Self-learning expert database construction method and system for power grid power-cut plan arrangement

Abstract

The invention discloses a self-learning expert database construction method and system for power grid power-cut plan arrangement, and relates to the technical field of power system power-cut maintenance plans, and the method comprises the steps: constructing an initial knowledge base based on historical power-cut data and maintenance log data; based on the initial knowledge base, self-learning is carried out, and a learning experience base containing simultaneous stop constraint, window period constraint and equipment linkage adjustment experience is constructed; performing data evaluation based on the learning experience library to form a power grid power-cut plan arrangement self-learning expert library; the method improves the accuracy and efficiency of arrangement, reduces the problems caused by human errors or experience deviation, improves the operation reliability and stability of a power grid, reduces the risk and loss of power failure, optimizes the resource configuration, and shortens the power failure time.

Description

Technical Field

[0001] The present invention relates to the technical field of power system outage maintenance planning, and particularly to a method and system for constructing a self-learning expert database for power grid outage plan scheduling. Background Art

[0002] With the continuous expansion of the power grid scale and the complexity of power demand, the scheduling of outage plans has become one of the core links in power grid management. Traditional outage plan scheduling methods mostly rely on manual experience, rule-based models or calculations based on simple algorithms. These methods often cannot meet the requirements of the high efficiency and accuracy of modern power grid operation, directly affecting the power supply reliability and safe and stable operation of the power grid.

[0003] Constructing an outage expert database has important practical significance. By constructing a systematic and structured outage expert database, manual experience can be solidified for centralized management and experience inheritance. This can not only effectively avoid duplicate work and waste of experience, but also improve the emergency response ability and processing efficiency for outage events. Through the mining and analysis of historical outage data, data can be fed back to the business. The expert database can provide data-driven decision support in the preparation of maintenance plans, and can record manual adjustment preferences and optimize algorithm models. This enables the expert database to be continuously improved and optimized, providing more accurate guidance for outage plans.

[0004] The application of outage expert databases in outage plan scheduling has been practiced for a long time. Traditional expert database research mainly focuses on constructing expert systems by abstracting and extracting manual experience, often having deficiencies such as large workload and limited summary scope. Self-learning expert databases can well solve these problems. They have the ability of autonomous learning and gradual optimization. By analyzing historical data and fully learning expert experience to guide outage plan scheduling, they can quickly find the time window with the least risk for outage plan arrangements with similar characteristics, improve the efficiency of plan scheduling, and reduce the operation risk of the power grid. However, the application of existing self-learning methods in power grid outage plan scheduling is still in its initial stage, lacking sufficient self-learning parameter training ability and algorithm adaptability. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the technical problem to be solved by the present invention is: how to construct a self-learning expert database, utilize historical outage data and maintenance logs, automatically learn and optimize outage plan scheduling, and improve the stability of power grid operation and power supply reliability.

[0007] To solve the above technical problem, the present invention provides the following technical solutions:

[0008] In a first aspect, an embodiment of the present invention provides a method for constructing a self-learning expert library for power grid outage plan scheduling, including:

[0009] Construct an initial knowledge base based on historical outage data and maintenance log data;

[0010] Based on the initial knowledge base, perform self-learning to construct a learning experience library including co-outage constraints, window period constraints, and equipment linkage adjustment experience;

[0011] Based on the learning experience library, conduct data evaluation to form a self-learning expert library for power grid outage plan scheduling.

[0012] As a preferred solution of the method for constructing a self-learning expert library for power grid outage plan scheduling, wherein:

[0013] The conducting data evaluation based on the learning experience library to form a self-learning expert library for power grid outage plan scheduling includes:

[0014] Input manual adjustment as a preference feature into the initial knowledge base. After the knowledge base is updated, the algorithm automatically conducts learning to iterate new versions of co-outage constraint experience, window period constraint experience, and equipment linkage adjustment experience.

[0015] As a preferred solution of the method for constructing a self-learning expert library for power grid outage plan scheduling, wherein:

[0016] The performing self-learning based on the initial knowledge base to construct a learning experience library including co-outage constraints, window period constraints, and equipment linkage adjustment experience includes:

[0017] Conduct self-learning of window period constraint experience, including: taking the intervals between the start and end times of the equipment's own power outage maintenance and the surrounding festivals as the mining criteria, discretely quantifying the power outage time periods and the annual time, performing interval weighted calculation, and establishing a power outage holiday discrete interval model.

[0018] As a preferred solution of the method for constructing a self-learning expert library for power grid outage plan scheduling, wherein:

[0019] The conducting self-learning of window period constraint experience further includes:

[0020] For the ambiguity values of the power outage time periods when the power outage maintenance work straddles festivals, perform identification and elimination processing, and process them based on the criterion that the power outage time period of the equipment exceeds the range of its corresponding festival segment to obtain an ambiguity value identification model for power outage holidays.

[0021] As a preferred solution of the method for constructing a self-learning expert library for power grid outage plan scheduling, wherein:

[0022] The performing self-learning based on the initial knowledge base to construct a learning experience library including co-outage constraints, window period constraints, and equipment linkage adjustment experience further includes:

[0023] Perform self-learning of co-stop constraint experience:

[0024] Scan the initial knowledge base data, extract non-repeated items and calculate the support of frequent items;

[0025] Define the minimum support threshold based on the support of frequent items, delete the items with support less than the minimum support, and sort the remaining data in descending order by item set;

[0026] For each frequent item, collect the set of paths ending with this item to form a conditional pattern base;

[0027] Use the conditional pattern base to construct a conditional FP-tree;

[0028] Recursively perform frequent item set mining on each conditional FP-tree until there are no more frequent item sets to mine in the conditional FP-tree;

[0029] Sort out and output all mined frequent item sets.

[0030] As an optimal solution for the construction method of the self-learning expert database for power grid outage plan scheduling, wherein:

[0031] The self-learning based on the initial knowledge base to construct a learning experience database including co-stop constraint, window period constraint and equipment linkage adjustment experience further includes:

[0032] Perform self-learning of equipment linkage adjustment experience:

[0033] Perform sequential pattern mining to extract the equipment linkage adjustment experience from historical maintenance data;

[0034] Scan the data set S in the initial database, obtain the frequent 1-item set and sort it in the order of maintenance time, and delete the infrequent items;

[0035] Count the prefixes of length 1, delete the items corresponding to the prefixes with support lower than the threshold α from the data set S, and at the same time obtain all the frequent 1-item sequences, and let i = 1;

[0036] Recursively mine for each prefix of length i that meets the support requirement:

[0037] Find the projection database corresponding to the prefix. If the projection database is empty, recursively return;

[0038] Count the support counts of each item in the corresponding projection database. If the support counts of all items are lower than the threshold α, recursively return;

[0039] Merge each single item that meets the support count with the current prefix to obtain a new prefix;

[0040] Let \(i = i + 1\). The prefixes are the respective prefixes after merging single items. Recursively execute the merging of each single item that meets the support count with the current prefix to obtain a new prefix.

[0041] Output all frequent sequence sets that meet the support requirements.

[0042] As an optimal solution for the construction method of the self - learning expert library for power grid outage plan scheduling, where:

[0043] Based on the historical outage data and maintenance log data, constructing the initial knowledge base includes:

[0044] After data extraction, predict the missing values, fill in the incomplete data through adjacent non - empty values, and construct a regression prediction model based on the original data.

[0045] Integrate and process the duplicate data, and perform real - time monitoring and identification of the future state model through data monitoring.

[0046] In a second aspect, an embodiment of the present invention provides a self - learning expert library construction system for power grid outage plan scheduling, including:

[0047] A data processing module, used to construct an initial knowledge base based on historical outage data and maintenance log data;

[0048] An intelligent learning module, used to perform self - learning based on the initial knowledge base to construct a learning experience library including co - outage constraints, window period constraints, and equipment linkage adjustment experience;

[0049] A data evaluation module, used to perform data evaluation based on the learning experience library to form a self - learning expert library for power grid outage plan scheduling.

[0050] In a third aspect, an embodiment of the present invention provides a computing device, including:

[0051] A memory and a processor;

[0052] The memory is used to store computer - executable instructions, and the processor is used to execute the computer - executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the method for constructing a self - learning expert library for power grid outage plan scheduling as described in any embodiment of the present invention.

[0053] In a fourth aspect, an embodiment of the present invention provides a computer - readable storage medium, which stores computer - executable instructions, and when the computer - executable instructions are executed by a processor, the method for constructing a self - learning expert library for power grid outage plan scheduling is implemented.

[0054] Advantages of the present invention: By means of algorithms such as FP-Growth, DGWA, and PrefixSpan, the present invention automatically mines key experiences in the outage plan scheduling; introduces a data monitoring and real-time update mechanism to ensure that the expert database has dynamic adaptability to future-state data; combines manual experience and self-learning algorithms to significantly improve the accuracy and practicality of the expert database; can automatically extract effective rules from historical data through autonomous learning, form an autonomous learning expert database after safety check data evaluation and manual adjustment and feedback of preferences, improve the accuracy and efficiency of scheduling, reduce problems caused by human errors or experience deviations, simultaneously enhance the operation reliability and stability of the power grid, reduce the risk and loss of power outages, optimize resource allocation, and shorten the outage time. Description of the Drawings

[0055] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0056] Figure 1 is the overall flowchart of the method for constructing an autonomous learning expert database for power grid outage plan scheduling according to the present invention;

[0057] Figure 2 is a schematic diagram of the outage holiday discrete interval model of the method for constructing an autonomous learning expert database for power grid outage plan scheduling according to the present invention;

[0058] Figure 3 is a schematic diagram of the outage holiday ambiguity value identification model of the method for constructing an autonomous learning expert database for power grid outage plan scheduling according to the present invention;

[0059] Figure 4 is a schematic diagram of the outage holiday weather temperature classification model of the method for constructing an autonomous learning expert database for power grid outage plan scheduling according to the present invention. Detailed Embodiments

[0060] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0061] In the following description, many specific details are set forth to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways than those specifically described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0062] Secondly, as used herein, "one embodiment" or "an embodiment" refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other.

[0063] Embodiment 1

[0064] Referring to Figure 1 , which is the first embodiment of the present invention, this embodiment provides a method for constructing a self-learning expert library for power grid outage plan scheduling, including

[0065] S1: Based on historical outage data and maintenance log data, construct an initial knowledge base;

[0066] S2: Based on the initial knowledge base, perform self-learning to construct a learning experience library including co-outage constraints, window period constraints, and equipment linkage adjustment experience;

[0067] S3: Based on the learning experience library, perform data evaluation to form a self-learning expert library for power grid outage plan scheduling.

[0068] It should be noted that through steps S1 - S3, effective rules can be automatically extracted from historical data through self-learning. After safety verification data evaluation and manual adjustment and feedback of preferences, a self-learning expert library is formed, which improves the accuracy and efficiency of scheduling, reduces problems caused by human errors or experience biases, simultaneously enhances the operation reliability and stability of the power grid, reduces the risk and loss of power outages, and optimizes resource allocation and shortens the power outage time.

[0069] Embodiment 2

[0070] Referring to Figures 1 - 4 , which is an embodiment of the present invention. Based on the previous embodiment, a method for constructing a self-learning expert library for power grid outage plan scheduling is provided, including:

[0071] In the embodiment of the present application, in the above step S1, constructing the initial knowledge base based on historical outage data and maintenance log data includes:

[0072] Select historical outage plan data and maintenance log data, and use the SAtt - SGRU - CNN model to perform data extraction on them.

[0073] It should be noted that when the SAtt-SGRU-CNN model processes historical power outage data and maintenance log data, it can effectively extract multi-level features from the text, improving the accuracy and efficiency of data processing. Through Skip-GRU, the model can capture the global information contained in power outage events and maintenance logs, thereby identifying key factors. The improved self-attention mechanism can reassign weights to the key feature elements related to power outages in the text, ensuring attention to key content. In addition, the optimized multi-channel CNN further enhances the extraction of local features and can identify the detailed information of different types of power outages and maintenance events. The model extracts features such as power outage equipment, power outage time, equipment type, maintenance frequency, triggering section, section control requirements, triggering grid risks, maintenance category, and opinions on mode arrangements, providing strong support for subsequent analysis and decision-making. In summary, through multi-level feature extraction and fusion, the model makes the processing and analysis of historical power outage data and maintenance logs more comprehensive and accurate.

[0074] In another possible implementation, in addition to using the SAtt-SGRU-CNN model for data extraction, the following models or methods can also be considered:

[0075] BERT (Bidirectional Encoder Representations from Transformers):

[0076] BERT is a pre-trained language model based on Transformer that can understand text information through bidirectional context. It can be used to extract semantic features from historical power outage data and maintenance logs, identify key information (such as power outage equipment, time, risks, etc.), and generate high-quality text representations.

[0077] LSTM (Long Short-Term Memory):

[0078] LSTM is a classic variant of the Recurrent Neural Network (RNN) suitable for processing time series data. It can capture the time-dependent relationships in power outage events and maintenance logs and extract features such as equipment status changes and maintenance frequencies.

[0079] Transformer:

[0080] The Transformer model captures the global dependencies in the text through the self-attention mechanism and is suitable for processing long texts and complex contexts. It can be used to extract key features in power outage events, such as section control requirements and grid risks.

[0081] GNN (Graph Neural Network):

[0082] If the power outage data and maintenance logs involve the power grid topology or the association relationships between devices, GNN can be used to model the dependencies between devices and extract features such as the experience of coordinated adjustment of devices.

[0083] After data extraction, data denoising is performed:

[0084] Clean and associate the historical power outage plan data and relevant maintenance log data to optimize the data quality and ensure that the sample data used for experience learning is more accurate and reliable.

[0085] Filling in missing data: Predict the missing values and fill in the incomplete data through adjacent non-empty values. Based on the original data, construct a regression prediction model, and its expression is:

[0086]

[0087] Among them, b0 represents the initial regression coefficient, b ui represents the regression coefficient of the first half of dataset i, b di represents the regression coefficient of the second half of dataset i, c1 and c2 respectively represent the regression coefficients of the first half and the second half of the dataset, H ui represents the average value of the first half of dataset i, H di represents the average value of the second half of dataset i, and α represents the regression timeliness.

[0088] Integrating duplicate data: Considering that the data sources of historical power outage data and maintenance logs are not unique, and there may be the same data in different data sources, it is necessary to perform integration processing on duplicate data. For entity e, its duplicate data integration feature vector F e is:

[0089] F e ={a i , f i , w i , [h i1 , h i2 |i = 1, 2, …, k}

[0090] Perform data monitoring:

[0091] Considering the operation of the future state model, real-time monitoring and identification of the future state model are carried out through data monitoring to ensure that unique tags can be assigned to future data, enabling effective processing and analysis of this data. This function is completed through an agent-based data synchronization model. The model consists of three layers: the application layer, the proxy layer, and the data layer. The local proxy uses a data synchronization listener to monitor data changes in real time, and the global proxy uses a data synchronization detector to periodically detect whether the data is synchronized and generates corresponding synchronization commands for distribution and execution by the global proxy, solving the problem of data monitoring and synchronization.

[0092] The initial knowledge base is constructed from the data after extraction, cleaning, and monitoring.

[0093] In the embodiment of the present application, based on the above-mentioned initial knowledge base in step S2, self-learning is carried out to construct a learning experience base including co-stop constraints, window period constraints, and equipment linkage adjustment experience, which includes:

[0094] Carry out self-learning of co-stop constraint experience:

[0095] It should be noted that the power system is a highly complex and dynamic network, involving multiple links such as power generation, transmission, transformation, and distribution. The interdependent relationship between each link makes it possible for phenomena such as accompanying outages, rolling outages, and hitchhiking to occur when the system experiences power outages. To reduce the number of power outages and avoid repeated power outages, the present invention uses the FP-Growth algorithm to carry out autonomous learning of co-stop constraint experience, and the subsequent results will be applied to the guidance of power outage plan scheduling.

[0096] Specifically, the first step: construct the FP tree

[0097] Scan all the data in the initial knowledge base to obtain the non-repeated items therein, that is, the item set with a frequency of 1. Define the minimum support degree, delete the items with a support degree less than the minimum support degree, and sort the remaining data in descending order according to the item set. Construct the FP tree for the equipment in the sorted order: first construct the root node, and then successively accumulate and count the nodes that appear until all the nodes in the plan are constructed.

[0098] The second step: mine the frequent item sets

[0099] The second step is divided into three sub-steps:

[0100] 1. Construct the conditional pattern base: Define the conditional pattern base as the set of paths ending with the searched element item. Starting from each frequent element item in the head pointer table, obtain the conditional pattern base corresponding to each element item.

[0101] It should be noted that starting from the node of each frequent item in the FP-tree, trace back upward along the parent nodes until the root node, and collect all the paths passed through. These paths are called conditional pattern bases.

[0102] 2. Generate conditional FP-trees: For each frequent item, a conditional FP-tree needs to be created. Use the obtained conditional pattern bases as input data and construct conditional FP-trees through the FP-tree construction method.

[0103] It should be noted that a conditional FP-tree needs to be generated for each frequent item. The process of constructing a conditional FP-tree is similar to that of constructing an FP-tree, but only uses the transaction data in the conditional pattern bases. The specific steps are: (1) Sort all the transactions in the conditional pattern bases according to the frequency of the frequent item; (2) For each transaction, sort by frequency and insert it into the FP-tree.

[0104] 3. Recursive search: Recursively mine each conditional FP-tree to find deeper frequent item sets.

[0105] If the root node of the conditional FP-tree has no children, it means that there are no more frequent item sets to be mined in this conditional FP-tree, and the recursion ends.

[0106] Otherwise, continue to recursively construct conditional FP-trees and mine frequent item sets.

[0107] The third step: Return the frequent item sets, that is, the output result.

[0108] In another possible implementation, the Apriori algorithm can also be used for co-stop constraint empirical self-learning. Apriori is a classic frequent item set mining algorithm that finds frequent item sets through a layer-by-layer search method. Its core idea is to utilize the property that "subsets of frequent item sets are also frequent" to reduce the search space. Although the computational complexity of Apriori is relatively high, it still has application value in certain scenarios.

[0109] In another possible implementation, the Eclat algorithm (Equivalence Class Transformation) can also be used for co-stop constraint empirical self-learning. The Eclat algorithm performs frequent item set mining through equivalence class transformation, adopts a vertical data format (i.e., stores the occurrence of items according to transaction IDs), and performs intersection operations, which is suitable for efficient frequent item set mining on large datasets.

[0110] In another possible implementation, the H-Mine algorithm can also be used for co-stop constraint empirical self-learning. H-Mine is an efficient frequent item set mining algorithm that combines the advantages of horizontal and vertical data formats and can quickly mine frequent item sets on large datasets.

[0111] Perform self-learning of window period constraint experience:

[0112] It should be noted that considering that not all holiday dates are fixed, when mining the power outage window period, the power outage time period cannot be simply used as the standard. The present invention uses the discrete segment gap weighted algorithm DGWA to establish a discrete interval model for power outage holidays, taking the intervals before and after the holidays at the start and end times of the equipment's own power outage maintenance as the mining standard, discretely quantifying the power outage time period and the annual time, and performing interval weighted calculation.

[0113] The discrete interval model for power outage holidays can predict that during the New Year's Day - Spring Festival period of that year for this equipment, the time intervals from the power outage window to New Year's Day and the Spring Festival are:

[0114]

[0115] Such as Figure 2 shown, T1, T2, T3, and T4 represent the power outage durations of the same equipment during the New Year's Day - Spring Festival period in four years, and ts1, ts 2、 ts 3、 ts4 are the intervals between the start of the power outage and the New Year's Day holiday in four years, and te1, te 2、 te 3、 te4 are the intervals between the end of the power outage and the Spring Festival holiday in four years. Here, ε is taken as 3.

[0116] For the ambiguity value of the power outage time period when the power outage maintenance work straddles the holiday, identification and elimination processing are performed. The identification method of the power outage time period ambiguity value will be processed based on the standard that the power outage time period of this equipment exceeds the range of its holiday segment, and an ambiguity value identification model for power outage holidays is obtained;

[0117] Such as Figure 3 shown, the ambiguity value identification model for power outage holidays can predict that during the Dragon Boat Festival - National Day period of that year for this equipment, the time intervals from the power outage window to the Dragon Boat Festival and the National Day are:

[0118]

[0119] Among them, S1 is the power outage time period, S s1 is the interval between the start of the power outage and the Dragon Boat Festival holiday, and S e1 is the interval between the end of the power outage and the National Day holiday; d3 represents the interval between the end time of the power outage and the start time of the National Day holiday, which is used to determine whether this data needs to be eliminated. Set a value. When d3 / S3 is greater than this value, this piece of data is an ambiguity value and needs to be eliminated and not included in the calculation.

[0120] Since the equipment may have maintenance due to weather conditions, the weather temperature factor also needs to be considered during the self-learning of window period constraint experience, and a weather temperature classification model for power outage holidays is established;

[0121] As Figure 4 shown, during the period from New Year's Day to the Spring Festival of that year, the time intervals from the power outage windows under good and poor weather conditions to New Year's Day and the Spring Festival for the device are respectively:

[0122]

[0123] In another possible implementation, a time series analysis algorithm (such as ARIMA) can also be used for self-learning of window period constraint experience:

[0124] ARIMA (AutoRegressive Integrated Moving Average) is a classic time series analysis algorithm that can capture trends and periodicities in time series. It can be used to analyze the relationship between power outage periods and holidays and predict power outage windows.

[0125] In another possible implementation, dynamic time warping (DTW) can also be used for self-learning of window period constraint experience:

[0126] DTW is a time series similarity measurement algorithm that can handle time series of different lengths and is suitable for analyzing the dynamic relationship between power outage periods and holidays. It can be used to identify the interval patterns between power outage periods and festivals.

[0127] In another possible implementation, weighted moving average (WMA) can also be used for self-learning of window period constraint experience:

[0128] WMA is a time series smoothing technique that can perform weighted calculations on power outage periods, combine historical data and current data, and generate more accurate power outage windows.

[0129] Perform self-learning of equipment linkage adjustment experience:

[0130] It should be noted that in the equipment maintenance business, due to specific linkage relationships between some equipment, specific sequences need to be followed during maintenance. To analyze the maintenance linkage relationships and their sequences between these equipment, the present invention uses the PrefixSpan algorithm for sequence pattern mining to extract the linkage adjustment experience between equipment from historical maintenance data.

[0131] Specifically, scan the dataset S in the initial database to obtain frequent 1-itemsets and arrange them in the order of maintenance time, and delete non-frequent items;

[0132] Count the prefixes of length 1, delete the items corresponding to the prefixes with support less than the threshold α from the dataset S, and obtain all frequent 1-item sequences. Let i = 1.

[0133] Recursively mine for each prefix of length i that meets the support requirement:

[0134] Find the projection database corresponding to the prefix. If the projection database is empty, recursively return.

[0135] Count the support counts of each item in the corresponding projection database. If the support counts of all items are less than the threshold α, recursively return.

[0136] Merge each single item that meets the support count with the current prefix to obtain several new prefixes;

[0137] Let i = i + 1, and the prefixes be the prefixes after merging the single items. Recursively execute the merging of each single item that meets the support count with the current prefix respectively to obtain several new prefixes.

[0138] Finally, output all frequent sequence sets that meet the support requirement.

[0139] In another possible implementation, the GSP (Generalized Sequential Pattern) algorithm can also be used for self-learning of device linkage adjustment experience:

[0140] GSP is a classic sequential pattern mining algorithm that mines frequent sequence patterns by gradually expanding sequences. It is suitable for processing experience learning of device linkage adjustment, especially in terms of the equipment maintenance sequence.

[0141] In another possible implementation, the SPADE (Sequential PAttern Discovery using Equivalence classes) algorithm can also be used for self-learning of device linkage adjustment experience:

[0142] The SPADE algorithm mines sequential patterns through equivalence class transformation and can efficiently handle frequent sequence patterns in large-scale datasets.

[0143] In another possible implementation, the CloSpan algorithm can also be used for self-learning of device linkage adjustment experience:

[0144] CloSpan is a closed sequential pattern mining algorithm that can avoid generating redundant frequent sequence patterns and is suitable for processing complex patterns in device linkage adjustment experience.

[0145] It should be noted that by applying the FP-Growth algorithm, DGWA algorithm, and PrefixSpan algorithm for empirical self-learning, maintenance experiences have been mined in three aspects: simultaneous outage, window period, and equipment linkage adjustment, thus completing the construction of the learning experience library. However, these experiences are completely extracted based on historical data and may not match the actual business requirements, and the risk of not conforming to the actual operation scenario cannot be completely avoided. To avoid the above risks, manual adjustment is still required after evaluation to complete the construction of the expert library.

[0146] In the embodiment of the present application, in step S3, based on the learning experience library, data evaluation is performed to form a self-learning expert library for power grid outage plan scheduling, including:

[0147] Since the power system is a highly rigorous system with a limited fault tolerance margin, the learning results obtained by applying data mining and analysis methods must be verified by the operating principles of the power system. The safety check data evaluation part mainly conducts safety inspections and evaluations on the experiences related to simultaneous outage equipment, maintenance window periods, and equipment maintenance linkages obtained from empirical self-learning. Only the learning results that meet the safety specifications can be approved and passed the verification.

[0148] In view of the fact that the self-learning method cannot comprehensively consider all factors and cannot guarantee full compliance with business requirements, therefore, the present invention will perform manual adjustment on the self-learning experiences that pass the evaluation, and incorporate the manual adjustment method as a preference feature into the self-learning parameter training model to increase the degree of fit between the model and the business, and improve the intelligence level of the expert library. In the manual adjustment part, business personnel can publish the results that meet the business requirements, or require the results that do not meet the requirements to be adjusted and re-learned, and output the adjusted results as a self-learning expert library for power grid outage plan scheduling to provide guiding suggestions for outage plan scheduling work.

[0149] It should be noted that the self-learning parameter training model refers to the model that obtains the learning experience library through self-learning from the initial knowledge base. The manual adjustment preference is input into the initial knowledge base as a feature, and then new experiences are obtained through training by the self-learning parameter training model.

[0150] Specifically, the manual adjustment will be recorded by the self-learning parameter training model and input into the initial knowledge base as a preference feature. After the knowledge base is updated, the algorithm automatically conducts learning to iterate new versions of the simultaneous outage constraint experience, window period constraint experience, and equipment linkage adjustment experience.

[0151] Embodiment 3

[0152] The above is a schematic solution of the method for constructing a self-learning expert database for power grid outage plan scheduling. It should be noted that the technical solution of the system for constructing a self-learning expert database for power grid outage plan scheduling belongs to the same concept as the above method for constructing a self-learning expert database for power grid outage plan scheduling. For the details not described in detail in the technical solution of the system for constructing a self-learning expert database for power grid outage plan scheduling in this embodiment, reference can be made to the description of the technical solution of the above method for constructing a self-learning expert database for power grid outage plan scheduling.

[0153] This embodiment also provides a system based on the method for constructing a self-learning expert database for power grid outage plan scheduling, including:

[0154] A data processing module, configured to construct an initial knowledge base based on historical outage data and maintenance log data;

[0155] An intelligent learning module, configured to perform self-learning based on the initial knowledge base to construct a learning experience base including co-outage constraints, window period constraints, and equipment linkage adjustment experience;

[0156] A data evaluation module, configured to perform data evaluation based on the learning experience base to form a self-learning expert database for power grid outage plan scheduling.

[0157] This embodiment also provides a computing device applicable to the situation of the method for constructing a self-learning expert database for power grid outage plan scheduling, including:

[0158] A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the method for constructing a self-learning expert database for power grid outage plan scheduling as proposed in the above embodiment.

[0159] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the method for constructing a self-learning expert database for power grid outage plan scheduling as proposed in the above embodiment.

[0160] The storage medium proposed in this embodiment and the method for constructing a self-learning expert database for power grid outage plan scheduling proposed in the above embodiment belong to the same inventive concept. For the technical details not described in detail in this embodiment, reference can be made to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0161] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for constructing a self-learning expert database for power grid outage plan scheduling, characterized in that, Including: Construct an initial knowledge base based on historical power outage data and maintenance log data; Based on the initial knowledge base, conduct self-learning to construct a learning experience base including co-outage constraints, window period constraints, and equipment linkage adjustment experience; Based on the learning experience base, conduct data evaluation to form a self-learning expert database for power grid outage plan scheduling.

2. The method for constructing a self-learning expert library for power grid outage plan scheduling according to claim 1, characterized in that The conducting data evaluation based on the learning experience base to form a self-learning expert database for power grid outage plan scheduling includes: Input manual adjustment as a preference feature into the initial knowledge base. After the knowledge base is updated, the algorithm automatically conducts learning to iterate new versions of co-outage constraint experience, window period constraint experience, and equipment linkage adjustment experience.

3. The method for constructing a self-learning expert library for power grid outage plan scheduling according to claim 2, characterized in that, The conducting self-learning based on the initial knowledge base to construct a learning experience base including co-outage constraints, window period constraints, and equipment linkage adjustment experience includes: Conduct self-learning of window period constraint experience, including: taking the intervals between the start and end times of the equipment's own power outage maintenance and the surrounding festivals as the mining criteria, discretely quantifying the power outage time periods and the annual time, conducting interval weighted calculation, and establishing a power outage holiday discrete interval model.

4. The method for constructing a self-learning expert database for power grid outage plan scheduling according to claim 3, characterized in that, The conducting self-learning of window period constraint experience further includes: For the ambiguity values of the power outage time periods when the power outage maintenance work straddles the festival, conduct identification and elimination processing, and process them based on the criterion that the equipment's power outage time period exceeds the range of its corresponding festival segment to obtain a power outage holiday ambiguity value identification model.

5. The method for constructing a self-learning expert database for power grid outage plan scheduling according to claim 4, characterized in that, The conducting self-learning based on the initial knowledge base to construct a learning experience base including co-outage constraints, window period constraints, and equipment linkage adjustment experience further includes: Conduct self-learning of co-outage constraint experience: Scan the data in the initial knowledge base, extract non-repeated items and calculate the support degree of frequent items; Define a minimum support degree threshold based on the support degree of frequent items, delete the items with a support degree less than the minimum support degree, and sort the remaining data in descending order according to the item set; For each frequent item, collect the path set ending with this item to form a conditional pattern base; Use the conditional pattern base to construct a conditional FP tree; Recursively conduct frequent item set mining on each conditional FP tree until there are no more frequent item sets to be mined in the conditional FP tree; Sort out and output all mined frequent item sets.

6. The method for constructing a self-learning expert library for power grid outage plan scheduling according to claim 5, wherein The conducting self-learning based on the initial knowledge base to construct a learning experience base including co-outage constraints, window period constraints, and equipment linkage adjustment experience further includes: Conduct self-learning of equipment linkage adjustment experience: Conduct sequential pattern mining to extract the equipment linkage adjustment experience from the historical maintenance data; Scan the data set S in the initial database to obtain frequent 1-item sets and sort them in the order of maintenance time, and delete the infrequent items; Count the prefixes of length 1, delete the items corresponding to the prefixes with a support degree lower than the threshold α from the data set S, and at the same time obtain all frequent 1-item sequences, and let i = 1; Conduct recursive mining for each prefix of length i that meets the support degree requirement: Find the projection database corresponding to the prefix. If the projection database is empty, recursively return; Count the support degree counts of each item in the corresponding projection database. If the support degree counts of all items are lower than the threshold α, recursively return; Merge each single item that meets the support degree count with the current prefix to obtain a new prefix; Let \(i = i + 1\). The prefixes are the respective prefixes after merging single items. Recursively execute the merging of each single item that meets the support count with the current prefix to obtain a new prefix. Output all frequent sequence sets that meet the support requirements.

7. The method for constructing a self-learning expert library for power grid outage plan scheduling according to claim 6, wherein Based on the historical power outage data and maintenance log data, constructing the initial knowledge base includes: After data extraction, predict the missing values, fill in the incomplete data through adjacent non-empty values, and based on the original data, construct a regression prediction model. Integrate the duplicate data and perform real-time monitoring and identification of the future state model through data listening.

8. A system adopting the method for constructing a self-learning expert database for power grid outage plan scheduling as described in any one of claims 1 to 7, characterized in that, Include: A data processing module for constructing an initial knowledge base based on historical power outage data and maintenance log data. An intelligent learning module for self-learning based on the initial knowledge base to construct a learning experience base containing co-outage constraints, window period constraints, and equipment linkage adjustment experience. A data evaluation module for performing data evaluation based on the learning experience base to form a self-learning expert library for power grid outage plan scheduling.

9. A computing device, including: A memory and a processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method for constructing a self-learning expert library for power grid outage plan scheduling according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, and when the computer-executable instructions are executed by a processor, the steps of the method for constructing a self-learning expert library for power grid outage plan scheduling according to any one of claims 1 to 7 are implemented.

Images