Metering asset management system and method for electric power company

Abstract

The invention relates to the technical field of metering asset management, and particularly discloses a metering asset management system and method for an electric power company, and the method comprises the steps: firstly carrying out the statistics of the total inventory and the inventory proportion of all types of equipment of the electric power company; meanwhile, a big data processing technology based on deep learning is adopted to carry out time sequence analysis on monthly average usage data of various devices, periodic change characteristics of monthly average usage amount of the devices are extracted, and the reasonable inventory of the devices is intelligently deduced according to the periodic change characteristics; and determining inventory management opinions according to comparison between the deduced reasonable inventory of the equipment and the actual total inventory of the equipment. Therefore, an electric power company can be helped to control the inventory level more accurately, unnecessary waste is avoided, and the overall level and economic benefits of asset management are effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of metering asset management, and more specifically, to a metering asset management system and method for a power company. Background Art

[0002] With the rapid development of social economy and the continuous improvement of people's living standards, the demand for electricity is growing, and the level of modern management of the power industry is also facing higher requirements. As an important entity providing power services, the effective management and utilization of power companies' assets are directly related to the quality and efficiency of power supply. Among them, metering assets (such as meters, mutual inductors, etc.) are important components of the power system. They are not only huge in number and variety, but also their accuracy and reliability directly affect the fairness of power transactions and the accuracy of user electricity fee calculation.

[0003] Traditionally, power companies have managed metering assets mainly through regular inventory, manual recording, and using experience to estimate inventory demand, which is not only inefficient and slow to respond, but also prone to human errors and low accuracy. In addition, this estimation method that relies on human experience may not be able to effectively predict future equipment demand trends due to the lack of scientific data analysis methods, often leading to excess or shortage of inventory, which not only increases the company's operating costs, but may also affect normal power supply services.

[0004] Therefore, an intelligent metering asset management system and method for power companies is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a metering asset management system and method for an electric power company, which first counts the total inventory and inventory proportion of various types of equipment of the electric power company, and uses big data processing technology based on deep learning to perform time series analysis on the monthly average usage data of various types of equipment, extracts the periodic change characteristics of the monthly average usage of the equipment, and uses this to intelligently infer the reasonable inventory of the equipment, and then determines the inventory management opinions based on the comparison between the inferred reasonable inventory of the equipment and the actual total inventory of the equipment. In this way, it can help electric power companies control inventory levels more accurately and avoid unnecessary waste, thereby effectively improving the overall level of asset management and economic benefits.

[0006] Accordingly, according to one aspect of the present application, a metering asset management system for a power company is provided, comprising:

[0007] The total inventory input module is used to count the total inventory of various types of equipment of the power company, including electric energy meters, communication modules, mutual inductors and acquisition terminals;

[0008] An inventory structure analysis module, used to calculate the inventory proportion of each type of equipment based on the total inventory of each type of equipment;

[0009] A big data analysis module, used to confirm the reasonable inventory of each type of equipment based on the data set of the monthly average usage of each type of equipment;

[0010] The management module is used to determine inventory management opinions based on the comparison between the reasonable inventory of each type of equipment and the total inventory of each type of equipment.

[0011] In the above-mentioned metering asset management system for power companies, the big data analysis module includes: an electric energy meter usage data extraction unit, which is used to extract the data set of the monthly average usage of the electric energy meter from the data set of the monthly average usage of the various types of equipment; a periodic feature extraction unit, which is used to perform periodic feature extraction on the data set of the monthly average usage of the electric energy meter to obtain multiple electric energy meter monthly average usage periodic feature vectors; a feature enhancement unit, which is used to perform feature enhancement based on feature field effect distribution on the multiple electric energy meter monthly average usage periodic feature vectors to obtain multiple enhanced electric energy meter monthly average usage periodic feature vectors; a reasonable inventory estimation unit, which is used to determine the reasonable inventory of the electric energy meter based on the multiple enhanced electric energy meter monthly average usage periodic feature vectors.

[0012] According to another aspect of the present application, a metering asset management method for a power company is provided, comprising:

[0013] Count the total inventory of various types of equipment of the power company, including electric energy meters, communication modules, mutual inductors and data collection terminals;

[0014] Based on the total inventory of each type of equipment, calculate the inventory ratio of each type of equipment;

[0015] Determine the reasonable inventory of each type of equipment based on a data set of average monthly usage of each type of equipment;

[0016] Based on the comparison between the reasonable inventory quantity of each type of equipment and the total inventory quantity of each type of equipment, the inventory management opinion is determined.

[0017] Compared with the prior art, the metering asset management system and method for power companies provided by this application first counts the total inventory and inventory proportion of various types of equipment of the power company, and uses big data processing technology based on deep learning to perform time series analysis on the monthly average usage data of various types of equipment, extracts the periodic change characteristics of the monthly average usage of the equipment, and uses this to intelligently infer the reasonable inventory of the equipment, and then determines the inventory management opinions based on the comparison between the inferred reasonable inventory of the equipment and the actual total inventory of the equipment. In this way, it can help power companies control inventory levels more accurately and avoid unnecessary waste, thereby effectively improving the overall level of asset management and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 A block diagram of a metering asset management system for an electric power company according to an embodiment of the present application.

[0020] Figure 2 It is a block diagram of a big data analysis module in a metering asset management system for an electric power company according to an embodiment of the present application.

[0021] Figure 3 This is a schematic diagram of data flow of a big data analysis module in a metering asset management system for an electric power company according to an embodiment of the present application.

[0022] Figure 4 It is a block diagram of a periodic feature extraction unit in a metering asset management system for an electric power company according to an embodiment of the present application.

[0023] Figure 5 It is a block diagram of a feature enhancement unit in a metering asset management system for an electric power company according to an embodiment of the present application.

[0024] Figure 6 The present invention is a flowchart of a metering asset management method for a power company according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] Below, the embodiments of the present application will be described in more detail in conjunction with the accompanying drawings, and the above and other purposes, features and advantages of the present application will become more apparent. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0026] In response to the technical problems described in the above background technology, the present application proposes a metering asset management system for power companies. Figure 1 FIG. 1 is a block diagram of a metering asset management system for a power company according to an embodiment of the present application. Figure 1 As shown, according to the embodiment of the present application, the metering asset management system 100 for an electric power company includes: a total inventory entry module 110, which is used to count the total inventory of various types of equipment of the electric power company, and the various types of equipment include electric energy meters, communication modules, mutual inductors and acquisition terminals; an inventory structure analysis module 120, which is used to calculate the inventory proportion of each type of equipment based on the total inventory of each type of equipment; a big data analysis module 130, which is used to confirm the reasonable inventory of each type of equipment based on a data set of the monthly average usage of each type of equipment; a management module 140, which is used to determine inventory management opinions based on a comparison between the reasonable inventory of each type of equipment and the total inventory of each type of equipment.

[0027] In the above-mentioned metering asset management system 100 for power companies, the total inventory input module 110 is used to count the total inventory of various types of equipment of the power company, and the various types of equipment include electric energy meters, communication modules, mutual inductors, and acquisition terminals. It should be understood that by statistically analyzing the total inventory of various types of equipment, the current inventory status can be clarified, which is helpful to formulate a more effective procurement plan to avoid unnecessary cost expenditures.

[0028] In the process of entering the total inventory of various types of equipment, it is necessary to ensure the accuracy, completeness and real-time nature of the data, so as to provide reliable data support for subsequent inventory management and decision-making. First of all, data collection is the basic step for entering the total inventory. In actual applications, the equipment of power companies is distributed in different warehouses and storage locations, so it is necessary to obtain equipment inventory records from multiple data sources, such as the power company's equipment management system, warehouse management system, supply chain management system, etc. At the same time, in order to ensure the comprehensiveness and consistency of the data, a complete data collection mechanism needs to be established.

[0029] Specifically, first of all, it is necessary to connect with various data source systems to ensure smooth data transmission. This is usually achieved through an API interface, which can pull data from various systems in real time or at a scheduled time. For example, you can use RESTful API or SOAP protocol to achieve data transmission. In addition, you can also import data files in batches, such as Excel or CSV files, through file transfer protocols such as FTP and SFTP.

[0030] Secondly, a set of standardized data formats needs to be designed to ensure that data obtained from different systems can be processed uniformly. Standardized data formats usually include key fields such as equipment model, quantity, storage date, storage date, storage location, etc. Through standardized processing, processing problems caused by inconsistent data formats can be effectively avoided and the efficiency of data processing can be improved.

[0031] In addition, data cleaning is a key step to ensure data quality. In practical applications, data obtained from different systems may have various problems, such as duplicate records, inconsistent formats, outliers, etc. If not handled in time, it may seriously affect subsequent data analysis and inventory management. Therefore, data cleaning is an indispensable part of the inventory total entry module.

[0032] In the data cleaning process, first of all, deduplication is required. Duplicate records will lead to inaccurate statistical results of total inventory, so it is necessary to identify and delete duplicate records through unique identifiers (such as the serial number or barcode of the device). In specific implementation, this can be achieved by writing SQL queries or using data processing tools (such as Pandas).

[0033] Secondly, format standardization is required. In actual applications, the data formats in different data systems may be inconsistent, so all data fields need to be converted to a unified format. For example, the date field may have multiple formats (such as YYYY-MM-DD, MM / DD / YYYY, etc.), and it needs to be uniformly converted to a standard date format (such as YYYY-MM-DD). In specific implementation, this can also be achieved by writing scripts or using data processing tools.

[0034] In addition, data anomaly detection and correction are also required. It should be understood that outliers may cause deviations in statistical data, so outliers need to be detected and processed. Common outliers include negative inventory, unreasonable dates, abnormally large values, etc. In specific implementations, outliers can be detected by setting reasonable thresholds, for example, inventory quantities cannot be negative, dates cannot exceed the current date, etc. Detected outliers need to be corrected or deleted. For example, SQL queries can be used to delete records with negative inventory quantities, etc.

[0035] Next, data collation is to further process and summarize the data on the basis of data cleaning to generate the required total inventory statistics. Data collation mainly includes two parts: classification and summary and total inventory calculation. First, the equipment needs to be classified by type. The power company has a wide variety of equipment, including energy meters, communication modules, transformers, and acquisition terminals. In order to facilitate statistics and analysis, the equipment needs to be classified by type. In the specific implementation, this can also be achieved by writing SQL queries.

[0036] At the same time, data storage is an important part of the total inventory entry module. In order to ensure that the statistical results can be saved for a long time and queried at any time, data storage usually uses relational databases (such as MySQL, PostgreSQL, etc.) or NoSQL databases (such as MongoDB, Cassandra, etc.).

[0037] In the data storage process, first of all, you need to design a reasonable database table structure. The total inventory data usually includes fields such as equipment type, total inventory, and statistical date. For example, you can design the following table structure:

[0038] CREATE TABLE inventory_summary(

[0039] id INT AUTO_INCREMENT PRIMARY KEY,

[0040] device_type VARCHAR(255)NOT NULL,

[0041] total_quantity INT NOT NULL,

[0042] statistic_date DATE NOT NULL )

[0044] Next, insert the statistical results into the database. To ensure the integrity and consistency of the data, transactions can be used to handle data insertion operations. At the same time, inventory data needs to be backed up regularly to prevent data loss. In the specific implementation, a scheduled task (such as a cron job) can be set to automatically back up the database.

[0045] In addition, data verification is the last line of defense to ensure the accuracy of total inventory data. Through data verification, errors in the data can be discovered and corrected to ensure the accuracy of inventory management. In practical applications, regular physical inventory checks to check whether the actual inventory is consistent with the system records are important means of data verification. Specifically, first, a regular inventory plan needs to be formulated. According to the actual situation of the power company, a physical inventory can be conducted monthly or quarterly. Before the inventory, an inventory form needs to be prepared to list all the equipment that needs to be counted and its related information. During the inventory, the staff needs to check the actual number of equipment one by one and record it in the inventory form. Secondly, the inventory results need to be compared with the system records, a difference report needs to be written, and the differences need to be analyzed to find out the reasons for the differences and make corrections. Common reasons for differences include data entry errors, equipment loss, equipment damage, etc. According to the reasons for the differences, take appropriate measures to make corrections. For example, if it is a data entry error, the system record needs to be updated; if the equipment is lost or damaged, it needs to be replenished or repaired.

[0046] In summary, the total inventory entry module is an important part of the power company's metering asset management, and its main function is to count the total inventory of various types of equipment in the power company. In specific implementation, the total inventory entry module efficiently and accurately counts the total inventory of various types of equipment through data collection, data cleaning, data collation, data storage and data verification, providing reliable data support for subsequent inventory management and decision-making.

[0047] In the above-mentioned metering asset management system 100 for power companies, the inventory structure analysis module 120 is used to calculate the inventory proportion of each type of equipment based on the total inventory of each type of equipment. It should be understood that by calculating the inventory proportion of each type of equipment, the inventory structure, that is, the distribution of each type of equipment in the entire inventory system, can be clearly seen, so as to optimize and adjust the inventory structure. In an embodiment of the present application, the statistical results of the inventory proportion of each type of equipment can be displayed in a graphical manner on the user interface, so that managers can intuitively understand and analyze the inventory proportion.

[0048] In the above-mentioned metering asset management system 100 for power companies, the big data analysis module 130 is used to confirm the reasonable inventory of the various types of equipment based on the data set of the monthly average usage of the various types of equipment. It should be understood that by analyzing the monthly average usage data of various types of equipment, the cyclical patterns and trends of equipment usage can be revealed, thereby providing a scientific basis for determining the reasonable inventory. For example, for equipment with seasonal fluctuations in usage, future market demand can be predicted based on historical data, and then a reasonable inventory level can be calculated.

[0049] Figure 2It is a block diagram of a big data analysis module in a metering asset management system for an electric power company according to an embodiment of the present application. Figure 3 Schematic diagram of data flow in a big data analysis module in a metering asset management system for a power company according to an embodiment of the present application. Figure 2 and Figure 3 As shown, the big data analysis module 130 includes: an electric energy meter usage data extraction unit 131, which is used to extract a data set of monthly average usage of the electric energy meter from the data set of monthly average usage of the various types of equipment; a periodic feature extraction unit 132, which is used to perform periodic feature extraction on the data set of the monthly average usage of the electric energy meter to obtain a plurality of periodic feature vectors of monthly average usage of the electric energy meter; a feature enhancement unit 133, which is used to perform feature enhancement on the plurality of periodic feature vectors of monthly average usage of the electric energy meter based on feature field effect distribution to obtain a plurality of enhanced periodic feature vectors of monthly average usage of the electric energy meter; and a reasonable inventory estimation unit 134, which is used to determine a reasonable inventory of the electric energy meter based on the plurality of enhanced periodic feature vectors of monthly average usage of the electric energy meter.

[0050] Specifically, the electric energy meter usage data extraction unit 131 is used to extract the data set of the monthly average usage of the electric energy meter from the data set of the monthly average usage of the various types of equipment. It should be understood that since different types of equipment have different usage characteristics and demand patterns, in order to more accurately manage inventory, it is necessary to analyze the usage pattern of each device separately. This application takes the electric energy meter as an example, and analyzes and learns the usage pattern of the electric energy meter by extracting the data set of the monthly average usage of the electric energy meter, so as to determine its reasonable inventory according to the usage pattern and demand changes of the electric energy meter.

[0051] Specifically, the data set of the average monthly usage of the electric energy meter is extracted from the data set of the average monthly usage of the various types of equipment, and the core step is to filter out the relevant data of the electric energy meter according to the device type field. In this process, clear screening criteria are required to ensure the accuracy and relevance of the screening results. Specifically, the device type field is the key to filtering the electric energy meter data. Normally, the device type field should contain the name or type code of the device. For example, the device type field of the electric energy meter may contain keywords such as "electricity meter", "electricity meter", and "Meter". Then, the electric energy meter data is filtered by keyword matching. In specific implementations, SQL queries or other data processing tools (such as Pandas) can be used to implement this function. For example, the following is a simple SQL query example:

[0052] SELECT *

[0053] FROM device_usage

[0054] WHERE device_type IN('Energy meter','Electric meter','Meter');

[0055] In addition to the device type field, you can also combine other conditions for filtering. For example, you can exclude data from a certain time period, or select only data from a certain region. This can be achieved by adding more filtering conditions. For example:

[0056] SELECT *

[0057] FROM device_usage

[0058] WHERE device_type IN('Energy meter','Electric meter','Meter')

[0059] AND usage_date BETWEEN'2023-01-01'AND'2023-12-31'

[0060] AND region='East China';

[0061] Through the above screening steps, the monthly average usage data set of the electric energy meter can be effectively extracted from a large amount of equipment usage data.

[0062] Specifically, the periodic feature extraction unit 132 is used to extract periodic features from the data set of the monthly average usage of the electric energy meter to obtain multiple periodic feature vectors of the monthly average usage of the electric energy meter. Figure 4 FIG. 1 is a block diagram of a periodic feature extraction unit in a metering asset management system for a power company according to an embodiment of the present application. Figure 4 As shown, the periodic feature extraction unit 132 includes: a data regularization subunit 1321, which is used to regularize the data set of the monthly average usage of the electric energy meter according to the time dimension to obtain a time queue of the monthly average usage of the electric energy meter; a singular spectrum decomposition subunit 1322, which is used to extract the periodic features in the time queue of the monthly average usage of the electric energy meter using a singular spectrum decomposition method to obtain multiple monthly average usage quantum sequences of the electric energy meter; a time series encoding subunit 1323, which is used to perform time series encoding on each of the monthly average usage quantum sequences of the multiple electric energy meters to obtain the periodic feature vectors of the monthly average usage of the multiple electric energy meters.

[0063] Specifically, the data regularization subunit 1321 is used to regularize the data set of the monthly average usage of the electric energy meter according to the time dimension to obtain the time queue of the monthly average usage of the electric energy meter. It should be understood that since the usage of the electric energy meter usually has obvious periodic characteristics, such as seasonal changes, holiday effects, etc., in order to accurately capture the temporal variation law of the usage of the electric energy meter, the present application further regularizes the data set of the monthly average usage of the electric energy meter, and arranges it according to the time dimension into the time queue of the monthly average usage of the electric energy meter, so as to retain the continuity and order of the data in the time dimension.

[0064] Specifically, the singular spectrum decomposition subunit 1322 is used to extract the periodic features in the time queue of the monthly average usage of the electric energy meter using the singular spectrum decomposition method to obtain multiple monthly average usage quantum sequences of the electric energy meter. In the technical solution of the present application, it is considered that the original time queue of the monthly average usage of the electric energy meter usually has a certain amount of random noise, which interferes with the accurate extraction of the time series change characteristics of the monthly average usage of the electric energy meter. Therefore, in order to improve the effect of feature extraction, the present application uses a singular spectrum decomposition method to process the time queue of the monthly average usage of the electric energy meter to filter out the noise components and extract the periodic signal, and decompose the time queue of the monthly average usage of the electric energy meter into multiple subsequences. Among them, each monthly average usage quantum sequence of the electric energy meter represents a specific periodic pattern, such as seasonal fluctuations or long-term trends, thereby providing higher quality input data for subsequent reasonable inventory inference.

[0065] In a specific example of the present application, first, a preset sliding window is used to perform sliding sampling with a step size of 1 on the time queue of the monthly average usage of the electricity meter to generate a series of overlapping time series samples, and they are arranged into a time series matrix of the monthly average usage of the electricity meter, wherein each column of the time series matrix of the monthly average usage of the electricity meter is a sliding window.

[0066] Next, the singular value decomposition (SVD) of the time series matrix of the monthly average usage of the electric energy meter is performed to obtain a set of singular values ​​and the left singular vectors and right singular vectors corresponding to each singular value, so as to reveal the main structural patterns in the time series of the monthly average usage of the electric energy meter. The singular value decomposition formula is: T = U∑Y T , U is a matrix composed of multiple left singular vectors, ∑ is a diagonal matrix of singular values, and Y is a matrix composed of multiple right singular vectors.

[0067] Then, according to the size of each singular value, the singular values ​​and their corresponding left singular vectors and right singular vectors are grouped, and the singular values ​​and singular vectors of each group are reconstructed to generate a reconstructed time series matrix: T i =U i Σi Y i T , where U i is a matrix composed of the left singular vectors of the same group, Y i is a matrix composed of the right singular vectors of the same group, Σ i is the singular value diagonal matrix composed of the permutations of the singular values ​​of the same group, T i To reconstruct the time series sub-matrix of the monthly average usage of the electricity meter.

[0068] Finally, the column averages of the reconstructed monthly average usage time series submatrices of each group of electric energy meters are calculated to convert each reconstructed monthly average usage time series submatrix back to the form of a time queue, thereby obtaining multiple monthly average usage quantum sequences of electric energy meters. In this way, the noise components in the original data can be effectively filtered out, and the periodicity and trend components can be separated, making the data smoother and more reliable, thereby providing a more accurate prediction basis for subsequent inventory management.

[0069] Specifically, the time series encoding subunit 1323 is used to perform time series encoding on each of the multiple electric energy meter monthly average usage quantum sequences to obtain the multiple electric energy meter monthly average usage period feature vectors. In a specific example of the present application, the time series encoding subunit 1323 is used to: input each of the multiple electric energy meter monthly average usage quantum sequences into a sequence encoder based on the LSTM-RNN hybrid model to obtain the multiple electric energy meter monthly average usage period feature vectors. That is, in order to fully capture the periodic change pattern contained in each electric energy meter monthly average usage quantum sequence, the present application uses the LSTM-RNN hybrid model with outstanding performance in time series processing tasks to perform sequence encoding on each electric energy meter monthly average usage quantum sequence. Those of ordinary skill in the art should know that the RNN (recurrent neural network) model can memorize the previous element information in the sequence and capture the time series association relationship in the sequence by recursively performing in the evolution direction of the time series. The LSTM (Long Short-Term Memory Network) model is an improvement of the RNN model. By introducing a gating mechanism, it solves the gradient vanishing problem encountered by traditional RNN when processing long sequences, so that it can better capture long-term dependencies in time series. The LSTM-RNN hybrid model combines the LSTM's ability to capture long-term dependencies and the RNN's ability to respond quickly to short-term dependencies, so that it can more comprehensively learn and understand the periodic change patterns and trends in the monthly average usage quantum sequences of each electricity meter, and dig out deeper usage rules, thereby generating multiple monthly average usage period feature vectors of electricity meters.

[0070] Specifically, the feature enhancement unit 133 is used to perform feature enhancement on the multiple electric energy meter monthly average usage cycle feature vectors based on the feature field effect distribution to obtain multiple enhanced electric energy meter monthly average usage cycle feature vectors. Here, considering that the multiple electric energy meter monthly average usage cycle feature vectors respectively represent the usage characteristics of the electric energy meter under different periodic modes, there is a certain contextual correlation and information redundancy. Therefore, in order to further improve the expressive power and distinguishability of the features, the present application further performs feature enhancement processing on the multiple electric energy meter monthly average usage cycle feature vectors, and analyzes the feature field effects of each electric energy meter monthly average usage cycle feature vector to identify the key information therein and further enhance its description capability of the electric energy meter usage pattern.

[0071] Figure 5 FIG. 1 is a block diagram of a feature enhancement unit in a metering asset management system for a power company according to an embodiment of the present application. Figure 5 As shown, the feature enhancement unit 133 includes: a feature updating subunit 1331, which is used to update the multiple electric energy meter monthly average usage cycle feature vectors based on the feature field effect distribution of the multiple electric energy meter monthly average usage cycle feature vectors to obtain updated multiple electric energy meter monthly average usage cycle feature vectors; a bidirectional context association reinforcement subunit 1332, which is used to perform bidirectional context association reinforcement on the updated multiple electric energy meter monthly average usage cycle feature vectors to obtain the multiple reinforced electric energy meter monthly average usage cycle feature vectors.

[0072] More specifically, the feature updating subunit includes: a feature reconstruction secondary subunit, which is used to use a feature reconstructor to perform feature reconstruction on each of the plurality of monthly average usage period feature vectors of the electric energy meters to obtain a plurality of reconstructed monthly average usage period feature vectors of the electric energy meters, which is expressed by the formula:

[0073] T={w1,w2,...,w n}

[0074] h i =W1w i W2

[0075] Wherein, T represents the monthly average usage period feature vectors of the multiple electric energy meters, w1, w2, w i and w n represents the first, second, i-th and n-th monthly average usage cycle feature vectors of the multiple monthly average usage cycle feature vectors of the electric energy meters, respectively, n is the number of the monthly average usage cycle feature vectors of the electric energy meters, W1 and W2 represent the first weight parameter matrix and the second weight parameter matrix, respectively, h iis the i-th reconstructed monthly average usage cycle feature vector of the electric energy meter. That is, the monthly average usage cycle feature vector of the electric energy meter is continuously linearly transformed by using the first weight parameter matrix and the second weight parameter matrix to obtain the reconstructed monthly average usage cycle feature vector of the electric energy meter.

[0076] Here, the feature reconstructor is first used to perform domain transformation on the monthly average usage cycle feature vectors of each electricity meter, mapping them to a unified feature space to achieve feature dimension unification, thereby providing an effective data basis for subsequent feature enhancement.

[0077] More specifically, the feature updating subunit also includes: an information depth calculation secondary subunit, which is used to calculate the information depth implicit characterization value of each reconstructed electricity meter monthly average usage cycle feature vector in the multiple reconstructed electricity meter monthly average usage cycle feature vectors to obtain multiple electricity meter monthly average usage information depth implicit characterization values; a reordering secondary subunit, which is used to reorder the multiple electricity meter monthly average usage cycle feature vectors based on the ordering of the multiple electricity meter monthly average usage information depth implicit characterization values ​​to obtain the updated multiple electricity meter monthly average usage cycle feature vectors. In a specific example, the information depth calculation secondary subunit is used to: calculate the square of the Euclidean norm of the reconstructed periodic feature vector of the average monthly usage of the electricity meter to obtain an initial information depth implicit representation value of the average monthly usage of the electricity meter; calculate the ratio of the sum of the initial information depth implicit representation value of the average monthly usage of the electricity meter and the difference between the sum and the initial information depth implicit representation value of the average monthly usage of the electricity meter to obtain an information depth representation modulation coefficient; calculate the logarithmic function value of the information depth representation modulation coefficient with base 2 to obtain the information depth implicit representation value of the average monthly usage of the electricity meter.

[0078] The above process can be expressed as:

[0079]

[0080] T order ={v1,v2,...,v n}

[0081] Among them, ||·|| represents the Euclidean norm of the calculated vector, log(·) represents the logarithmic function with base 2, and e i represents the deep implicit representation value of the monthly average usage information of the i-th electric energy meter, To rder Represents the updated monthly average usage cycle feature vectors of multiple electric energy meters, v1, v2, v i and v nRespectively represent the first, second, i-th and n-th electric energy meter monthly average usage cycle feature vectors among the updated multiple electric energy meter monthly average usage cycle feature vectors.

[0082] That is, by calculating the information depth implicit representation value of each reconstructed electric energy meter monthly average usage cycle feature vector, the information density or complexity of each electric energy meter monthly average usage cycle feature in the overall information of the data set is measured, and its importance and influence in the overall information of the data set are quantified. Then, according to the information depth implicit representation value of each electric energy meter monthly average usage cycle feature, its feature order is further adjusted, and the information structure of the data set is reorganized to ensure that key information is given priority in the subsequent contextual semantic association reinforcement encoding process.

[0083] More specifically, the bidirectional context association reinforcement subunit is used to: input the updated multiple electric energy meter monthly average usage cycle feature vectors into a first semantic association encoder based on a forward LSTM model to obtain multiple forward electric energy meter monthly average usage context semantic coding feature vectors; input the updated multiple electric energy meter monthly average usage cycle feature vectors into a second semantic association encoder based on a backward LSTM model to obtain multiple backward electric energy meter monthly average usage context semantic coding feature vectors; merge the multiple forward electric energy meter monthly average usage context semantic coding feature vectors and the multiple backward electric energy meter monthly average usage context semantic coding feature vectors to obtain the multiple reinforced electric energy meter monthly average usage cycle feature vectors, which can be expressed as follows:

[0084]

[0085] T f =M m v forward +H m v backward +b m

[0086] in, represents the forward LSTM model, represents the backward LSTM model, v forward Represents the context semantic encoding feature vector of the average monthly usage of multiple forward electric energy meters, v backward Represents the context semantic encoding feature vector of the average monthly usage of multiple backward electric energy meters, M m and H m denote the first weighting parameter and the second weighting parameter respectively, b m is the bias term, T f is the monthly average usage cycle feature vector of the multiple enhanced electric energy meters.

[0087] That is, a bidirectional long short-term memory (LSTM) model is used to encode the updated feature vector of the monthly average usage cycle of the electric energy meter from both the forward and backward directions. Among them, the forward LSTM captures the information flow pattern from front to back, while the backward LSTM operates in the opposite direction to identify the dependency from back to front. This dual-perspective encoding mechanism can capture the global dependency of information at different levels, and use it to strengthen the semantic expression of the monthly average usage cycle characteristics of each electric energy meter. Finally, the bidirectional contextual semantic information of the monthly average usage cycle characteristics of each electric energy meter is integrated by fusing the encoding results of the bidirectional LSTM model, forming a richer and more comprehensive number of enhanced monthly average usage cycle feature vectors of the electric energy meter, thereby achieving a deep understanding of the usage pattern of the electric energy meter.

[0088] Specifically, the reasonable inventory estimation unit 134 is used to determine the reasonable inventory of the electric energy meter based on the multiple enhanced monthly average usage cycle feature vectors of the electric energy meter. In a specific example of the present application, the reasonable inventory estimation unit 134 is used to: after cascading the multiple enhanced monthly average usage cycle feature vectors of the electric energy meter into a multi-cycle cascade feature vector of the monthly average usage of the electric energy meter, the multi-cycle cascade feature vector of the monthly average usage of the electric energy meter is input into a decoder-based reasonable inventory inference module to obtain a decoded value of the reasonable inventory of the electric energy meter. That is, by performing a cascade operation on the multiple enhanced monthly average usage cycle feature vectors of the electric energy meter, the periodic characteristics of the monthly average usage of the electric energy meter on different time scales are integrated to form a more comprehensive data view, thereby more comprehensively reflecting the global change law of the usage of the electric energy meter. Then, the decoder model is used to perform feature analysis and inventory inference on the multi-cycle cascade feature vector of the monthly average usage of the electric energy meter. In a specific example of the present application, the decoder adopts a multi-layer perceptron structure, and processes the multi-period cascade feature vector of the monthly average usage of the electricity meter layer by layer through its internal multi-layer neural network structure to learn the deep-level characteristics of the changes in the usage of the electricity meter, and map it to the target numerical space to generate a reasonable inventory of the electricity meter, thereby providing important data reference for inventory management planning.

[0089] The decoder model is a deep learning model that is mainly used to convert the encoded high-level features into target outputs. In the metering asset management system of the power company, the decoder model predicts the reasonable inventory of the electric energy meter based on the multi-period cascade feature vector of the monthly average usage of the electric energy meter. The following is an example of the network structure of a decoder model:

[0090] Input layer: The input layer of the decoder model is used to receive the multi-period cascade feature vector of the monthly average usage of the electric energy meter and pass it to the next layer for processing.

[0091] Hidden layer: The hidden layer is the core part of the decoder model and is usually composed of multiple fully connected layers. Each fully connected layer contains several neurons, and each neuron performs a nonlinear transformation on the input features through an activation function. The main task of the hidden layer is to learn and extract the deep representation of the input features, thereby capturing the complex patterns and dependencies in the data. In the example of this application, the hidden layer includes multiple fully connected layers, activation functions, batch normalization layers, and dropout layers.

[0092] Specifically, the fully connected layer is the most basic layer in the decoder model, and each neuron is connected to all neurons in the previous layer. Through the weight matrix and bias term, the fully connected layer linearly transforms the input features, and the output of each fully connected layer is a linear combination of the input features of the previous layer. The activation function is used to introduce nonlinear factors so that the model can learn more complex feature representations. Common activation functions include ReLU (Rectified Linear Unit), sigmoid, and tanh. Among them, the ReLU function is one of the most commonly used activation functions, which can effectively alleviate the gradient vanishing problem and improve the training efficiency of the model. Batch Normalization is a common technique used to accelerate the training process of the model and improve the generalization ability of the model. Batch Normalization makes the distribution of input features more stable by standardizing each small batch of data, which helps to reduce internal covariate shift and improve the training speed and stability of the model. Dropout is a common regularization technique used to prevent model overfitting. During the training process, Dropout randomly discards the output of a part of the neurons, thereby forcing the model to learn more independent features. During the testing phase, the outputs of all neurons are retained but scaled to keep the desired output unchanged.

[0093] The output layer is the last layer of the decoder model, which is responsible for converting the feature representation of the hidden layer into the target output. It usually contains one or more neurons, and the number depends on the dimension of the prediction target. In the metering asset management system of the power company, the output layer uses a linear activation function to map the feature representation of the hidden layer output to the value of the reasonable inventory amount, so as to provide accurate inventory management decision support for the power company.

[0094] In a preferred example of the present application, the multi-period cascade feature vector of the monthly average usage of the electric energy meter is passed through a reasonable inventory inference module based on a decoder to obtain a decoded value of the reasonable inventory of the electric energy meter, including: first, determining the feature mean and feature variance of the multi-period cascade feature vector of the monthly average usage of the electric energy meter, and dividing the feature mean by the feature variance to obtain a multi-period cascade probability statistic value of the monthly average usage of the electric energy meter:

[0095]

[0096] Wherein, μ and σ represent the characteristic mean and characteristic variance of the multi-period cascade characteristic vector of the monthly average usage of the electric energy meter, respectively, and ρ represents the multi-period cascade probability statistic value of the monthly average usage of the electric energy meter.

[0097] Next, the multi-period cascade feature vector of the monthly average usage of the electric energy meter is multiplied by the inverse of the maximum eigenvalue in the multi-period cascade feature vector of the monthly average usage of the electric energy meter to obtain the multi-period cascade probability constraint vector of the monthly average usage of the electric energy meter:

[0098] V1=V⊙v max -1

[0099] Where V represents the multi-period cascade feature vector of the monthly average usage of the electric energy meter, v max represents the maximum eigenvalue in the multi-period cascade feature vector of the monthly average usage of the electric energy meter, and V1 represents the multi-period cascade probability constraint vector of the monthly average usage of the electric energy meter.

[0100] Then, the multi-period cascade probability constraint vector of the monthly average usage of the electric energy meter is added to the multi-period cascade probability statistical value of the monthly average usage of the electric energy meter, and the logarithmic value with base 2 is calculated to obtain the multi-period cascade information interaction vector of the monthly average usage of the electric energy meter:

[0101]

[0102] in, represents point addition, log(·) represents the logarithmic function with base 2, and V2 represents the multi-period cascade information interaction vector of the monthly average usage of the electric energy meter.

[0103] Furthermore, after performing point subtraction on the multi-period cascade probability constraint vector of the monthly average usage of the electric energy meter with the multi-period cascade probability statistics of the monthly average usage of the electric energy meter, the inverse of each eigenvalue is calculated to obtain the multi-period cascade sequence constraint vector of the monthly average usage of the electric energy meter:

[0104]

[0105] in, represents point reduction, and V3 represents the multi-period cascade sequence constraint vector of the monthly average usage of the electric energy meter.

[0106] Finally, the multi-period cascade information interaction vector of the monthly average usage of the electricity meter and the multi-period cascade sequence constraint vector of the monthly average usage of the electricity meter are interpolated to obtain an optimized multi-period cascade feature vector of the monthly average usage of the electricity meter; the optimized multi-period cascade feature vector of the monthly average usage of the electricity meter is passed through a reasonable inventory inference module based on a decoder to obtain a decoded value of the reasonable inventory of the electricity meter.

[0107] In the technical solution of the present application, the multiple electric energy meter monthly average usage cycle feature vectors respectively represent the time series pattern features within different time period windows of the electric energy meter monthly average usage. After the multiple electric energy meter monthly average usage cycle feature vectors are subjected to feature enhancement based on field effect distribution analysis, the time period significance of the multiple enhanced electric energy meter monthly average usage cycle feature vectors will be enhanced, but the feature expression differences between the enhanced electric energy meter monthly average usage cycle feature vectors will also be amplified, which makes the multi-period cascade feature vector of the electric energy meter monthly average usage obtained by cascading the multiple enhanced electric energy meter monthly average usage cycle feature vectors have a more significant aggregation diversity of feature distributions of different time periods, thereby affecting the accuracy of the decoding result when the multi-period cascade feature vector of the electric energy meter monthly average usage is decoded and regressed by a decoder. That is, considering that the weight matrix of the decoder acts on the multi-period cascade feature vector of the average monthly usage of the electricity meter, the distribution diversity of the multi-period cascade feature vector of the average monthly usage of the electricity meter causes the uncertainty of the weight stimulus parameters of the weight matrix of the decoder, resulting in the lack of a posteriori probability density inference of the multi-period cascade feature vector of the average monthly usage of the electricity meter through the weight matrix, which will affect the accuracy of the decoding result.

[0108] In this case, the probability statistical characteristics of the multi-period cascade feature vector of the monthly average usage of the electricity meter are used to simulate the mesoscale interaction structure under probability constraints between the eigenvalue scale and the eigenvector scale of the multi-period cascade feature vector of the monthly average usage of the electricity meter, so as to construct a bidirectional latent variable motif based on the mesoscale short sequence relative to the probability statistical value to perform mesoscale bidirectional migration, and to perform posterior recovery based on the sequence constraints on the interactive information, thereby improving the convergence effect in the probability density domain, and improving the accuracy of the decoded value of the reasonable inventory of the electricity meter obtained by the multi-period cascade feature vector of the monthly average usage of the electricity meter through the decoder-based reasonable inventory inference module.

[0109] In the above-mentioned metering asset management system 100 for power companies, the management module 140 is used to determine inventory management opinions based on the comparison between the reasonable inventory of each type of equipment and the total inventory of each type of equipment. For example, when the reasonable inventory of the equipment is lower than the total inventory of the equipment, it indicates that there is excess inventory. At this time, you should consider reducing purchases or production, and you may even need to take promotional measures to speed up inventory turnover. On the contrary, if the reasonable inventory of the equipment is higher than the total inventory of the equipment, it means that the inventory is insufficient, and you need to increase the purchase volume or adjust the production plan to replenish the inventory. In this way, it can effectively help decision makers optimize inventory levels, reduce resource waste, and improve operational efficiency.

[0110] In summary, the metering asset management system for power companies according to the embodiment of the present application is explained, which first counts the total inventory and inventory proportion of various types of equipment of the power company, and uses big data processing technology based on deep learning to perform time series analysis on the monthly average usage data of various types of equipment, extracts the periodic change characteristics of the monthly average usage of the equipment, and uses this to intelligently infer the reasonable inventory of the equipment, and then determines the inventory management opinions based on the comparison between the inferred reasonable inventory of the equipment and the actual total inventory of the equipment. In this way, it can help power companies control inventory levels more accurately and avoid unnecessary waste, thereby effectively improving the overall level of asset management and economic benefits.

[0111] Figure 6 FIG. 1 is a flow chart of a metering asset management method for a power company according to an embodiment of the present application. Figure 6 As shown, according to the embodiment of the present application, the metering asset management method for an electric power company includes the following steps: S1, counting the total inventory of various types of equipment of the electric power company, wherein the various types of equipment include electric energy meters, communication modules, mutual inductors, and acquisition terminals; S2, based on the total inventory of the various types of equipment, calculating the inventory ratio of the various types of equipment; S3, confirming the reasonable inventory of the various types of equipment based on a data set of the monthly average usage of the various types of equipment; S4, determining inventory management opinions based on a comparison between the reasonable inventory of the various types of equipment and the total inventory of the various types of equipment.

[0112] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned metering asset management method for a power company have been described in detail above. Figures 1 to 5 The present invention has been described in detail in the description of the metering asset management system for electric power companies, and therefore, its repeated description will be omitted.

[0113] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

Claims

1. A metering asset management system for a power company, characterized in that: include: The total inventory input module is used to count the total inventory of various types of equipment of the power company, including electric energy meters, communication modules, mutual inductors and acquisition terminals; An inventory structure analysis module, used to calculate the inventory proportion of each type of equipment based on the total inventory of each type of equipment; A big data analysis module, used to confirm the reasonable inventory of each type of equipment based on the data set of the monthly average usage of each type of equipment; A management module, configured to determine inventory management suggestions based on a comparison between the reasonable inventory of each type of equipment and the total inventory of each type of equipment; Among them, the big data analysis module includes: an electricity meter usage data extraction unit, which is used to extract the data set of the monthly average usage of the electricity meter from the data set of the monthly average usage of the various types of equipment; a periodic feature extraction unit, which is used to perform periodic feature extraction on the data set of the monthly average usage of the electricity meter to obtain multiple periodic feature vectors of the monthly average usage of the electricity meter; a feature enhancement unit, which is used to perform feature enhancement based on the feature field effect distribution on the multiple periodic feature vectors of the monthly average usage of the electricity meter to obtain multiple enhanced periodic feature vectors of the monthly average usage of the electricity meter; and a reasonable inventory estimation unit, which is used to determine the reasonable inventory of the electricity meter based on the multiple enhanced periodic feature vectors of the monthly average usage of the electricity meter.

2. The metering asset management system for a power company according to claim 1, characterized in that: The periodic feature extraction unit comprises: A data regularization subunit, used for regularizing the data set of the monthly average usage of the electric energy meter according to the time dimension to obtain a time queue of the monthly average usage of the electric energy meter; A singular spectrum decomposition subunit, used for extracting the periodic features in the time queue of the monthly average usage of the electric energy meter by using a singular spectrum decomposition method to obtain a plurality of monthly average usage quantum sequences of the electric energy meter; The time series encoding subunit is used to perform time series encoding on each of the monthly average usage quantum sequences of the multiple electric energy meters to obtain the monthly average usage period feature vectors of the multiple electric energy meters.

3. The metering asset management system for a power company according to claim 2, characterized in that: The timing coding subunit is used for: Each of the monthly average usage quantum sequences of the multiple electric energy meters is input into a sequence encoder based on an LSTM-RNN hybrid model to obtain the monthly average usage period feature vectors of the multiple electric energy meters.

4. The metering asset management system for a power company according to claim 3, characterized in that: The feature enhancement unit comprises: A feature updating subunit, configured to update the plurality of monthly average usage period feature vectors of the plurality of electric energy meters based on the feature field effect distribution of the plurality of monthly average usage period feature vectors of the electric energy meters to obtain updated plurality of monthly average usage period feature vectors of the electric energy meters; The bidirectional context association strengthening subunit is used to perform bidirectional context association strengthening on the updated multiple electric energy meter monthly average usage cycle feature vectors to obtain the multiple strengthened electric energy meter monthly average usage cycle feature vectors.

5. The metering asset management system for a power company according to claim 4, characterized in that: The feature updating subunit comprises: A feature reconstruction secondary subunit is used to use a feature reconstructor to perform feature reconstruction on each of the plurality of monthly average usage period feature vectors of the electric energy meters to obtain a plurality of reconstructed monthly average usage period feature vectors of the electric energy meters; The information depth calculation secondary subunit is used to calculate the information depth implicit representation value of each reconstructed electric energy meter monthly average usage cycle feature vector in the multiple reconstructed electric energy meter monthly average usage cycle feature vectors to obtain multiple electric energy meter monthly average usage information depth implicit representation values; The reordering secondary subunit is used to reorder the multiple electric energy meter monthly average usage period feature vectors based on the order of the deep implicit characterization values ​​of the multiple electric energy meter monthly average usage information to obtain the updated multiple electric energy meter monthly average usage period feature vectors.

6. The metering asset management system for a power company according to claim 5, characterized in that: The information depth calculation secondary subunit is used to: Calculating the square of the Euclidean norm of the reconstructed monthly average usage period feature vector of the electric energy meter to obtain a deep implicit representation value of the initial information of the monthly average usage of the electric energy meter; Calculating a ratio of a sum of an initial information depth implicit characterization value of the monthly average usage of the electric energy meter and a difference between an initial information depth implicit characterization value of the monthly average usage of the electric energy meter to obtain an information depth characterization modulation coefficient; The logarithmic function value with base 2 of the information depth representation modulation coefficient is calculated to obtain the implicit representation value of the information depth of the monthly average usage of the electric energy meter.

7. The metering asset management system for a power company according to claim 6, characterized in that: The bidirectional contextual association reinforcement subunit is used to: Inputting the updated multiple electric energy meter monthly average usage cycle feature vectors into a first semantic association encoder based on a forward LSTM model to obtain multiple forward electric energy meter monthly average usage context semantic encoding feature vectors; Inputting the updated multiple electric energy meter monthly average usage cycle feature vectors into a second semantic association encoder based on a backward LSTM model to obtain multiple backward electric energy meter monthly average usage context semantic encoding feature vectors; The plurality of forward electric energy meter monthly average usage context semantic coding feature vectors and the plurality of backward electric energy meter monthly average usage context semantic coding feature vectors are fused to obtain the plurality of enhanced electric energy meter monthly average usage period feature vectors.

8. The metering asset management system for a power company according to claim 7, characterized in that: The reasonable inventory estimation unit is used to: After the multiple enhanced monthly average usage period feature vectors of the electric energy meter are cascaded into a multi-period cascade feature vector of the monthly average usage of the electric energy meter, the multi-period cascade feature vector of the monthly average usage of the electric energy meter is input into a reasonable inventory inference module based on a decoder to obtain a decoded value of the reasonable inventory of the electric energy meter.

9. A metering asset management method for an electric power company, characterized in that: include: Count the total inventory of various types of equipment of the power company, including electric energy meters, communication modules, mutual inductors and data collection terminals; Based on the total inventory of each type of equipment, calculate the inventory ratio of each type of equipment; Determine the reasonable inventory of each type of equipment based on the data set of average monthly usage of each type of equipment; Based on the comparison between the reasonable inventory quantity of each type of equipment and the total inventory quantity of each type of equipment, the inventory management opinion is determined.

Images