graph facing the power market may include: collecting key data in the electric power market; the format of the critical data may include structured data, semi-structured data, and unstructured data; extracting information including power market index names, power market index relationships and power market index attributes from the key data; and constructing a knowledge graph facing the electric power market according to the extracted information.
Particularly, key data related to the electric power market of market-related people are combed, and main data sources are government-related departments including a national statistical bureau, a commission for transformation (energy agency) and the like, a Beijing electric power trading center, a national power grid company, a provincial power grid company, various power generation enterprises, provincial electric power users, various power selling companies and the like. The data types and data of various market relatives are various, and comprise energy data of the national energy bureau, economic operation data of the national statistical bureau, operation data of a power grid, operation data of a power generation enterprise, transaction data and power consumption data of power consumers and power selling companies, market management data of a transaction center, analysis data of a third-party professional consulting organization and the like.
As various market relatives are involved, the data volume is huge, and the data apertures are inconsistent. The data formats involved include various types including structured data, semi-structured data, unstructured data. The structured data is from a new generation of electric power transaction platform, a financial ERP system, a scheduling system and a financial system; the semi-structured data is derived from XML files, JSON files and the like; the unstructured data mainly comprises various Word files, Excel files, PDF files, pictures, videos, audios and other format files. The data format of the various types of data involved needs to be processed. For structured data and semi-structured data, processing can be completed through tools such as D2R (DRF format converter), wrapper (format analysis tool) and the like, while unstructured data such as text class and the like needs to be extracted through certain technical means, such as related technology of natural language processing, and all data are finally converted into standard data through processing.
Then, information such as the power market index name, the power market index relationship, and the power market index attribute is extracted.
The extraction of the power market index name can take the power market operation index as a father index, and each index is decomposed layer by layer according to the index definition to form each level of sub-index. And the entity set E of all the parent indexes and the child indexes forming the knowledge graph is { index 1, index 2, … and index n }. For example, the unit type E1 is { gas turbine, coal turbine, nuclear power generator, wind turbine, photovoltaic unit }, and the electric power trading index system E2 is { medium and long term market trading, spot market trading, power generation right market }. As shown in fig. 3, by index extraction, it can be known that the unit 1 is a gas unit, the unit 2 is a coal-fired unit, and the unit 3 is a nuclear power unit.
The power market indicator relationship extraction is performed from two directions. Firstly, extracting the dependency relationship between a father index and a son index according to the index definition; and secondly, extracting the association relationship between the father index and the father index according to the possibly same child indexes under different father indexes. As shown in fig. 3, by the relationship extraction, the (unit 1, a1, B1) is obtained, it is known that the unit type of the unit 1 is a gas unit, and the spot market transaction is participated in.
The electric power market index attribute extraction mainly extracts the index levels according to the index definition, the electric power market operation index is a first-level index, and each sub-index corresponds to a corresponding sub-level.
In specific implementation, in the above multi-modal data fusion method provided by the embodiment of the present invention, after the step S101 is executed to construct the knowledge graph facing the power market, the method may further include: coreference resolution and entity disambiguation are performed on the knowledge graph to fuse knowledge.
Specifically, the common meaning analysis is mainly used for solving the problem that a plurality of designated items correspond to the same entity object, for example, entities of 'Zhe energy', 'Zhe energy chang' and 'chang power plant' in an index entity set are the same and can be uniformly named as 'chang power plant'; the entity disambiguation is to process the problem that the entity of a part of indexes is unknown, and if entities of a power generation enterprise, a power plant and a power generation company in the index entity set are the same, the entity disambiguation can be uniformly named as a power generation enterprise.
It should be understood that the ontology construction of the knowledge graph refers to a three-group representation structure of indexes, as shown in fig. 4, taking "day-ahead market data index" ontology construction as an example, an entity set of the indexes is { day-ahead market data index, unit parameter index, power generation side declaration index, clear data index, unit starting cost, generator set declaration electricity price, generator set declaration output, electric energy node electricity price of each transaction period of the day-ahead market }, a relationship set is { inclusion }, and an attribute set is { primary, secondary, tertiary }. The line segment in FIG. 4 shows the parent and child index dependencies; the circle size and color distinguish the attribute characterizing the index hierarchy.
In specific implementation, in the above multi-modal data fusion method provided in the embodiment of the present invention, the step S102 of preprocessing the entity features in the knowledge graph may include: converting the picture into a feature vector by adopting pre-trained fast-RCNN; and converting the text into the feature vector by adopting a pre-trained BERT model.
Specifically, for pictures, the picture content is converted into vector representation by adopting pre-trained fast-RCNN; for text, the present invention characterizes documents as vectors using a pre-trained BERT model. The processing mode converts various types of data into uniform mathematical symbols, and is beneficial to subsequent knowledge fusion.
In a specific implementation manner, in the above-mentioned multi-modal data fusion method provided in the embodiment of the present invention, the identifying, in step S103, an anchor entity from the knowledge graph may include: and extracting key entities from the pictures and the question texts which are converted into the feature vectors to serve as anchor entities.
Specifically, for the question text, the extraction of the key Entity is actually the keyword extraction task, and the nouns, verbs and Named entities in the key Entity are extracted by using some existing tools such as Stanford NLP Dependency Parser and Stanford Named Entity Recognizer. For the picture, extracting the entities in the picture needs to use a target detection technology, detect the objects in the picture by using pre-trained fast-RCNN, and simultaneously obtain the names of the objects, and align the names to the entities in the knowledge graph. The invention takes the key entity extracted from the image and text as the anchor entity.
In specific implementation, in the multi-modal data fusion method provided in the embodiment of the present invention, the fusing processing performed on the anchor entity information in step S103 may include: acquiring the importance degree of neighbors and relations corresponding to the anchor entity; and (4) taking the importance degree as weight, and fusing the information of the multi-order neighbors in a graph neural network mode to serve as the fused anchor point entity.
In particular, to reduce the number of map entities, increasing computational efficiency, the present invention therefore chooses to aggregate anchor entities in the map. Aggregation is to be performed by weighted summation, and for an anchor entity e, its associated entity and relationship pair is denoted as (e)i,ri) Two rules are given: the first rule is that the more similar the question text vector, the higher the importance level; the second rule is that if the neighbor is also the anchor entity, the importance level is high.
Based on this, all importance levels s (r) that need to aggregate neighbor entities and relationships for anchor entity ei,ei) Written as the formula:
wherein E is0A set of anchor entities is represented that is,
is represented by (e, e)i) Learnable vector of relationships, hqIs the last hidden layer feature extracted using LSTM, representing the feature vector of the entire problem text, I ═ 1,2, …, n, representing the I-th anchor entity. After the importance levels are modeled, aggregation can be performed by the importance levels. A shallow graph neural network mode is adopted, information of multi-order neighbors can be aggregated, and various information is mapped to a uniform representation space, so that knowledge aggregation is realized, and operation of downstream tasks is facilitated, and the method specifically comprises the following steps:
firstly, aggregating the neighbor information of the entities and the relations of the t-1 round to obtain the information from the neighbors of the t round
And then integrating entity information and neighbor characteristics as new entity characteristics of the t round:
first, the weights of self and neighbor are calculated
And then weighted summation is carried out to obtain the entity characteristics of the t rounds. Therefore, the anchor entities of the knowledge graph are aggregated by the algorithm, and the information in the entity field in the knowledge graph is coded into the representation vector of each entity, so that the number of the entities is reduced, the efficiency is improved, and knowledge fusion is realized.
In specific implementation, in the above multi-modal data fusion method provided by the embodiment of the present invention, while performing step S104 to convert the knowledge graph into an anchor graph composed of fused anchor entities, the method may further include: when the endpoints of the two sides in the knowledge graph are the edges on the anchor point entity, the edges are reserved in the anchor point graph; when the end points of the two sides in the knowledge graph are not on the side of the anchor point entity, deleting the side; when one end point in the knowledge graph is on the anchor point entity u and the other end point is on the side of the non-anchor point entity v, searching another anchor point entity g which is closest to the non-anchor point entity v on the knowledge graph, and calculating the distance l between the non-anchor point entity v and the other anchor point entity g; if the distance l is smaller than a set value h, adding an anchor point entity u to one edge of another anchor point entity g in the anchor point graph; if the distance is not less than the set value, the edge is deleted.
In specific implementation, in the above multi-modal data fusion method provided by the embodiment of the present invention, after performing step S104 to convert the knowledge-graph into an anchor point graph composed of fused anchor point entities, the method may further include: and quantizing the confidence degrees of the anchor point entities in the anchor point graph, and discarding the anchor point entities corresponding to the confidence degrees smaller than the preset threshold value. Therefore, the quality of the anchor point diagram is guaranteed by abandoning the knowledge with lower confidence coefficient.
Further, after the anchor point image is obtained, the index coverage, the index hierarchy depth and the correlation strength between the father index and the father index can be evaluated to ensure the quality of the image map, and the evaluation is mainly completed based on the judgment of related service experts. Taking a 'day-ahead market data index' as an example, according to actual services, the sub-indexes 'power generation side declaration data' can be further supplemented into 'a generator set declaration minimum power price', 'a generator set declaration maximum power price', 'a generator set declaration median power price', 'a generator set declaration average power price', 'a generator set declaration minimum output', 'a generator set declaration maximum output', 'different types of generator set declaration total output', 'declaration total output of the same price section', and the three-level indexes are further enriched.
It can be understood that after two steps of index extraction and knowledge fusion, a complete index entity set and an attribute set can be established, knowledge reasoning can be performed on the index entity set and the attribute set, only the index relation set is perfected, and implicit knowledge needs to be further mined on the basis of index definition, so that a knowledge base is enriched and expanded. The index relation reasoning is used for mining value information from two types of original information, namely index business logic and index historical data.
The index-based business logic mainly excavates the actual relation between the business and the business by understanding the business corresponding to each level of indexes in the entity set, thereby enriching the incidence relation between the indexes and the indexes. For example, the income of the power generation enterprise is the income of power selling, so that the 'generating capacity of the unit' and the 'trading electric quantity' have an association relation on the service logic, and further the father index 'power supply side index' and the father index 'power trading index' have an association relation.
And (3) mining a potential relation based on the index historical data by adopting mathematical statistical methods such as cluster analysis, correlation analysis, association rules and the like and on the basis of the index historical data of each level. For example, based on historical data of two indexes of "supply-demand ratio" and "medium-and-long term monthly price", a correlation coefficient is calculated, and the two indexes are highly positively correlated, namely, the "supply-demand ratio" and the "medium-and-long term monthly price" have a certain positive correlation.
FIG. 5 shows an anchor point diagram frame corresponding to the electric power market index fusion relationship. The boxes in the diagram represent specific indexes, namely the entities of the knowledge graph, the hierarchy attributes of the indexes are distinguished through arrows and colors, and the relation among the indexes is represented through a dotted line connection. The map can visually display the structural relationship between the father and son indexes and the mutual relationship between the father and father indexes.
Correspondingly, the embodiment of the invention also discloses a multi-mode data fusion device facing the electric power market, which comprises a processor and a memory; wherein the processor, when executing the computer program stored in the memory, implements the multimodal data fusion method disclosed in the foregoing embodiments.
For more specific processes of the above method, reference may be made to corresponding contents disclosed in the foregoing embodiments, and details are not repeated here.
Further, the present invention also discloses a computer readable storage medium for storing a computer program; the computer program, when executed by a processor, implements the multimodal data fusion method disclosed above.
For more specific processes of the above method, reference may be made to corresponding contents disclosed in the foregoing embodiments, and details are not repeated here.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other. The device and the storage medium disclosed by the embodiment correspond to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
To sum up, the multi-modal data fusion method for the power market provided by the embodiment of the invention comprises the following steps: constructing a knowledge graph facing the electric power market; entities of the knowledge graph represent various indicators in the electricity market; the edges of the knowledge graph represent the relationship between the indexes; preprocessing entity features in the knowledge graph to convert various data into uniform multidimensional vectors; identifying an anchor entity from the knowledge graph, and fusing anchor entity information; and converting the knowledge graph into an anchor point graph formed by the fused anchor point entities. The invention adopts the knowledge graph to realize multi-mode fusion in the electric power market, expresses various indexes into a graph structure, and represents the indexes as uniform multi-dimensional vectors, thereby promoting the effective integration of multi-mode multi-field data, converts the large knowledge graph into a smaller anchor point graph consisting of anchor points, reduces the number of entities, improves the efficiency, and can realize quick and cross-platform heterogeneous data retrieval and integrated management. In addition, the invention also provides corresponding equipment and a computer readable storage medium aiming at the multi-mode data fusion method, so that the method has higher practicability, and the equipment and the computer readable storage medium have corresponding advantages.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The multimodal data fusion method, device and medium for the power market provided by the invention are described in detail above, and a specific example is applied in the text to explain the principle and the implementation of the invention, and the description of the above embodiment is only used to help understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (10)
1. A multimodal data fusion method oriented to a power market is characterized by comprising the following steps:
constructing a knowledge graph facing the electric power market; entities of the knowledge graph represent a plurality of indicators in the electricity market; the edges of the knowledge graph represent the relationship between the indexes;
preprocessing the entity characteristics in the knowledge graph to convert various data into uniform multidimensional vectors;
identifying an anchor entity from the knowledge graph, and fusing anchor entity information;
and converting the knowledge graph into an anchor point graph formed by the fused anchor point entities.
2. The multimodal data fusion method of claim 1, wherein the building a power market oriented knowledge graph comprises:
collecting key data in the electric power market; the format of the key data comprises structured data, semi-structured data and unstructured data;
extracting information including power market index names, power market index relationships and power market index attributes from the key data;
and constructing a knowledge graph facing the electric power market according to the extracted information.
3. The multimodal data fusion method of claim 2 wherein the preprocessing of the entity features in the knowledge-graph comprises:
converting the picture into a feature vector by adopting pre-trained fast-RCNN;
and converting the text into the feature vector by adopting a pre-trained BERT model.
4. The multimodal data fusion method of claim 3 wherein the identifying an anchor entity from the knowledge-graph comprises:
and extracting key entities from the pictures and the question texts which are converted into the feature vectors to serve as anchor entities.
5. The multimodal data fusion method of claim 4 wherein the fusing anchor entity information comprises:
acquiring the importance degree of neighbors and relations corresponding to the anchor entity;
and taking the importance degree as weight, and fusing the information of the multi-order neighbors in a graph neural network mode to serve as a fused anchor point entity.
6. The multimodal data fusion method of claim 5 further comprising, while converting the knowledge-graph into an anchor graph of fused anchor entities:
when the end points of the knowledge graph are edges on the anchor point entity, keeping the edges in the anchor point graph;
when the endpoints of the two sides in the knowledge graph are not on the edge of the anchor point entity, deleting the edge;
when one end point in the knowledge graph is on the anchor point entity and the other end point is on the side of the non-anchor point entity, searching another anchor point entity closest to the non-anchor point entity on the knowledge graph, and calculating the distance between the non-anchor point entity and the other anchor point entity; if the distance is smaller than a set value, adding an anchor point entity to one edge of another anchor point entity in the anchor point graph; if the distance is not less than the set value, the edge is deleted.
7. The multimodal data fusion method of claim 6, further comprising, after constructing the power market oriented knowledge graph:
performing coreference resolution and entity disambiguation on the knowledge graph to fuse knowledge.
8. The multimodal data fusion method of claim 7 further comprising, after converting the knowledge-graph into an anchor graph of fused anchor entities:
and quantizing the confidence degrees of the anchor point entities in the anchor point graph, and discarding the anchor point entities corresponding to the confidence degrees smaller than a preset threshold value.
9. A multimodal data fusion apparatus oriented to an electric power market, comprising a processor and a memory, wherein the processor, when executing a computer program stored in the memory, implements the multimodal data fusion method of any one of claims 1 to 8.
10. A computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the multimodal data fusion method of any of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111612678.1A CN114281878A (en) | 2021-12-27 | 2021-12-27 | Multimode data fusion method, device and medium for power market |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111612678.1A CN114281878A (en) | 2021-12-27 | 2021-12-27 | Multimode data fusion method, device and medium for power market |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114281878A true CN114281878A (en) | 2022-04-05 |
Family
ID=80876245
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111612678.1A Pending CN114281878A (en) | 2021-12-27 | 2021-12-27 | Multimode data fusion method, device and medium for power market |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114281878A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115329158A (en) * | 2022-10-17 | 2022-11-11 | 湖南能源大数据中心有限责任公司 | Data association method based on multi-source heterogeneous power data |
| CN115829058A (en) * | 2022-12-23 | 2023-03-21 | 北京百度网讯科技有限公司 | Training sample processing method, cross-modal matching method, device, equipment and medium |
| CN117034126A (en) * | 2023-10-09 | 2023-11-10 | 广东电力交易中心有限责任公司 | Green power user classification and feature extraction method and system based on knowledge graph |
-
2021
- 2021-12-27 CN CN202111612678.1A patent/CN114281878A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115329158A (en) * | 2022-10-17 | 2022-11-11 | 湖南能源大数据中心有限责任公司 | Data association method based on multi-source heterogeneous power data |
| CN115829058A (en) * | 2022-12-23 | 2023-03-21 | 北京百度网讯科技有限公司 | Training sample processing method, cross-modal matching method, device, equipment and medium |
| CN115829058B (en) * | 2022-12-23 | 2024-04-23 | 北京百度网讯科技有限公司 | Training sample processing method, cross-modal matching method, device, equipment and medium |
| CN117034126A (en) * | 2023-10-09 | 2023-11-10 | 广东电力交易中心有限责任公司 | Green power user classification and feature extraction method and system based on knowledge graph |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114281878A (en) | Multimode data fusion method, device and medium for power market | |
| Liu et al. | Some q‐rung orthopair fuzzy aggregation operators and their applications to multiple‐attribute decision making | |
| Chatterjee et al. | A multi-criteria decision making for renewable energy selection using Z-numbers in uncertain environment | |
| Mateo | Multi criteria analysis in the renewable energy industry | |
| Krishnan et al. | Efficient database generation for decision tree based power system security assessment | |
| Wu et al. | Risk assessment of wind-photovoltaic-hydrogen storage projects using an improved fuzzy synthetic evaluation approach based on cloud model: A case study in China | |
| Kuleli Pak et al. | Renewable energy perspective for Turkey using sustainability indicators | |
| Zhou | Fuzzy analytical network process implementation with matlab | |
| CN111552813A (en) | Power knowledge graph construction method based on power grid full-service data | |
| CN111260505B (en) | Big data analysis method and device based on power Internet of things and computer equipment | |
| CN111061679A (en) | Method and system for rapid configuration of technological innovation policy based on rete and drools rules | |
| CN111327046A (en) | Economic dispatching method for power system containing large-scale wind power scene | |
| CN107220742B (en) | Information system development general supporting method and platform based on system vulnerability analysis | |
| CN116596574A (en) | Power grid user portrait construction method and system | |
| CN115827885A (en) | Operation and maintenance knowledge graph construction method and device and electronic equipment | |
| Yu et al. | Forecasting digital economy of China using an Adaptive Lasso and grey model optimized by particle swarm optimization algorithm | |
| Xiao et al. | Comprehensive Evaluation Index System of Distribution Network for Distributed Photovoltaic Access | |
| CN115544871A (en) | Distributed robust energy storage planning method considering renewable power supply space correlation | |
| CN115048528A (en) | Method and device for constructing knowledge graph of new energy electric field operation data | |
| CN106097141A (en) | The ternary interval number appraisal procedure of photovoltaic generating system confidence capacity | |
| Ma et al. | Knowledge graph based recommendation algorithm for educational resource | |
| Balin et al. | A Multi-Criteria Decision-Making Methodology Suggestion for Turkey Energy Planning Based Type-2 Fuzzy Sets | |
| CN110837922A (en) | Global energy comprehensive optimization configuration method considering national risk | |
| Li et al. | Integrating unimodality into distributionally robust optimal power flow | |
| CN117934209B (en) | Regional power system carbon emission big data analysis method based on knowledge graph |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |