CN117235265A - Processing system and method for power data file - Google Patents

Abstract

The invention provides a processing system and a processing method of an electric power data file, and relates to the technical field of data processing. In the invention, for each power data file to be stored, marking the power data file to be stored as a power data file to be processed; analyzing target power abnormality characterization data corresponding to a power data file to be processed by utilizing a plurality of power data analysis networks; performing first classification processing on a plurality of power data files to be stored based on corresponding target power abnormality characterization data to form at least one first classification set; based on the similarity between the power data files to be stored, respectively carrying out second classification processing in each first classification set to form at least one second classification set corresponding to each first classification set; and respectively storing each second classification set in a classification way. Based on the method, the reliability of classified storage can be improved.

Description

A system and method for processing electric power data files

Technical field

The present invention relates to the field of data processing technology, and specifically to a system and method for processing electric power data files.

Background technique

As the accuracy of data processing technology becomes higher and higher, the application scenarios of data processing technology are also constantly expanding. For example, it can be used in the field of electric power. Specifically, the collected or formed electric power data files can be Data processing technology is used to perform data feature analysis to determine the correlation or difference between data, so that classified storage can be performed. However, in the existing technology, there is a problem of poor reliability of classified storage.

Contents of the invention

In view of this, the object of the present invention is to provide a power data file processing system and method to improve the reliability of classified storage.

In order to achieve the above objects, the embodiments of the present invention adopt the following technical solutions:

A method for processing power data files, including:

For each of the plurality of power data files to be stored, marking the power data file to be stored as a power data file to be processed;

Utilize multiple power data analysis networks to analyze the target power abnormality characterization data corresponding to the power data file to be processed, and the target power abnormality characterization data is used to reflect the abnormal state of the power system corresponding to the power data file to be processed;

Based on the corresponding target power abnormality characterization data, a first classification process is performed on the plurality of power data files to be stored to form at least one first classification set, each of the first classification sets including at least one power data file to be stored. ;

Based on the similarity between the power data files to be stored, a second classification process is performed within each of the first classification sets to form at least one second classification set corresponding to each of the first classification sets. The two-category set includes at least one power data file to be stored;

Each obtained second classification set is classified and stored separately.

In some preferred embodiments, in the above method for processing power data files, the plurality of power data files to be stored are subjected to a first classification process based on the corresponding target power abnormality characterization data to form at least a first classification process. The steps for classifying a collection include:

Perform consistency or similarity analysis on the target power abnormality characterization data corresponding to each two power data files to be stored in the plurality of power data files to be stored;

The analyzed power data files to be stored whose corresponding target power abnormality characterization data are consistent or whose corresponding target power abnormality characterization data belong to the same parameter interval are assigned to the same first classification set to form At least one first category set.

In some preferred embodiments, in the above method for processing power data files, the second classification process is performed within each of the first classification sets based on the similarity between the power data files to be stored, forming The steps of at least one second classification set corresponding to each first classification set include:

For each of the first classification sets, a quantitative statistical operation is performed on the power data files to be stored included in the first classified set to form a corresponding statistical value of the number of files, and when the statistical value of the number of files is less than or equal to a predetermined In the case of the first reference value, the first classification set is determined as the corresponding second classification set, and in the case where the file quantity statistical value is greater than the first reference value, the first classification set is determined as The corresponding third classification set;

Based on the similarity between the power data files to be stored, a second classification process is performed within each of the third classification sets to form at least one second classification set corresponding to each of the third classification sets.

In some preferred embodiments, in the above method for processing power data files, the second classification process is performed within each of the third classification sets based on the similarity between the power data files to be stored, forming The steps of at least one second classification set corresponding to each third classification set include:

Perform a keyword extraction operation on each power data file to be stored in the third classification set to form a keyword sequence corresponding to each power data file to be stored. Each keyword in the keyword sequence belongs to A collection of reference keywords for power system domain configuration;

Perform feature mining processing on the corresponding keyword sequences respectively to form a keyword feature representation corresponding to the power data file to be stored;

Based on the keyword feature representation, the similarity between the corresponding power data files to be stored is calculated, and based on the similarity between the power data files to be stored, it is performed within each of the third classification sets. The second classification process forms at least one second classification set corresponding to each third classification set.

In some preferred embodiments, in the above method for processing power data files, the corresponding keyword sequences are respectively subjected to feature mining processing to form a keyword feature representation corresponding to the power data file to be stored. steps, including:

For each keyword in the keyword sequence, embedding processing is performed on the keyword to form a word embedding feature representation corresponding to the keyword; and, based on the target power data corpus, determining the keyword in the keyword sequence For each keyword, whether there are related keywords in the keyword sequence, and the co-occurrence probability of the related keywords and the corresponding keywords in the target power data corpus is greater than the preset probability;

Each keyword in the keyword sequence that does not have a relevant keyword is marked as a first keyword, and each keyword in the keyword sequence that has a relevant keyword is marked as a second keyword. words, and mark the word embedding feature representation of each first keyword to mark the target word embedding feature representation of the first keyword;

For each second keyword, mark the word embedding feature representation corresponding to the related keyword corresponding to the second keyword to mark the related word embedding feature representation corresponding to the second keyword, and, The related word embedding feature representation is subjected to a transposition operation to form a transposed word embedding feature representation corresponding to the second keyword;

Calculate the word embedding feature representation corresponding to each of the second keywords, the corresponding transposed word embedding feature representation and the related word embedding feature representation and perform a fusion operation to form a target word embedding corresponding to each of the second keywords. Feature representation;

The target word embedding feature representation corresponding to each keyword in the keyword sequence is spliced to form a keyword feature representation corresponding to the power data file to be stored.

In some preferred embodiments, in the above method for processing power data files, the similarity between corresponding power data files to be stored is calculated based on the keyword feature representation, and the similarity between the corresponding power data files to be stored is calculated based on the power to be stored. The step of performing a second classification process on each of the third classification sets based on the similarity between the data files and forming at least one second classification set corresponding to each of the third classification sets includes:

Perform mean calculation on the keyword feature representation corresponding to each power data file to be stored in the third classification set to output the corresponding mean keyword feature representation;

For each power data file to be stored in the third classification set, calculate the cosine similarity between the keyword feature representation corresponding to the power data file to be stored and the mean keyword feature representation to obtain the power to be stored Cosine similarity corresponding to the data file;

Based on multiple continuous similarity intervals configured by cosine, perform a second classification process on each power data file to be stored in the third classification set to form at least one second classification set corresponding to the third classification set, In the second classification set, the cosine similarity corresponding to each included power data file to be stored belongs to the same similarity interval.

In some preferred embodiments, in the above method for processing a power data file, the step of using multiple power data analysis networks to analyze the target power abnormality characterization data corresponding to the power data file to be processed includes:

Utilizing multiple power data analysis networks to perform feature mining operations on the power data files to be processed to output corresponding multiple initial data feature representations, each of the multiple power data analysis networks is used to perform feature mining based on the loaded data, output corresponding power abnormality characterization data, and the power data file to be processed belongs to the operating text data of the power system;

Perform a feature representation fusion operation on the plurality of initial data feature representations to form a corresponding aggregated data feature representation;

Based on the aggregated data feature representation, the target power abnormality characterization data corresponding to the power data file to be processed is analyzed, and the target power abnormality characterization data is used to reflect the abnormal state of the power system corresponding to the power data file to be processed.

An embodiment of the present invention also provides a power data file processing system, including:

A data file marking module, configured to mark each of the plurality of power data files to be stored as a power data file to be stored as a power data file to be processed;

A power anomaly analysis module is used to utilize multiple power data analysis networks to analyze the target power anomaly characterization data corresponding to the power data file to be processed, and the target power anomaly characterization data is used to reflect the corresponding power data file to be processed. Abnormal status of the power system;

A first classification processing module, configured to perform first classification processing on the plurality of power data files to be stored based on the corresponding target power abnormality characterization data to form at least one first classification set, each of the first classification sets Includes at least one power data file to be stored;

The second classification processing module is configured to perform second classification processing within each of the first classification sets based on the similarity between the power data files to be stored, and form at least one corresponding to each of the first classification sets. second classification sets, each second classification set including at least one power data file to be stored;

A classification storage module, configured to classify and store each obtained second classification set respectively.

In some preferred embodiments, in the above power data file processing system, the first classification processing module is specifically used to:

Perform consistency or similarity analysis on the target power abnormality characterization data corresponding to each two power data files to be stored in the plurality of power data files to be stored;

The analyzed power data files to be stored whose corresponding target power abnormality characterization data are consistent or whose corresponding target power abnormality characterization data belong to the same parameter interval are assigned to the same first classification set to form At least one first category set.

In some preferred embodiments, in the above power data file processing system, the second classification processing module is specifically used to:

For each of the first classification sets, a quantitative statistical operation is performed on the power data files to be stored included in the first classified set to form a corresponding statistical value of the number of files, and when the statistical value of the number of files is less than or equal to a predetermined In the case of the first reference value, the first classification set is determined as the corresponding second classification set, and in the case where the file quantity statistical value is greater than the first reference value, the first classification set is determined as The corresponding third classification set;

Based on the similarity between the power data files to be stored, a second classification process is performed within each of the third classification sets to form at least one second classification set corresponding to each of the third classification sets.

An embodiment of the present invention provides a system and method for processing power data files. For each power data file to be stored, the power data file to be stored is marked as a power data file to be processed; multiple power data analyzes are used The network analyzes the target power abnormality characterization data corresponding to the power data file to be processed; based on the corresponding target power abnormality characterization data, performs first classification processing on multiple power data files to be stored to form at least one first classification set; based on The similarity between the power data files to be stored is subjected to second classification processing within each first classification set to form at least one second classification set corresponding to each first classification set; each second classification set is processed separately Carry out classified storage. Based on the foregoing content, two-level classification processing can be performed through the similarity between data files and the corresponding target power anomaly characterization data, making the classification more accurate and thereby improving the reliability of classification storage.

In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, preferred embodiments are given below and described in detail with reference to the accompanying drawings.

Description of drawings

Figure 1 is a structural block diagram of a power data file processing platform provided by an embodiment of the present invention.

FIG. 2 is a schematic flowchart of each step included in the power data file processing method provided by an embodiment of the present invention.

FIG. 3 is a schematic diagram of each module included in the power data file processing system provided by the embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, 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 described embodiments These are only some of the embodiments of the present invention, not all of them. The components of the embodiments of the invention generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations.

Therefore, the following detailed description of the embodiments of the invention provided in the appended drawings is not intended to limit the scope of the claimed invention, but rather to represent selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.

As shown in Figure 1, an embodiment of the present invention provides a power data file processing platform. Wherein, the power data file processing platform may include a memory and a processor.

In detail, the memory and the processor are electrically connected directly or indirectly to realize data transmission or interaction. For example, they can be electrically connected to each other through one or more communication buses or signal lines. The memory may store at least one software function module (computer program) that may exist in the form of software or firmware. The processor may be configured to execute an executable computer program stored in the memory, thereby implementing the power data file processing method provided by embodiments of the present invention (as described later).

Alternatively, in some implementations, the memory may be, but is not limited to, random access memory (Random Access Memory, RAM), read only memory (Read Only Memory, ROM), programmable read only memory ( Programmable Read-Only Memory (PROM), Erasable ProgrammableRead-Only Memory (EPROM), Electrically Erasable ProgrammableRead-Only Memory (EEPROM), etc.

Alternatively, in some implementations, the processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), a system on a chip (System on Chip). Chip, SoC), etc.; it can also be a digital signal processor (DSP), application specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

Alternatively, in some implementations, the power data file processing platform may be a server with data processing capabilities.

With reference to FIG. 2 , an embodiment of the present invention also provides a method for processing power data files, which can be applied to the above-mentioned power data file processing platform. Wherein, the method steps defined in the process related to the power data file processing method can be implemented by the power data file processing platform.

The specific process shown in Figure 2 will be elaborated below.

Step S100: For each power data file to be stored among the multiple power data files to be stored, mark the power data file to be stored as a power data file to be processed.

In an embodiment of the present invention, the power data file processing platform may mark each of the multiple power data files to be stored as power data to be processed. files (can be done sequentially or in parallel).

Step S200: Use multiple power data analysis networks to analyze target power abnormality characterization data corresponding to the power data file to be processed.

In an embodiment of the present invention, the power data file processing platform can utilize multiple power data analysis networks to analyze the target power abnormality characterization data corresponding to the power data file to be processed. The target power abnormality characterization data is used to reflect the abnormal state of the power system corresponding to the power data file to be processed.

Step S300: Perform a first classification process on the plurality of power data files to be stored based on the corresponding target power abnormality characterization data to form at least one first classification set.

In an embodiment of the present invention, the power data file processing platform may perform a first classification process on the plurality of power data files to be stored based on the corresponding target power anomaly characterization data to form at least one first classification set. Each of the first classification sets includes at least one power data file to be stored.

Step S400: Based on the similarity between the power data files to be stored, perform a second classification process within each of the first classification sets to form at least one second classification set corresponding to each of the first classification sets.

In an embodiment of the present invention, the power data file processing platform may perform second classification processing within each of the first classification sets based on the similarity between the power data files to be stored, forming each of the first classification sets. At least one second classification set corresponding to the first classification set. Each second classification set includes at least one power data file to be stored.

Step S500: Classify and store each obtained second classification set respectively.

In this embodiment of the present invention, the power data file processing platform may separately classify and store each of the obtained second classification sets. In this way, each power data file to be stored in a second classification set is stored in the same storage device to facilitate subsequent calls.

Based on the foregoing content, that is, the above-mentioned steps S100 to S500, two-level classification processing can be performed through the similarity between data files and the corresponding target power abnormality characterization data, so that the classification accuracy is higher, thereby improving the efficiency of classification storage. Reliability.

Optionally, in some embodiments, the step of using multiple power data analysis networks to analyze the target power abnormality characterization data corresponding to the power data file to be processed may further include the following content, such as step S110. , step S120 and step S130.

Step S110: Use multiple power data analysis networks to perform feature mining operations on the power data files to be processed to output multiple corresponding initial data feature representations.

In an embodiment of the present invention, the power data file processing platform can utilize multiple power data analysis networks to perform feature mining operations on the power data files to be processed to output multiple corresponding initial data feature representations. Each power data analysis network in the plurality of power data analysis networks is used to output corresponding power abnormality characterization data based on the loaded data, and the power data file to be processed belongs to the operating text data of the power system, that is to say , the power data file to be processed is used to describe the operating process of the power system.

Step S120: Perform a feature representation fusion operation on the plurality of initial data feature representations to form a corresponding aggregated data feature representation.

In this embodiment of the present invention, the power data file processing platform may perform a feature representation fusion operation on the multiple initial data feature representations to form a corresponding aggregated data feature representation.

Step S130: Based on the aggregated data feature representation, analyze the target power abnormality characterization data corresponding to the power data file to be processed.

In an embodiment of the present invention, the power data file processing platform can analyze the target power abnormality characterization data corresponding to the power data file to be processed based on the aggregated data feature representation, and the target power abnormality characterization data is used for Reflect the abnormal status of the power system corresponding to the power data file to be processed, such as whether it is abnormal, the degree of the abnormality, etc.

Based on the above content, since multiple power data analysis networks are first used to perform feature mining operations, multiple initial data feature representations can be obtained. In this way, the aggregated data feature representation of the target power anomaly representation data analyzed by the user can be further integrated to obtain, In other words, the basis for conducting power abnormality analysis is more sufficient, and therefore the reliability of power data analysis can be improved.

Optionally, in some embodiments, the step of using multiple power data analysis networks to perform feature mining operations on the power data files to be processed to output corresponding multiple initial data feature representations may further include the following content:

Determine the power data file to be processed, and analyze the power data file segments to be processed included in the power data file to be processed. For example, the power data file to be processed can be split to form corresponding power data files to be processed. Processing power data file fragments, for example, when forming multiple power data file fragments to be processed, there may be no temporal sequence relationship between the power data file fragments to be processed, that is, they respectively reflect the operating data of the power system at different times. Alternatively, each to-be-processed power data file fragment may have an on-device correspondence relationship, that is, each of them reflects the operating data of different power equipment;

The to-be-processed power data file fragment is marked to be loaded into each of the plurality of power data analysis networks. For example, one of the plurality of power data analysis networks The network parameters between different power data analysis networks can be different. For example, the size of the filter matrix between different power data analysis networks can be different, the network architecture can also be different, and the number of filter matrices between different power data analysis networks can be different;

Utilize each of the plurality of power data analysis networks to mine an initial data feature representation among a plurality of initial data feature representations, where an initial data feature representation includes an initial feature corresponding to a segment of the power data file to be processed. express.

Optionally, in some implementations, the power data file to be processed includes a plurality of power data file fragments to be processed. Based on this, the power data in the network is analyzed using the plurality of power data. The step of analyzing the network and mining out one initial data feature representation among multiple initial data feature representations may further include the following:

Utilize a data mining sub-network included in each of the power data analysis networks to perform data mining operations on multiple power data file segments to be processed, and the data mining sub-network is used to perform data mining operations on multiple power data file segments to be processed. Multiple initial feature representations are mined from the fragments, and the data mining operation may refer to mapping and filtering of feature space, etc.;

Determine the correlation description data of multiple power data file segments to be processed, and the correlation description data is used to reflect the distribution correlation relationship of the multiple power data file segments to be processed in the power data file to be processed. , such as forming a sequential relationship between times;

Based on the correlation description data, a correlation mining operation is performed on a plurality of the initial feature representations to output corresponding initial data feature representations.

Optionally, in some embodiments, the step of using a data mining sub-network included in each of the power data analysis networks to perform data mining operations on multiple power data file fragments to be processed may further include: The following content:

Load multiple power data file fragments to be processed into multiple data mining sub-networks included in each of the power data analysis networks, and the multiple data mining sub-networks are used to perform analysis on multiple data mining sub-networks. Multiple sets of intermediate feature representations are mined from the power data file fragments to be processed. There is a one-to-one correspondence between the multiple data mining sub-networks and the multiple sets of intermediate feature representations. Each set of intermediate feature representations in the multiple sets of intermediate feature representations The feature representation includes a plurality of intermediate feature representations, and there is a one-to-one correspondence between the plurality of intermediate feature representations and the plurality of power data file fragments to be processed;

Among the multiple sets of intermediate feature representations, the intermediate feature representations corresponding to the same power data file fragment to be processed are merged to form multiple initial feature representations. For example, the same power data file fragment to be processed can be corresponding to The intermediate feature representation is spliced.

Optionally, in some embodiments, the step of performing correlation mining operations on multiple initial feature representations based on the correlation relationship description data to output corresponding initial data feature representations may further include the following: content:

Based on the correlation description data, load the initial feature representation in order to load it into the data association mining unit;

Based on the data association mining unit, the associated data feature representation is mined. For example, the data association mining unit can describe data according to the relevant relationship and splice multiple initial feature representations to form corresponding associated data. Feature representation;

Utilize a focused feature analysis unit to perform a focused feature analysis operation on the associated data feature representation to output a plurality of data feature representations to be processed. The focused feature analysis unit is used to characterize important content based on each power data file segment to be processed. parameters to analyze the data feature representation to be processed corresponding to each power data file segment to be processed. For example, based on the adjacent initial feature representation, the initial feature representation in the associated data feature representation can be compared between modalities. Focus feature analysis operation to obtain the corresponding feature representation of the data to be processed, wherein the focus feature weight parameter obtained by performing the focus feature analysis operation can be used as the important content representation parameter, so that weighting can be performed based on the important content representation parameter, To obtain the corresponding feature representation of the data to be processed;

Utilize the feature integration unit included in each of the power data analysis networks to perform a feature integration operation on the plurality of data feature representations to be processed to output the corresponding initial data feature representation. The processing process of the feature integration unit can be related to The process of feature mining is reversed, such as performing inverse filtering (such as upsampling) to obtain the corresponding initial data feature representation.

Optionally, in some embodiments, the power data file to be processed includes a plurality of power data file fragments to be processed; each of the multiple initial data feature representations includes a file corresponding to a plurality of the power data file segments to be processed. Process multiple preliminary feature representations that have a one-to-one correspondence between power data file segments. Based on this, the step of performing a feature representation fusion operation on the multiple initial data feature representations to form a corresponding aggregated data feature representation , can further include the following content:

Among the plurality of initial data feature representations, a plurality of preliminary feature representation clusters having a one-to-one correspondence with a plurality of the power data file fragments to be processed are selected, and each of the plurality of preliminary feature representation clusters is A preliminary feature representation cluster includes a preliminary feature representation corresponding to one of the plurality of power data file segments to be processed in the plurality of initial data feature representations;

Determining the mean preliminary feature representation of each preliminary feature representation cluster in the plurality of preliminary feature representation clusters (that is, performing a mean superposition on each preliminary feature representation in the preliminary feature representation cluster to obtain a mean preliminary feature representation), Represented by multiple mean preliminary features that have a one-to-one correspondence between the output and multiple power data file segments to be processed;

The feature representations including the plurality of mean preliminary feature representations are marked to be marked as corresponding aggregated data feature representations. That is to say, the aggregated data feature representation may include the plurality of mean preliminary feature representations.

Optionally, in some embodiments, the power data file to be processed includes a plurality of power data file fragments to be processed; each of the multiple initial data feature representations includes a combination of the multiple initial data feature representations. A plurality of preliminary feature representations having a one-to-one correspondence between the power data file segments to be processed are described. Based on this, the plurality of initial data feature representations are subjected to a feature representation fusion operation to form a corresponding aggregated data feature representation. The steps may further include the following:

Among the plurality of initial data feature representations, a plurality of preliminary feature representation clusters having a one-to-one correspondence with a plurality of the power data file fragments to be processed are selected, and each of the plurality of preliminary feature representation clusters is A preliminary feature representation cluster includes a preliminary feature representation corresponding to one of the plurality of power data file segments to be processed in the plurality of initial data feature representations;

Determine the most relevant preliminary feature representation of each preliminary feature representation cluster in the plurality of preliminary feature representation clusters (that is, cluster all preliminary feature representations in the preliminary feature representation cluster to determine the cluster center, as The most relevant preliminary feature representation), to output a plurality of most relevant preliminary feature representations that have a one-to-one correspondence with a plurality of the power data file fragments to be processed;

Mark the feature representation including the plurality of most relevant preliminary feature representations to mark the corresponding aggregated data feature representation, that is to say, the aggregated data feature representation may include the plurality of most relevant preliminary feature representations, For example, the plurality of most relevant preliminary feature representations may be spliced to form a corresponding aggregated data feature representation.

Optionally, in some implementations, the power data file to be processed includes N power data file segments to be processed, and the aggregate data feature representation includes N corresponding to the N power data file segments to be processed. Feature representation. Based on this, the step of analyzing the target power abnormality characterization data corresponding to the power data file to be processed based on the aggregated data feature representation may further include the following content:

Perform a fully connected operation on the N feature representations to obtain a fully connected feature representation;

Perform similarity calculations on the fully connected feature representation and multiple central feature representations respectively to output the corresponding multiple feature representation similarities;

Determine one feature representation similarity (such as the largest feature representation similarity) among the plurality of feature representation similarities, and use the mark as the target feature representation similarity;

The reference power abnormality characterization data corresponding to the central feature representation corresponding to the similarity of the target feature representation is marked as the target power abnormality characterization data corresponding to the power data file to be processed. Each of the central feature representations is based on the corresponding reference power abnormality representation data. The characteristic representation corresponding to at least one typical power data file of the power abnormality characterization data is determined (clustering is performed to determine).

Optionally, in some embodiments, before the step of using multiple power data analysis networks to perform a feature mining operation on the power data file to be processed to output corresponding multiple initial data feature representations, the power data file The processing method can further include the following content:

Based on typical power data files (and corresponding actual power anomaly characterization data), perform a network update operation on each of the multiple power data analysis networks to be updated to form multiple corresponding updated power data analysis networks. ;

Based on the typical power data file, a network update operation is performed on each of the multiple associated networks to form corresponding multiple updated associated networks. Each of the multiple associated networks includes a The updated power data analysis network and a feature representation restoration network are used to restore the feature representation corresponding to the typical power data file based on the power abnormality characterization data analyzed by the updated power data analysis network (such as , the corresponding error parameter can be determined based on the difference between the feature representation mined by the updated power data analysis network and the feature representation restored by the feature representation restoration network, and then network update processing is performed based on the error parameter);

Based on the multiple updated associated networks, multiple power data analysis networks are determined. For example, based on the network parameters of the multiple updated associated networks, a power data analysis network is constructed.

Optionally, in some embodiments, the step of performing a first classification process on the plurality of power data files to be stored based on the corresponding target power abnormality characterization data to form at least one first classification set may be Further includes the following:

Perform consistency or similarity analysis on the target power abnormality characterization data corresponding to each two power data files to be stored in the plurality of power data files to be stored;

The analyzed power data files to be stored whose corresponding target power abnormality characterization data are consistent or whose corresponding target power abnormality characterization data belong to the same parameter interval are assigned to the same first classification set to form At least one first category set.

Optionally, in some embodiments, based on the similarity between the power data files to be stored, a second classification process is performed within each of the first classification sets to form each of the first classification sets. The step of assembling at least one second classification set corresponding to the collection may further include the following content:

For each of the first classification sets, a quantitative statistical operation is performed on the power data files to be stored included in the first classified set to form a corresponding statistical value of the number of files, and when the statistical value of the number of files is less than or equal to a predetermined In the case of the first reference value (such as 5), the first classification set is determined as the corresponding second classification set, and in the case where the file quantity statistical value is greater than the first reference value, the first classification set is determined as the corresponding second classification set. The classification set is determined as the corresponding third classification set;

Based on the similarity between the power data files to be stored, a second classification process is performed within each of the third classification sets to form at least one second classification set corresponding to each of the third classification sets.

Optionally, in some embodiments, based on the similarity between the power data files to be stored, a second classification process is performed within each of the third classification sets to form each of the third classifications. The step of assembling at least one second classification set corresponding to the collection may further include the following content:

Perform a keyword extraction operation on each power data file to be stored in the third classification set to form a keyword sequence corresponding to each power data file to be stored. Each keyword in the keyword sequence belongs to A collection of reference keywords for power system domain configuration;

Perform feature mining processing on the corresponding keyword sequences respectively to form a keyword feature representation corresponding to the power data file to be stored;

Based on the keyword feature representation, the similarity between the corresponding power data files to be stored is calculated, and based on the similarity between the power data files to be stored, it is performed within each of the third classification sets. The second classification process forms at least one second classification set corresponding to each third classification set.

Optionally, in some embodiments, the step of performing feature mining processing on the corresponding keyword sequences to form a keyword feature representation corresponding to the power data file to be stored may further include the following: content:

For each keyword in the keyword sequence, embedding processing is performed on the keyword to form a word embedding feature representation corresponding to the keyword; and, based on the target power data corpus, determining the keyword in the keyword sequence For each keyword, whether there are related keywords in the keyword sequence, and the co-occurrence probability of the related keywords and the corresponding keywords in the target power data corpus is greater than the preset probability;

Each keyword in the keyword sequence that does not have a relevant keyword is marked as a first keyword, and each keyword in the keyword sequence that has a relevant keyword is marked as a second keyword. words, and mark the word embedding feature representation of each first keyword to mark the target word embedding feature representation of the first keyword;

For each second keyword, mark the word embedding feature representation corresponding to the related keyword corresponding to the second keyword to mark the related word embedding feature representation corresponding to the second keyword, and, The related word embedding feature representation is subjected to a transposition operation to form a transposed word embedding feature representation corresponding to the second keyword;

Calculate the word embedding feature representation corresponding to each of the second keywords, the corresponding transposed word embedding feature representation and the related word embedding feature representation and perform a fusion operation to form a target word embedding corresponding to each of the second keywords. Feature representation, for example, the transposed word embedding feature representation and the word embedding feature representation can be multiplied, divided by the number of dimensions of the word embedding feature representation, and then normalized parameters, and then multiplied Use the related word embedding feature representation to achieve fusion and obtain the corresponding target word embedding feature representation;

The target word embedding feature representation corresponding to each keyword in the keyword sequence is spliced to form a keyword feature representation corresponding to the power data file to be stored.

Optionally, in some embodiments, the similarity between the corresponding power data files to be stored is calculated based on the keyword feature representation, and the similarity between the power data files to be stored is calculated based on the similarity between the power data files to be stored. , the step of performing a second classification process within each of the third classification sets to form at least one second classification set corresponding to each of the third classification sets may further include the following content:

Perform mean calculation on the keyword feature representation corresponding to each power data file to be stored in the third classification set to output the corresponding mean keyword feature representation;

For each power data file to be stored in the third classification set, calculate the cosine similarity between the keyword feature representation corresponding to the power data file to be stored and the mean keyword feature representation to obtain the power to be stored Cosine similarity corresponding to the data file;

Based on multiple continuous similarity intervals configured by cosine, perform a second classification process on each power data file to be stored in the third classification set to form at least one second classification set corresponding to the third classification set, In the second classification set, the cosine similarity corresponding to each power data file to be stored belongs to the same similarity interval. The similarity interval can be configured in advance according to actual needs, and is not specifically limited here. .

With reference to FIG. 3 , an embodiment of the present invention also provides a power data file processing system, which can be applied to the above-mentioned power data file processing platform. Wherein, the power data file processing system may include the following software function modules:

A data file marking module, configured to mark each of the plurality of power data files to be stored as a power data file to be stored as a power data file to be processed;

A power anomaly analysis module is used to utilize multiple power data analysis networks to analyze the target power anomaly characterization data corresponding to the power data file to be processed, and the target power anomaly characterization data is used to reflect the corresponding power data file to be processed. Abnormal status of the power system;

A first classification processing module, configured to perform first classification processing on the plurality of power data files to be stored based on the corresponding target power abnormality characterization data to form at least one first classification set, each of the first classification sets Includes at least one power data file to be stored;

The second classification processing module is configured to perform second classification processing within each of the first classification sets based on the similarity between the power data files to be stored, and form at least one corresponding to each of the first classification sets. second classification sets, each second classification set including at least one power data file to be stored;

A classification storage module, configured to classify and store each obtained second classification set respectively.

Optionally, in some implementations, the first classification processing module is specifically used to:

Perform consistency or similarity analysis on the target power abnormality characterization data corresponding to each two power data files to be stored in the plurality of power data files to be stored;

The analyzed power data files to be stored whose corresponding target power abnormality characterization data are consistent or whose corresponding target power abnormality characterization data belong to the same parameter interval are assigned to the same first classification set to form At least one first category set.

Optionally, in some implementations, the second classification processing module is specifically used to:

For each of the first classification sets, a quantitative statistical operation is performed on the power data files to be stored included in the first classified set to form a corresponding statistical value of the number of files, and when the statistical value of the number of files is less than or equal to a predetermined In the case of the first reference value, the first classification set is determined as the corresponding second classification set, and in the case where the file quantity statistical value is greater than the first reference value, the first classification set is determined as The corresponding third classification set;

Based on the similarity between the power data files to be stored, a second classification process is performed within each of the third classification sets to form at least one second classification set corresponding to each of the third classification sets.

In summary, the present invention provides a power data file processing system and method. For each power data file to be stored, the power data file to be stored is marked as a power data file to be processed; using multiple The power data analysis network analyzes the target power abnormality characterization data corresponding to the power data file to be processed; based on the corresponding target power abnormality characterization data, performs first classification processing on multiple power data files to be stored to form at least one first classification Sets; based on the similarity between the power data files to be stored, perform second classification processing within each first classification set to form at least one second classification set corresponding to each first classification set; for each second classification set Classified collections are stored separately. Based on the foregoing content, two-level classification processing can be performed through the similarity between data files and the corresponding target power anomaly characterization data, making the classification more accurate and thereby improving the reliability of classification storage.

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims (10)

1. A method for processing power data files, which is characterized by including: For each of the plurality of power data files to be stored, marking the power data file to be stored as a power data file to be processed; Utilize multiple power data analysis networks to analyze the target power abnormality characterization data corresponding to the power data file to be processed, and the target power abnormality characterization data is used to reflect the abnormal state of the power system corresponding to the power data file to be processed; Based on the corresponding target power abnormality characterization data, a first classification process is performed on the plurality of power data files to be stored to form at least one first classification set, each of the first classification sets including at least one power data file to be stored. ; Based on the similarity between the power data files to be stored, a second classification process is performed within each of the first classification sets to form at least one second classification set corresponding to each of the first classification sets. The two-category set includes at least one power data file to be stored; Each obtained second classification set is classified and stored separately. 2. The method of processing power data files according to claim 1, wherein the plurality of power data files to be stored are subjected to a first classification process based on the corresponding target power abnormality characterization data to form at least The steps for a first classification set include: Perform consistency or similarity analysis on the target power abnormality characterization data corresponding to each two power data files to be stored in the plurality of power data files to be stored; The analyzed power data files to be stored whose corresponding target power abnormality characterization data are consistent or whose corresponding target power abnormality characterization data belong to the same parameter interval are assigned to the same first classification set to form At least one first category set. 3. The method for processing power data files according to claim 1, wherein the second classification process is performed within each of the first classification sets based on the similarity between the power data files to be stored. , the step of forming at least one second classification set corresponding to each first classification set includes: For each of the first classification sets, a quantitative statistical operation is performed on the power data files to be stored included in the first classified set to form a corresponding statistical value of the number of files, and when the statistical value of the number of files is less than or equal to a predetermined In the case of the first reference value, the first classification set is determined as the corresponding second classification set, and in the case where the file quantity statistical value is greater than the first reference value, the first classification set is determined as The corresponding third classification set; Based on the similarity between the power data files to be stored, a second classification process is performed within each of the third classification sets to form at least one second classification set corresponding to each of the third classification sets. 4. The method for processing power data files according to claim 3, wherein the second classification process is performed within each of the third classification sets based on the similarity between the power data files to be stored. , the step of forming at least one second classification set corresponding to each of the third classification sets includes: Perform a keyword extraction operation on each power data file to be stored in the third classification set to form a keyword sequence corresponding to each power data file to be stored. Each keyword in the keyword sequence belongs to A collection of reference keywords for power system domain configuration; Perform feature mining processing on the corresponding keyword sequences respectively to form a keyword feature representation corresponding to the power data file to be stored; Based on the keyword feature representation, the similarity between the corresponding power data files to be stored is calculated, and based on the similarity between the power data files to be stored, it is performed within each of the third classification sets. The second classification process forms at least one second classification set corresponding to each third classification set. 5. The method for processing power data files according to claim 4, wherein feature mining is performed on the corresponding keyword sequences to form keyword features corresponding to the power data files to be stored. The steps represented include: For each keyword in the keyword sequence, embedding processing is performed on the keyword to form a word embedding feature representation corresponding to the keyword; and, based on the target power data corpus, determining the keyword in the keyword sequence For each keyword, whether there are related keywords in the keyword sequence, and the co-occurrence probability of the related keywords and the corresponding keywords in the target power data corpus is greater than the preset probability; Each keyword in the keyword sequence that does not have a relevant keyword is marked as a first keyword, and each keyword in the keyword sequence that has a relevant keyword is marked as a second keyword. words, and mark the word embedding feature representation of each first keyword to mark the target word embedding feature representation of the first keyword; For each second keyword, mark the word embedding feature representation corresponding to the related keyword corresponding to the second keyword to mark the related word embedding feature representation corresponding to the second keyword, and, The related word embedding feature representation is subjected to a transposition operation to form a transposed word embedding feature representation corresponding to the second keyword; Calculate the word embedding feature representation corresponding to each of the second keywords, the corresponding transposed word embedding feature representation and the related word embedding feature representation and perform a fusion operation to form a target word embedding corresponding to each of the second keywords. Feature representation; The target word embedding feature representation corresponding to each keyword in the keyword sequence is spliced to form a keyword feature representation corresponding to the power data file to be stored. 6. The method for processing power data files according to claim 4, wherein the similarity between the corresponding power data files to be stored is calculated based on the keyword feature representation, and the similarity between the corresponding power data files to be stored is calculated based on the keyword feature representation. The steps of storing similarities between power data files, performing second classification processing within each of the third classification sets, and forming at least one second classification set corresponding to each of the third classification sets include: Perform mean calculation on the keyword feature representation corresponding to each power data file to be stored in the third classification set to output the corresponding mean keyword feature representation; For each power data file to be stored in the third classification set, calculate the cosine similarity between the keyword feature representation corresponding to the power data file to be stored and the mean keyword feature representation to obtain the power to be stored Cosine similarity corresponding to the data file; Based on multiple continuous similarity intervals configured by cosine, perform a second classification process on each power data file to be stored in the third classification set to form at least one second classification set corresponding to the third classification set, In the second classification set, the cosine similarity corresponding to each included power data file to be stored belongs to the same similarity interval. 7. The method for processing power data files according to any one of claims 1 to 6, characterized in that the target power abnormality representation corresponding to the power data file to be processed is analyzed using multiple power data analysis networks. Data steps include: Utilizing multiple power data analysis networks to perform feature mining operations on the power data files to be processed to output corresponding multiple initial data feature representations, each of the multiple power data analysis networks is used to perform feature mining based on the loaded data, output corresponding power abnormality characterization data, and the power data file to be processed belongs to the operating text data of the power system; Perform a feature representation fusion operation on the plurality of initial data feature representations to form a corresponding aggregated data feature representation; Based on the aggregated data feature representation, the target power abnormality characterization data corresponding to the power data file to be processed is analyzed, and the target power abnormality characterization data is used to reflect the abnormal state of the power system corresponding to the power data file to be processed. 8. A power data file processing system, characterized by including: A data file marking module, configured to mark each of the plurality of power data files to be stored as a power data file to be stored as a power data file to be processed; A power anomaly analysis module is used to utilize multiple power data analysis networks to analyze the target power anomaly characterization data corresponding to the power data file to be processed, and the target power anomaly characterization data is used to reflect the corresponding power data file to be processed. Abnormal status of the power system; A first classification processing module, configured to perform first classification processing on the plurality of power data files to be stored based on the corresponding target power abnormality characterization data to form at least one first classification set, each of the first classification sets Includes at least one power data file to be stored; The second classification processing module is configured to perform second classification processing within each of the first classification sets based on the similarity between the power data files to be stored, and form at least one corresponding to each of the first classification sets. second classification sets, each second classification set including at least one power data file to be stored; A classification storage module, configured to classify and store each obtained second classification set respectively. 9. The power data file processing system according to claim 8, wherein the first classification processing module is specifically used for: Perform consistency or similarity analysis on the target power abnormality characterization data corresponding to each two power data files to be stored in the plurality of power data files to be stored; The analyzed power data files to be stored whose corresponding target power abnormality characterization data are consistent or whose corresponding target power abnormality characterization data belong to the same parameter interval are assigned to the same first classification set to form At least one first category set. 10. The power data file processing system according to claim 8, characterized in that the second classification processing module is specifically used to: For each of the first classification sets, a quantitative statistical operation is performed on the power data files to be stored included in the first classified set to form a corresponding statistical value of the number of files, and when the statistical value of the number of files is less than or equal to a predetermined In the case of the first reference value, the first classification set is determined as the corresponding second classification set, and in the case where the file quantity statistical value is greater than the first reference value, the first classification set is determined as The corresponding third classification set; Based on the similarity between the power data files to be stored, a second classification process is performed within each of the third classification sets to form at least one second classification set corresponding to each of the third classification sets.