Power grid planning, construction, operation and maintenance integrated modeling method and system based on semantic system

Abstract

The invention provides a power grid planning, construction, operation and maintenance integrated modeling method and system based on a semantic system, and relates to the technical field of power grid planning, and the method comprises the steps: carrying out the semantic annotation of power grid planning historical data, constructing a semantic vector space, generating a multi-dimensional semantic feature library, and carrying out the feature extraction, fusion and decision output through a multilayer neural network. According to the method, real-time power grid planning demand data is received, an initial scheme including a load distribution prediction result, a grid structure configuration scheme and an investment scale suggestion is output, iterative optimization is performed by adopting a reinforcement learning algorithm, and finally an optimal power grid planning scheme is output. According to the invention, multi-source heterogeneous power grid data can be effectively fused, the accuracy and efficiency of power grid planning are improved, the investment cost is reduced, and the reliability, economy and environmental protection of a power grid are improved.

Description

Technical Field

[0001] The present invention relates to power grid planning technology, and in particular to a semantic system-based power grid planning, construction, operation and maintenance integrated modeling method and system. Background Art

[0002] Traditional grid planning, construction, and operation and maintenance methods mainly rely on expert experience and manual analysis, and adopt static planning methods, which are difficult to adapt to the challenges brought by the continuous expansion of grid scale, rapid growth of load, and large-scale access to new energy. The three stages of planning, construction, and operation and maintenance are relatively independent, lacking effective information sharing and coordination mechanisms, resulting in a disconnect between planning schemes and actual construction and operation and maintenance, making it difficult to maximize benefits throughout the life cycle. In addition, traditional planning methods do not make sufficient use of the mining of massive historical data, making it difficult to fully realize the value of data, resulting in a lack of scientificity and foresight in the formulation of planning schemes.

[0003] Traditional power grid planning methods are unable to effectively handle the complex and ever-changing power grid operating environment, and do not adequately consider uncertain factors such as load forecasting and fluctuations in new energy output, resulting in poor robustness and adaptability of planning schemes. Existing planning methods lack overall consideration of the entire process of planning, construction, and operation and maintenance. There is a serious phenomenon of information islands between various stages, making it difficult to achieve optimal resource allocation and coordinated operation, affecting the overall operating efficiency of the power grid. Traditional planning methods do not make sufficient use of the mining of massive data, making it difficult to effectively extract the valuable information contained in the data, resulting in a lack of data support for the formulation of planning schemes, making it difficult to achieve accurate planning and scientific decision-making. Summary of the invention

[0004] The embodiments of the present invention provide a semantic system-based integrated modeling method and system for power grid planning, construction, operation and maintenance, which can solve the problems in the prior art.

[0005] According to a first aspect of the embodiments of the present invention, Provides an integrated modeling method for power grid planning, construction, operation and maintenance based on a semantic system, including: Acquire historical data of power grid planning, the historical data of power grid planning including load forecast data, power grid topology data and equipment parameter data; semantically annotate the historical data of power grid planning and establish a vocabulary of power grid planning domain; construct a semantic vector space based on the vocabulary of power grid planning domain, map the historical data of power grid planning to the semantic vector space, and generate basic semantic features of power grid planning; perform multi-dimensional decomposition of the basic semantic features of power grid planning according to power grid planning business rules to form a multi-dimensional semantic feature library including time dimension features, space dimension features and business dimension features; Constructing a multi-layer neural network structure, the multi-layer neural network structure includes a feature extraction layer, a feature fusion layer and a decision output layer; inputting feature data in the multi-dimensional semantic feature library into the feature extraction layer, extracting local features through convolution operation; using an attention mechanism in the feature fusion layer to perform adaptive weight allocation and feature fusion on features of different dimensions; setting a multi-task learning module in the decision output layer to perform model training for power grid load prediction, grid structure optimization and investment benefit evaluation respectively; Receive real-time grid planning demand data, and input the real-time grid planning demand data into a trained deep learning model; output an initial grid planning plan based on the deep learning model, wherein the initial grid planning plan includes load distribution prediction results, grid structure configuration plans, and investment scale recommendations; use a reinforcement learning algorithm to iteratively optimize the initial grid planning plan, construct an optimization objective function based on grid reliability indicators, economic indicators, and environmental protection indicators, and output an optimal grid planning plan.

[0006] Acquire historical data of power grid planning, the historical data of power grid planning includes load forecast data, power grid topology data and equipment parameter data; semantically annotate the historical data of power grid planning, and establish a vocabulary of power grid planning domain including: Acquire historical data of power grid planning through a data acquisition interface, wherein the historical data of power grid planning includes load forecast data, power grid topology data, and equipment parameter data; establish a data quality scoring index system for the historical data of power grid planning, wherein the data quality scoring index system includes a data integrity index, a data accuracy index, and a data timeliness index; perform a quality assessment on the historical data of power grid planning based on the data quality scoring index system, and generate a quality assessment report; According to the evaluation results in the quality evaluation report, the historical data of the power grid planning is graded and divided into valid data, data to be repaired and invalid data; missing values ​​in the data to be repaired are repaired by using an interpolation method based on time series correlation, and abnormal values ​​in the data to be repaired are identified and corrected by using statistical features and expert rules to generate repaired data; the valid data and the repaired data are merged to form preprocessed data; Design a unified data model, which includes a data identification field, an attribute description field, and an association relationship field; map the preprocessed data to the unified data model through ETL transformation rules to form standardized data; establish a version identification for the standardized data, record data change information, and generate a data traceability chain; Extract basic terms for power grid planning from normative documents and use the basic terms as initial entries; use a word frequency-inverse document frequency algorithm to analyze the standardized data and extract characteristic words; submit the characteristic words to domain experts for review, and merge the reviewed characteristic words with the initial entries to form an initial vocabulary; A multi-level semantic annotation system is constructed, which includes a part-of-speech tagging layer, a concept category tagging layer and a relationship tagging layer; the initial vocabulary is tagged with parts of speech using a sequence tagging model based on conditional random fields to generate part-of-speech tagging results; the part-of-speech tagging results are tagged with concept categories based on domain ontology to generate concept tagging results; the concept tagging results are tagged with relationships through dependency syntactic analysis to extract semantic relationships between words and generate a power grid planning domain vocabulary.

[0007] A semantic vector space is constructed based on the grid planning domain vocabulary, and the grid planning historical data is mapped into the semantic vector space to generate basic semantic features of grid planning; according to grid planning business rules, the basic semantic features of grid planning are decomposed into multiple dimensions to form a multi-dimensional semantic feature library including time dimension features, space dimension features and business dimension features, including: Establish a semantic vector model in the field of power grid planning, wherein the semantic vector model in the field of power grid planning is constructed using the word2vec algorithm, and the training corpus of the word2vec algorithm includes entries in the vocabulary of the field of power grid planning and their context information; obtain a word vector mapping matrix based on the word2vec algorithm training, wherein the word vector mapping matrix records the high-dimensional vector representation corresponding to each entry; and construct a semantic vector space using the word vector mapping matrix; Performing word segmentation processing on the power grid planning historical data, identifying professional terms based on the power grid planning field vocabulary, and generating a word segmentation sequence; mapping the terms in the word segmentation sequence to the semantic vector space through the word vector mapping matrix to obtain the term semantic vector; performing weighted combination on the term semantic vector to generate basic semantic features of power grid planning; Acquire a power grid planning business rule base, the power grid planning business rule base includes a time dimension rule set, a space dimension rule set and a business dimension rule set; extract a time series correlation feature from the power grid planning basic semantic feature based on the time dimension rule set, the time series correlation feature describes the evolution law of power grid planning elements over time; Extracting spatial distribution features from the basic semantic features of power grid planning based on the spatial dimension rule set, wherein the spatial distribution features describe the geographical location relationship and topological connection relationship of power grid planning elements; extracting business logic features from the basic semantic features of power grid planning based on the business dimension rule set, wherein the business logic features describe the constraints and decision-making basis in the power grid planning process; The temporal correlation feature, the spatial distribution feature and the business logic feature are feature-fused to establish a multi-dimensional semantic feature index structure; and a multi-dimensional semantic feature library is constructed based on the multi-dimensional semantic feature index structure.

[0008] In the feature fusion layer, an attention mechanism is used to perform adaptive weight allocation and feature fusion on features of different dimensions; a multi-task learning module is set in the decision output layer to perform model training for power grid load prediction, grid structure optimization and investment benefit evaluation, including: In the feature fusion layer, adaptive weight allocation is performed on the multi-dimensional features of the power grid planning, wherein the multi-dimensional features include power grid load features, grid structure features, and investment benefit features; an attention submodule is constructed for each dimensional feature, and each of the attention submodules includes a query vector generation unit, a key vector generation unit, and a value vector generation unit; The multi-dimensional features are respectively input into the corresponding attention submodules, and the query vector, key vector and value vector of the multi-dimensional features are generated by the query vector generation unit, the key vector generation unit and the value vector generation unit; the query vector and the key vector are subjected to dot product operation and scaled to obtain an attention score; the attention score is subjected to Softmax normalization to obtain the attention weight of the dimensional feature; The attention weight of the dimensional feature and the value vector are weightedly summed to obtain a weighted representation of the dimensional feature; the weighted representations of different dimensional features are fused, and the correlation strength between dimensions is calculated using a cross-dimensional attention mechanism; the weights of different dimensional features are adaptively assigned according to the correlation strength between dimensions to generate a fused feature vector; A multi-task learning module is set at the decision output layer, wherein the multi-task learning module includes a shared feature extraction layer and a task-specific layer; the fused feature vector is input into the shared feature extraction layer to extract high-level semantic features shared between tasks; the high-level semantic features are respectively input into three parallel task-specific layers for model training; The three parallel task-specific layers include a load forecasting training layer, a grid structure optimization training layer and an investment benefit evaluation training layer; the load forecasting training layer uses a time series deep learning model to model the grid load characteristics; the grid structure optimization training layer uses a graph neural network to analyze the grid topology characteristics; the investment benefit evaluation training layer uses a deep reinforcement learning model to train the investment decision-making process.

[0009] Receive real-time grid planning demand data, input the real-time grid planning demand data into a trained deep learning model; output an initial grid planning plan based on the deep learning model, the initial grid planning plan including load distribution prediction results, grid structure configuration plan and investment scale recommendations including: Receive real-time power grid planning demand data through a data receiving module, wherein the real-time power grid planning demand data includes power grid operation status data, historical load distribution data, current grid topology data and investment constraint condition data; Preprocessing the real-time power grid planning demand data, cleaning the real-time power grid planning demand data, and removing abnormal values; performing time series alignment on the real-time power grid planning demand data; normalizing the real-time power grid planning demand data to obtain preprocessed power grid planning demand data; Input the preprocessed power grid planning demand data into a trained deep learning model, wherein the deep learning model includes a data encoding unit and a solution generation unit; the data encoding unit uses a long short-term memory network to process time series data features, uses a graph convolutional network to process network topology features, and uses a multi-layer perceptron to process investment constraint features; the solution generation unit generates an initial power grid planning solution based on the output features of the data encoding unit; The deep learning model generates the initial grid planning plan through a parallel processing mechanism, and the initial grid planning plan includes: using a probabilistic prediction method to generate load distribution prediction results, and the load distribution prediction results include a mean prediction and a confidence interval of future loads; using a graph optimization algorithm to generate a grid structure configuration plan, and the grid structure configuration plan includes a substation site selection plan and a line corridor planning plan; based on cost-benefit analysis, an investment scale proposal is generated, and the investment scale proposal includes a construction cost budget and a phased investment proposal.

[0010] The reinforcement learning algorithm is used to iteratively optimize the initial power grid planning scheme, and the optimization objective function is constructed according to the power grid reliability index, economic index and environmental protection index. The output of the optimal power grid planning scheme includes: The initial plan of the power grid planning is subjected to data standardization processing to obtain standardized planning data including substation layout data, line corridor data and investment configuration data; a reinforcement learning optimization model is established based on the standardized planning data to construct a state space, wherein the state space records the current substation operation status, line load status and investment execution status; an action space is constructed, wherein the action space includes a substation capacity adjustment amount, a line corridor adjustment amount and an investment scale adjustment amount; Based on the operation data in the state space, a power grid reliability index is constructed, and the power supply reliability rate, voltage qualification rate and network loss rate are obtained through power flow calculation and reliability analysis; based on the adjustment data in the action space, an economic index is constructed, and the project construction cost, operation and maintenance cost and benefit payback period are obtained through investment benefit analysis; based on the combined data of the state space and the action space, an environmental protection index is constructed, and the land occupation area, electromagnetic radiation intensity and carbon emissions are obtained through environmental impact assessment; the power grid reliability index, economic index and environmental protection index are combined into an optimization objective function, and the objective function value is calculated by weighted summation, and the objective function value is used as a reward signal for reinforcement learning; A deep Q learning network is constructed based on the state space, action space and optimization objective function, wherein the deep Q learning network maps the current state to the optimal action selection; the state transition sequence after the action is executed and the corresponding reward signal are stored in an experience replay pool; training samples are selected from the experience replay pool, and the time difference error of each sample is calculated; the samples are sorted according to the time difference error, and samples with larger errors are preferentially selected for network training; The deep Q learning network is updated based on the selected training samples, and the loss between the network prediction value and the target value is calculated using the optimization objective function; the network parameters are optimized by the back propagation algorithm so that the action selection output by the network is closer to the optimal strategy; the action selection, experience replay and network update processes are repeated, and when the change in the optimization objective function value of consecutive preset rounds is less than the convergence threshold, the optimal power grid planning scheme is output.

[0011] According to a second aspect of the embodiments of the present invention, Provides an integrated modeling system for power grid planning, construction, operation and maintenance based on a semantic system, including: The first unit is used to obtain historical data of power grid planning, wherein the historical data of power grid planning includes load forecast data, power grid topology data and equipment parameter data; semantically annotate the historical data of power grid planning and establish a vocabulary of power grid planning domain; construct a semantic vector space based on the vocabulary of power grid planning domain, map the historical data of power grid planning to the semantic vector space, and generate basic semantic features of power grid planning; perform multi-dimensional decomposition of the basic semantic features of power grid planning according to power grid planning business rules to form a multi-dimensional semantic feature library including time dimension features, space dimension features and business dimension features; The second unit is used to construct a multi-layer neural network structure, which includes a feature extraction layer, a feature fusion layer and a decision output layer; the feature data in the multi-dimensional semantic feature library is input into the feature extraction layer, and local features are extracted by convolution operation; the attention mechanism is used in the feature fusion layer to perform adaptive weight allocation and feature fusion on features of different dimensions; a multi-task learning module is set in the decision output layer to perform model training for power grid load prediction, grid structure optimization and investment benefit evaluation respectively; The third unit is used to receive real-time grid planning demand data, input the real-time grid planning demand data into a trained deep learning model; output an initial grid planning plan based on the deep learning model, the initial grid planning plan includes load distribution prediction results, grid structure configuration plan and investment scale recommendations; use a reinforcement learning algorithm to iteratively optimize the initial grid planning plan, construct an optimization objective function based on grid reliability indicators, economic indicators and environmental protection indicators, and output an optimal grid planning plan.

[0012] According to a third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0013] A fourth aspect of the embodiments of the present invention is: A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0014] The beneficial effects of this application are as follows: 1. Improve the efficiency of power grid planning: By building a semantic system and a multi-dimensional semantic feature library, the historical data of power grid planning is converted into machine-understandable semantic information, and combined with a deep learning model for rapid analysis and decision-making, thereby shortening the power grid planning cycle and improving planning efficiency.

[0015] 2. Improve the scientificity and rationality of power grid planning schemes: Utilize multi-layer neural networks and attention mechanisms, integrate multi-dimensional information such as time, space, and business, and consider multiple indicators such as power grid reliability, economy, and environmental protection, so as to generate more scientific and reasonable power grid planning schemes.

[0016] 3. Realize the integration of grid planning, construction, and operation and maintenance: Iterate and optimize the initial plan through reinforcement learning algorithm, and combine it with real-time grid planning demand data to make the generated planning plan closer to the actual operation situation, promote the coordination of planning, construction, and operation and maintenance, and realize integrated management. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart of a semantic system-based integrated modeling method for power grid planning, construction, operation and maintenance according to an embodiment of the present invention; Figure 2 It is a structural schematic diagram of a semantic system-based integrated modeling system for power grid planning, construction, operation and maintenance according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0020] Figure 1 FIG. 1 is a flow chart of a method for integrating power grid planning, construction, operation and maintenance based on a semantic system according to an embodiment of the present invention. Figure 1 As shown, the method includes: S11. Acquire historical data of power grid planning, the historical data of power grid planning including load forecast data, power grid topology data and equipment parameter data; semantically annotate the historical data of power grid planning and establish a vocabulary of power grid planning domain; construct a semantic vector space based on the vocabulary of power grid planning domain, map the historical data of power grid planning to the semantic vector space, and generate basic semantic features of power grid planning; perform multi-dimensional decomposition of the basic semantic features of power grid planning according to power grid planning business rules to form a multi-dimensional semantic feature library including time dimension features, space dimension features and business dimension features; S12. Construct a multi-layer neural network structure, which includes a feature extraction layer, a feature fusion layer and a decision output layer; input the feature data in the multi-dimensional semantic feature library into the feature extraction layer, and extract local features through convolution operation; use the attention mechanism in the feature fusion layer to perform adaptive weight allocation and feature fusion on features of different dimensions; set a multi-task learning module in the decision output layer to perform model training for power grid load prediction, grid structure optimization and investment benefit evaluation respectively; S13. Receive real-time grid planning demand data, and input the real-time grid planning demand data into a trained deep learning model; output an initial grid planning plan based on the deep learning model, wherein the initial grid planning plan includes load distribution prediction results, grid structure configuration plan and investment scale recommendations; use a reinforcement learning algorithm to iteratively optimize the initial grid planning plan, construct an optimization objective function based on grid reliability indicators, economic indicators and environmental protection indicators, and output an optimal grid planning plan.

[0021] In an optional implementation, obtaining historical data of power grid planning, the historical data of power grid planning including load forecast data, power grid topology data and equipment parameter data; semantically annotating the historical data of power grid planning, and establishing a vocabulary of power grid planning domain includes: Acquire historical data of power grid planning through a data acquisition interface, wherein the historical data of power grid planning includes load forecast data, power grid topology data, and equipment parameter data; establish a data quality scoring index system for the historical data of power grid planning, wherein the data quality scoring index system includes a data integrity index, a data accuracy index, and a data timeliness index; perform a quality assessment on the historical data of power grid planning based on the data quality scoring index system, and generate a quality assessment report; According to the evaluation results in the quality evaluation report, the historical data of the power grid planning is graded and divided into valid data, data to be repaired and invalid data; missing values ​​in the data to be repaired are repaired by using an interpolation method based on time series correlation, and abnormal values ​​in the data to be repaired are identified and corrected by using statistical features and expert rules to generate repaired data; the valid data and the repaired data are merged to form preprocessed data; Design a unified data model, which includes a data identification field, an attribute description field, and an association relationship field; map the preprocessed data to the unified data model through ETL transformation rules to form standardized data; establish a version identification for the standardized data, record data change information, and generate a data traceability chain; Extract basic terms for power grid planning from normative documents and use the basic terms as initial entries; use a word frequency-inverse document frequency algorithm to analyze the standardized data and extract characteristic words; submit the characteristic words to domain experts for review, and merge the reviewed characteristic words with the initial entries to form an initial vocabulary; A multi-level semantic annotation system is constructed, which includes a part-of-speech tagging layer, a concept category tagging layer and a relationship tagging layer; the initial vocabulary is tagged with parts of speech using a sequence tagging model based on conditional random fields to generate part-of-speech tagging results; the part-of-speech tagging results are tagged with concept categories based on domain ontology to generate concept tagging results; the concept tagging results are tagged with relationships through dependency syntactic analysis to extract semantic relationships between words and generate a power grid planning domain vocabulary.

[0022] Obtain historical data on power grid planning. By establishing a data collection interface with databases such as the power grid dispatching system and planning management system, historical load forecast data, power grid topology data, and equipment parameter data can be obtained. Load forecast data includes load forecast values ​​at different time granularities, such as hourly, daily, or monthly forecast data. Power grid topology data includes the connection relationship and geographic location information of power equipment such as transmission lines, substations, and generators. Equipment parameter data includes technical parameters such as rated voltage, rated capacity, and impedance of power equipment. For example, historical data from 2020 to 2022 can be exported from the database, including hourly load forecast values, connection information of each substation, and parameter data of all transformers.

[0023] Conduct a quality assessment on the historical data of power grid planning. Establish a data quality scoring index system, including three dimensions: data integrity, data accuracy, and data timeliness. The data integrity index is used to assess whether there are missing values ​​in the data, such as counting the number and proportion of missing values. The data accuracy index is used to assess the degree of deviation of the data, such as the error between the calculated data and the actual value. The data timeliness index is used to assess the freshness of the data, such as the time delay of calculating the data. Conduct a data quality assessment on the historical data collected from 2020 to 2022. For example, it was found that there were 5% missing values ​​in the load forecast data for July 2021, abnormal values ​​in the parameter data of some transformers, and the timeliness of the data in 2020 was poor. Generate a quality assessment report and record the assessment results.

[0024] Perform hierarchical processing on historical data of power grid planning. According to the results of the quality assessment report, historical data are divided into valid data, data to be repaired, and invalid data. For example, data with high integrity, high accuracy, and high timeliness are marked as valid data; data with a small number of missing values ​​or outliers are marked as data to be repaired; data with too many missing values ​​or obvious errors are marked as invalid data. Based on the previous assessment results, the load forecast data with a small number of missing values ​​in July 2021 are marked as data to be repaired, the transformer data with abnormal parameters are also marked as data to be repaired, the data in 2020 are marked as invalid data, and the rest of the data are marked as valid data.

[0025] Repair the data to be repaired. For missing values ​​in the data to be repaired, use an interpolation method based on time series correlation to repair them. For example, use the previous and next data for linear interpolation or spline interpolation to fill in the missing values. For outliers in the data to be repaired, use statistical features and expert rules to identify and correct them. For example, identify outliers based on the statistical distribution of the data and correct them based on expert experience. For example, use the linear interpolation method to fill in the missing values ​​in the July 2021 load forecast data, and correct the transformer data with abnormal parameters according to the rules formulated by experts. Merge the repaired data with the valid data to form preprocessed data.

[0026] Establish a unified data model. Design a unified data model to store and manage preprocessed data. The unified data model includes data identification fields, attribute description fields, and association fields. The data identification field is used to uniquely identify each piece of data, such as giving each data point a unique ID. The attribute description field is used to describe the specific meaning of the data, such as load value, voltage level, equipment type, etc. The association field is used to describe the association between data, such as the connection between a substation and a transmission line.

[0027] Map preprocessed data to a unified data model. Use ETL (Extract, Transform, Load) transformation rules to map preprocessed data to a unified data model to form standardized data. For example, convert historical data in different formats to a unified data format and store the data in a unified database. Create version identifiers for standardized data, record data change information, and generate a data traceability chain to track the data modification history.

[0028] Construct a vocabulary in the field of power grid planning. Extract basic terms for power grid planning from normative documents, industry standards, and technical literature and use them as initial entries. For example, extract terms such as "substation", "transmission line", and "load forecast". Use the term frequency-inverse document frequency (TF-IDF) algorithm to analyze the standardized data and extract feature words. For example, analyze the words that appear frequently and have distinctiveness in the standardized data, such as "peak-to-valley difference", "line loss", and "transformer capacity". Submit the extracted feature words to domain experts for review, and merge the reviewed feature words with the initial entries to form an initial vocabulary.

[0029] Perform multi-level semantic annotation on the initial vocabulary. Construct a multi-level semantic annotation system, including part-of-speech tagging layer, concept category tagging layer and relationship tagging layer. Use a sequence tagging model based on conditional random fields to perform part-of-speech tagging on the initial vocabulary, for example, tag "substation" as a noun and "connection" as a verb. Perform concept category tagging on the part-of-speech tagging results based on domain ontology, for example, tag "substation" as the "power equipment" category. Perform relationship tagging on the concept tagging results through dependency syntactic analysis to extract semantic relationships between words, such as extracting the "connection" relationship between "substation" and "transmission line". Finally, generate a grid planning domain vocabulary, which contains information such as the part-of-speech, concept category and semantic relationship of the vocabulary.

[0030] The solution of this application can: Improve data quality: Through data quality assessment and repair, the integrity, accuracy and timeliness of historical data of power grid planning are effectively improved, providing a reliable data foundation for subsequent planning analysis. Improve data utilization efficiency: By establishing a unified data model and standardized data, unified data management and sharing are achieved, data silos and data redundancy are avoided, and data utilization efficiency is improved. Promote knowledge accumulation and sharing: By constructing a vocabulary in the field of power grid planning, the standardized expression and accumulation of knowledge in the field of power grid planning are achieved, which promotes knowledge sharing and inheritance and provides support for the intelligent development of power grid planning.

[0031] In an optional implementation, a semantic vector space is constructed based on the power grid planning domain vocabulary, the power grid planning historical data is mapped to the semantic vector space, and basic semantic features of power grid planning are generated; according to power grid planning business rules, the basic semantic features of power grid planning are decomposed into multiple dimensions to form a multi-dimensional semantic feature library including time dimension features, space dimension features and business dimension features, including: Establish a semantic vector model in the field of power grid planning, wherein the semantic vector model in the field of power grid planning is constructed using the word2vec algorithm, and the training corpus of the word2vec algorithm includes entries in the vocabulary of the field of power grid planning and their context information; obtain a word vector mapping matrix based on the word2vec algorithm training, wherein the word vector mapping matrix records the high-dimensional vector representation corresponding to each entry; and construct a semantic vector space using the word vector mapping matrix; Performing word segmentation processing on the power grid planning historical data, identifying professional terms based on the power grid planning field vocabulary, and generating a word segmentation sequence; mapping the terms in the word segmentation sequence to the semantic vector space through the word vector mapping matrix to obtain the term semantic vector; performing weighted combination on the term semantic vector to generate basic semantic features of power grid planning; Acquire a power grid planning business rule base, the power grid planning business rule base includes a time dimension rule set, a space dimension rule set and a business dimension rule set; extract a time series correlation feature from the power grid planning basic semantic feature based on the time dimension rule set, the time series correlation feature describes the evolution law of power grid planning elements over time; Extracting spatial distribution features from the basic semantic features of power grid planning based on the spatial dimension rule set, wherein the spatial distribution features describe the geographical location relationship and topological connection relationship of power grid planning elements; extracting business logic features from the basic semantic features of power grid planning based on the business dimension rule set, wherein the business logic features describe the constraints and decision-making basis in the power grid planning process; The temporal correlation feature, the spatial distribution feature and the business logic feature are feature-fused to establish a multi-dimensional semantic feature index structure; and a multi-dimensional semantic feature library is constructed based on the multi-dimensional semantic feature index structure.

[0032] The method for extracting semantic features and building a multi-dimensional feature library for power grid planning is specifically implemented as follows: First, construct a vocabulary in the field of power grid planning. Collect and organize text materials such as professional terms, technical specifications, equipment names, etc. related to power grid planning, and after manual screening and sorting, form a vocabulary containing commonly used entries. For example, the vocabulary can include entries such as "transformer", "transmission line", "load forecast", and "distribution network planning". This vocabulary will serve as the basis for subsequent steps.

[0033] Then, a semantic vector space is constructed based on the vocabulary. The word2vec algorithm is used to train the vocabulary in the field of power grid planning. The training corpus includes the entries in the vocabulary and their context information, which can be extracted from documents, reports, specifications and other materials related to power grid planning. For example, for the entry "transformer", its context information can be "installation of transformer", "transformer capacity", "transformer maintenance", etc. Through the training of the word2vec algorithm, a word vector mapping matrix can be obtained. Using this matrix, a semantic vector space can be constructed, in which each entry is represented as a vector in the space, and the distance between entries with similar semantics in the space is also closer.

[0034] Next, the historical data of power grid planning is processed. Collect and organize the power grid planning data of previous years, including planning schemes, equipment information, operation data, etc. Perform word segmentation on these data, and identify professional terms based on the previously constructed power grid planning field vocabulary to generate a word segmentation sequence. For example, for a text describing "a new 110kV transmission line connecting substation A and substation B", after word segmentation and professional terminology recognition, the word segmentation sequence ["new", "110kV", "transmission line", "connection", "substation A", "substation B"] can be obtained. Each term in the word segmentation sequence is mapped to the semantic vector space through the word vector mapping matrix to obtain the semantic vector of each term. Then, the semantic vectors of these terms are weighted and combined to generate the basic semantic features of power grid planning. The weighted combination method can be adjusted according to factors such as the importance and frequency of occurrence of the terms.

[0035] Get the power grid planning business rule base. The rule base contains time dimension rule sets, space dimension rule sets, and business dimension rule sets. Time dimension rule sets include: load growth rate, equipment aging rate, etc.; space dimension rule sets include: substation distance, line length, etc.; business dimension rule sets include: power supply reliability requirements, investment budget restrictions, etc.

[0036] Extract multi-dimensional features based on business rules. Use the time dimension rule set to extract time-series correlation features from the basic semantic features of power grid planning, for example, analyze the load data of previous years and extract the load growth trend. Use the space dimension rule set to extract spatial distribution features, for example, analyze the geographical location information of substations and lines and extract the power grid topology. Use the business dimension rule set to extract business logic features, for example, analyze the power supply reliability requirements and investment budget and extract the constraints of the planning scheme.

[0037] Finally, a multi-dimensional semantic feature library is constructed. The extracted time-series correlation features, spatial distribution features, and business logic features are fused to establish a multi-dimensional semantic feature index structure. For example, features of different dimensions can be combined into a vector and indexed for easy retrieval. A multi-dimensional semantic feature library is constructed based on this index structure to facilitate subsequent power grid planning analysis and decision-making. For example, historical planning schemes that meet the conditions can be quickly retrieved based on the load growth trend and power grid topology of a certain area to provide a reference for new planning schemes.

[0038] The solution of this application can: Improve the efficiency of power grid planning: Through the construction of semantic vector space and the extraction of multi-dimensional features, complex power grid planning information can be converted into structured data, which is convenient for computer processing and analysis, thereby improving the efficiency of power grid planning. Improve the quality of power grid planning: The multi-dimensional semantic feature library contains rich historical planning information and business rules, which can provide reference for new planning schemes, avoid the limitations of human experience, and thus improve the quality of power grid planning. Support intelligent power grid planning: The multi-dimensional semantic feature library provides a data foundation for intelligent power grid planning, and can support the construction and application of power grid planning models based on artificial intelligence technology. For example, machine learning algorithms can be used to analyze historical planning data, predict future power grid development trends, and automatically generate optimized planning schemes.

[0039] In an optional implementation, an attention mechanism is used in the feature fusion layer to adaptively assign weights and fuse features of different dimensions; a multi-task learning module is set in the decision output layer to perform model training for power grid load prediction, grid structure optimization, and investment benefit evaluation, respectively, including: In the feature fusion layer, adaptive weight allocation is performed on the multi-dimensional features of the power grid planning, wherein the multi-dimensional features include power grid load features, grid structure features, and investment benefit features; an attention submodule is constructed for each dimensional feature, and each of the attention submodules includes a query vector generation unit, a key vector generation unit, and a value vector generation unit; The multi-dimensional features are respectively input into the corresponding attention submodules, and the query vector, key vector and value vector of the multi-dimensional features are generated by the query vector generation unit, the key vector generation unit and the value vector generation unit; the query vector and the key vector are subjected to dot product operation and scaled to obtain an attention score; the attention score is subjected to Softmax normalization to obtain the attention weight of the dimensional feature; The attention weight of the dimensional feature and the value vector are weightedly summed to obtain a weighted representation of the dimensional feature; the weighted representations of different dimensional features are fused, and the correlation strength between dimensions is calculated using a cross-dimensional attention mechanism; the weights of different dimensional features are adaptively assigned according to the correlation strength between dimensions to generate a fused feature vector; A multi-task learning module is set at the decision output layer, wherein the multi-task learning module includes a shared feature extraction layer and a task-specific layer; the fused feature vector is input into the shared feature extraction layer to extract high-level semantic features shared between tasks; the high-level semantic features are respectively input into three parallel task-specific layers for model training; The three parallel task-specific layers include a load forecasting training layer, a grid structure optimization training layer and an investment benefit evaluation training layer; the load forecasting training layer uses a time series deep learning model to model the grid load characteristics; the grid structure optimization training layer uses a graph neural network to analyze the grid topology characteristics; the investment benefit evaluation training layer uses a deep reinforcement learning model to train the investment decision-making process.

[0040] The intelligent power grid planning method aims to realize intelligent power grid planning through multi-dimensional feature fusion and multi-task learning of power grid load, grid structure and investment benefits.

[0041] First, collect and preprocess historical grid data. For example, collect power load data, grid topology data, and related investment cost and benefit data for the past five years. Clean the collected data, remove outliers and missing values, and perform normalization, such as scaling the data to between 0 and 1, to ensure that feature data of different dimensions are comparable.

[0042] Then, construct a multi-dimensional feature representation. Convert the preprocessed data into a feature vector. For example, take the load data of 24 hours a day as a 24-dimensional load feature vector; use the adjacency matrix to represent the power grid topology, and flatten the adjacency matrix into a vector as the grid structure feature vector; and form a vector of indicators such as investment cost and expected return as the investment benefit feature vector. Assuming that the load data of a certain day is [100,110,...,150], the load feature vector of that day is [100,110,...,150]. Assuming that the cost of an investment is 1 million and the expected return is 1.2 million, the investment benefit feature vector is [100,120].

[0043] Next, feature fusion is performed. Attention submodules are constructed for load features, grid structure features, and investment benefit features respectively. Each attention submodule contains a query vector generation unit, a key vector generation unit, and a value vector generation unit. For example, for the load feature vector, the corresponding query vector, key vector, and value vector are generated by the three units respectively. The query vector and the key vector are scaled after the dot product operation to obtain the attention score. The attention score is normalized by Softmax to obtain the attention weight of the load feature. The attention weight is weighted and summed with the value vector to obtain the weighted representation of the load feature. The same operation is performed on other dimensional features. Then, a cross-dimensional attention mechanism is used to calculate the correlation strength between dimensions, such as calculating the correlation strength between load features and grid structure features. The weighted representations of features of different dimensions are weighted summed according to the correlation strength between dimensions to obtain a fused feature vector.

[0044] Subsequently, the fused feature vector is input into the multi-task learning module. The module includes a shared feature extraction layer and a task-specific layer. The shared feature extraction layer is used to extract high-level semantic features shared between tasks. For example, a multi-layer perceptron is used to extract high-level features. The extracted high-level semantic features are respectively input into three parallel task-specific layers for model training. The three task-specific layers are load forecasting training layer, grid structure optimization training layer, and investment benefit evaluation training layer. The load forecasting training layer uses a time series deep learning model, such as a long short-term memory network, to model the load characteristics of the power grid. The grid structure optimization training layer uses a graph neural network to analyze the topological structure characteristics of the power grid. The investment benefit evaluation training layer uses a deep reinforcement learning model to train the investment decision-making process.

[0045] Finally, the power grid planning scheme is output, for example, the load forecast results for a period of time in the future, the optimized power grid topology structure and the investment benefit evaluation results are output.

[0046] The solution of this application can: Improve prediction accuracy: Through multi-dimensional feature fusion and attention mechanism, the correlation between different features can be captured, thereby improving the accuracy of load forecasting, grid structure optimization and investment benefit evaluation. Optimize planning scheme: Through multi-task learning, load forecasting, grid structure optimization and investment benefit evaluation can be considered at the same time, so as to obtain a more comprehensive and optimized power grid planning scheme. Enhance decision-making efficiency: Through intelligent model training and prediction, manual intervention can be reduced, thereby improving the efficiency of power grid planning.

[0047] In an optional implementation, real-time grid planning demand data is received, and the real-time grid planning demand data is input into a trained deep learning model; an initial grid planning plan is output based on the deep learning model, and the initial grid planning plan includes load distribution prediction results, grid structure configuration plans, and investment scale recommendations, including: Receive real-time power grid planning demand data through a data receiving module, wherein the real-time power grid planning demand data includes power grid operation status data, historical load distribution data, current grid topology data and investment constraint condition data; Preprocessing the real-time power grid planning demand data, cleaning the real-time power grid planning demand data, and removing abnormal values; performing time series alignment on the real-time power grid planning demand data; normalizing the real-time power grid planning demand data to obtain preprocessed power grid planning demand data; Input the preprocessed power grid planning demand data into a trained deep learning model, wherein the deep learning model includes a data encoding unit and a solution generation unit; the data encoding unit uses a long short-term memory network to process time series data features, uses a graph convolutional network to process network topology features, and uses a multi-layer perceptron to process investment constraint features; the solution generation unit generates an initial power grid planning solution based on the output features of the data encoding unit; The deep learning model generates the initial grid planning plan through a parallel processing mechanism, and the initial grid planning plan includes: using a probabilistic prediction method to generate load distribution prediction results, and the load distribution prediction results include a mean prediction and a confidence interval of future loads; using a graph optimization algorithm to generate a grid structure configuration plan, and the grid structure configuration plan includes a substation site selection plan and a line corridor planning plan; based on cost-benefit analysis, an investment scale proposal is generated, and the investment scale proposal includes a construction cost budget and a phased investment proposal.

[0048] Receive real-time grid planning demand data and input it into the trained deep learning model to generate an initial grid planning plan, which includes load distribution prediction results, grid structure configuration plan and investment scale recommendations.

[0049] First, the real-time grid planning demand data is obtained through the data receiving module. These data include grid operation status data (e.g., current power load, voltage level, line load, etc.), historical load distribution data (e.g., daily / hourly load data in the past few years, and factors affecting load, such as weather, holidays, etc.), current grid topology data (e.g., location and capacity of substations, connection and capacity of transmission lines, etc.), and investment constraint data (e.g., budget constraints, environmental regulations, land use restrictions, etc.).

[0050] For example, suppose we receive data including: hourly load data for a certain region over the past three years, the current grid topology (including 2 substations and 5 transmission lines), and an investment budget of RMB 100 million for the next five years.

[0051] Next, the received data is preprocessed. This includes data cleaning, such as removing outliers and missing values; time series alignment, such as aligning all time series data to the same time resolution; and normalization, such as scaling all data to the same numerical range.

[0052] For example, during data cleaning, we may find that the load data at certain time points is abnormally high, which may be caused by data collection errors or special events. We can replace these outliers by using the mean or median. In terms of time series alignment, we can convert all data into hourly data. In terms of normalization, we can scale all data to between 0 and 1.

[0053] The preprocessed data will be input into the trained deep learning model. The model consists of a data encoding unit and a solution generation unit. The data encoding unit uses a long short-term memory network (LSTM) to process time series data features (such as historical load data), a graph convolutional network (GCN) to process network topology features (such as current grid topology data), and a multi-layer perceptron (MLP) to process investment constraint features (such as budget constraints). The output features of these units will be integrated together.

[0054] For example, LSTM networks learn temporal patterns in historical load data, such as daily and seasonal fluctuations. GCN networks learn spatial relationships in the grid topology, such as the connections between substations and transmission lines. MLP networks learn the impact of investment constraints, such as the impact of budget constraints on grid structure planning.

[0055] The scheme generation unit generates an initial scheme for power grid planning based on the output features of the data encoding unit. This includes using a probabilistic forecasting method to generate load distribution forecast results, including the mean forecast and confidence interval of future loads; using a graph optimization algorithm to generate a grid structure configuration scheme, including a substation site selection scheme and a line corridor planning scheme; and generating an investment scale proposal based on a cost-benefit analysis, including a construction cost budget and a phased investment proposal.

[0056] For example, the model might predict that peak load will grow by 20% over the next five years and provide a 95% confidence interval. It might also recommend building a new substation at a specific location and building a new transmission line to connect to the existing grid. It might also recommend phasing the investment, for example, $30 million in the first year, $40 million in the second year, and $30 million in the remaining three years.

[0057] The deep learning model generates an initial plan for power grid planning through a parallel processing mechanism, thereby improving computational efficiency.

[0058] The solution of this application can: Improve grid planning efficiency: By automatically generating initial grid planning plans, manual intervention and time costs can be significantly reduced, thereby improving grid planning efficiency. Deep learning models can quickly process large amounts of historical data and real-time data and generate high-quality planning plans, thereby accelerating the planning cycle. Optimize grid planning plans: Deep learning models can learn complex grid operation rules and investment constraints and generate better grid planning plans. Compared with traditional methods, this technical solution can more accurately predict future loads and optimize grid structure configurations, thereby improving the reliability and economy of the grid. Enhance the adaptability of grid planning: This technical solution can adapt to the ever-changing grid operating environment and needs. By receiving real-time data and making dynamic adjustments, the solution can generate more adaptive grid planning plans to better cope with future uncertainties.

[0059] In an optional implementation, the initial grid planning scheme is iteratively optimized using a reinforcement learning algorithm, an optimization objective function is constructed according to grid reliability indicators, economic indicators, and environmental protection indicators, and the output of the optimal grid planning scheme includes: The initial plan of the power grid planning is subjected to data standardization processing to obtain standardized planning data including substation layout data, line corridor data and investment configuration data; a reinforcement learning optimization model is established based on the standardized planning data to construct a state space, wherein the state space records the current substation operation status, line load status and investment execution status; an action space is constructed, wherein the action space includes a substation capacity adjustment amount, a line corridor adjustment amount and an investment scale adjustment amount; Based on the operation data in the state space, a power grid reliability index is constructed, and the power supply reliability rate, voltage qualification rate and network loss rate are obtained through power flow calculation and reliability analysis; based on the adjustment data in the action space, an economic index is constructed, and the project construction cost, operation and maintenance cost and benefit payback period are obtained through investment benefit analysis; based on the combined data of the state space and the action space, an environmental protection index is constructed, and the land occupation area, electromagnetic radiation intensity and carbon emissions are obtained through environmental impact assessment; the power grid reliability index, economic index and environmental protection index are combined into an optimization objective function, and the objective function value is calculated by weighted summation, and the objective function value is used as a reward signal for reinforcement learning; A deep Q learning network is constructed based on the state space, action space and optimization objective function, wherein the deep Q learning network maps the current state to the optimal action selection; the state transition sequence after the action is executed and the corresponding reward signal are stored in an experience replay pool; training samples are selected from the experience replay pool, and the time difference error of each sample is calculated; the samples are sorted according to the time difference error, and samples with larger errors are preferentially selected for network training; The deep Q learning network is updated based on the selected training samples, and the loss between the network prediction value and the target value is calculated using the optimization objective function; the network parameters are optimized by the back propagation algorithm so that the action selection output by the network is closer to the optimal strategy; the action selection, experience replay and network update processes are repeated, and when the change in the optimization objective function value of consecutive preset rounds is less than the convergence threshold, the optimal power grid planning scheme is output.

[0060] The initial plan for grid planning can be obtained by collecting information such as the topology of the existing grid, equipment parameters, load data, etc., and combining it with future power demand forecasts and development plans. For example, the location and capacity of substations in a certain area, the path, length, and current carrying capacity of transmission lines, and power load data in different time periods can be collected. In addition, factors such as new energy access and distributed power generation development can also be considered to form an initial plan for grid planning.

[0061] Next, the initial plan for power grid planning is standardized. For example, the substation capacity data is scaled to between 0 and 1, the line length data is scaled to between 0 and 1, and the investment amount data is scaled to between 0 and 1. The purpose of standardization is to eliminate the influence of different data dimensions and dimensions, and to facilitate the training and optimization of subsequent models. Assuming that the capacity of a substation is 500 MVA and the maximum capacity is 1000 MVA, its standardized capacity value is 0.5.

[0062] A reinforcement learning optimization model is established based on standardized planning data. The state space records the current operating status of the power grid, such as the load rate of each substation, the load rate of the line, the amount of investment, etc. The action space includes adjustments to the power grid planning scheme, such as increasing or decreasing the capacity of the substation, adjusting the line path, increasing or decreasing the scale of investment, etc. Taking the substation capacity adjustment as an example, the action can be defined as increasing or decreasing the capacity by 10%.

[0063] Construct grid reliability indicators, economic indicators and environmental indicators. Reliability indicators can be obtained through power flow calculation and reliability analysis, such as power supply reliability, voltage qualification rate and network loss rate. Assume that the power supply reliability rate of a certain scheme is 99.5%, the voltage qualification rate is 99%, and the network loss rate is 2% through power flow calculation. Economic indicators can be obtained through investment benefit analysis, such as engineering construction cost, operation and maintenance cost and benefit payback period. Assume that the engineering construction cost of a certain scheme is 100 million yuan, the operation and maintenance cost is 1 million yuan per year, and the benefit payback period is 10 years. Environmental indicators can be obtained through environmental impact assessment, such as land occupation area, electromagnetic radiation intensity and carbon emissions. Assume that the land occupation area of ​​a certain scheme is 10 hectares, the electromagnetic radiation intensity meets the national standards, and the carbon emissions are 1,000 tons.

[0064] These indicators are combined into an optimization objective function, and the objective function value is calculated by weighted summation. The weight can be adjusted according to actual needs. For example, if reliability is given more attention, a higher weight can be given to the reliability indicator.

[0065] Build a deep Q-learning network to map the current state to the optimal action selection. Store the state transition sequence and the corresponding reward signal after the action is executed in the experience replay pool. For example, store the initial state, the action executed, the new state, and the reward obtained. Select training samples from the experience replay pool and calculate the time difference error of each sample. Sort the samples by importance according to the time difference error, and give priority to samples with larger errors for network training.

[0066] The deep Q learning network is updated based on the selected training samples, and the loss between the network prediction value and the target value is calculated using the optimization objective function. The network parameters are optimized through the back propagation algorithm to make the action selection output by the network closer to the optimal strategy. The action selection, experience playback and network update process are repeated. When the change of the optimization objective function value of consecutive preset rounds is less than the convergence threshold, for example, the change of the objective function value of 100 consecutive rounds is less than 0.01, the optimal power grid planning scheme is output.

[0067] The solution of this application can: Improve the efficiency of power grid planning: automatically search for the optimal planning scheme through reinforcement learning algorithms, reduce manual intervention, and shorten the planning cycle. Improve the quality of power grid planning: comprehensively consider reliability, economy, and environmental protection indicators, and find the optimal balance point through intelligent optimization algorithms to formulate more scientific and reasonable planning schemes. Reduce the cost of power grid planning: reduce engineering construction costs and operation and maintenance costs and improve investment benefits by optimizing power grid structure and investment strategies.

[0068] Figure 2 FIG. 1 is a schematic diagram of a structure of a power grid planning, construction, operation and maintenance integrated modeling system based on a semantic system according to an embodiment of the present invention. Figure 2 As shown, the system comprises: The first unit is used to obtain historical data of power grid planning, wherein the historical data of power grid planning includes load forecast data, power grid topology data and equipment parameter data; semantically annotate the historical data of power grid planning and establish a vocabulary of power grid planning domain; construct a semantic vector space based on the vocabulary of power grid planning domain, map the historical data of power grid planning to the semantic vector space, and generate basic semantic features of power grid planning; perform multi-dimensional decomposition of the basic semantic features of power grid planning according to power grid planning business rules to form a multi-dimensional semantic feature library including time dimension features, space dimension features and business dimension features; The second unit is used to construct a multi-layer neural network structure, which includes a feature extraction layer, a feature fusion layer and a decision output layer; the feature data in the multi-dimensional semantic feature library is input into the feature extraction layer, and local features are extracted by convolution operation; the attention mechanism is used in the feature fusion layer to perform adaptive weight allocation and feature fusion on features of different dimensions; a multi-task learning module is set in the decision output layer to perform model training for power grid load prediction, grid structure optimization and investment benefit evaluation respectively; The third unit is used to receive real-time grid planning demand data, input the real-time grid planning demand data into a trained deep learning model; output an initial grid planning plan based on the deep learning model, the initial grid planning plan includes load distribution prediction results, grid structure configuration plan and investment scale recommendations; use a reinforcement learning algorithm to iteratively optimize the initial grid planning plan, construct an optimization objective function based on grid reliability indicators, economic indicators and environmental protection indicators, and output an optimal grid planning plan.

[0069] According to a third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0070] A fourth aspect of the embodiments of the present invention is: A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0071] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A semantic system-based integrated modeling method for power grid planning, construction, operation and maintenance, characterized by: include: Acquiring historical data of power grid planning, wherein the historical data of power grid planning includes load forecast data, power grid topology data and equipment parameter data; Performing semantic annotation on the historical data of power grid planning and establishing a vocabulary in the field of power grid planning; Constructing a semantic vector space based on the grid planning domain vocabulary, mapping the grid planning historical data into the semantic vector space, and generating basic semantic features of the grid planning; performing multi-dimensional decomposition of the basic semantic features of the grid planning according to grid planning business rules, and forming a multi-dimensional semantic feature library including time dimension features, space dimension features, and business dimension features; Constructing a multi-layer neural network structure, the multi-layer neural network structure includes a feature extraction layer, a feature fusion layer and a decision output layer; inputting feature data in the multi-dimensional semantic feature library into the feature extraction layer, extracting local features through convolution operation; using an attention mechanism in the feature fusion layer to perform adaptive weight allocation and feature fusion on features of different dimensions; setting a multi-task learning module in the decision output layer to perform model training for power grid load prediction, grid structure optimization and investment benefit evaluation respectively; Receiving real-time power grid planning demand data, and inputting the real-time power grid planning demand data into the trained deep learning model; Outputting an initial plan for power grid planning based on the deep learning model, wherein the initial plan for power grid planning includes load distribution prediction results, grid structure configuration plan and investment scale recommendations; The reinforcement learning algorithm is used to iteratively optimize the initial power grid planning plan, and an optimization objective function is constructed according to power grid reliability indicators, economic indicators and environmental protection indicators to output the optimal power grid planning plan.

2. The method according to claim 1, characterized in that: Acquiring historical data of power grid planning, wherein the historical data of power grid planning includes load forecast data, power grid topology data and equipment parameter data; The power grid planning historical data is semantically annotated to establish a power grid planning domain vocabulary including: Acquire historical data of power grid planning through a data acquisition interface, wherein the historical data of power grid planning includes load forecast data, power grid topology data, and equipment parameter data; establish a data quality scoring index system for the historical data of power grid planning, wherein the data quality scoring index system includes a data integrity index, a data accuracy index, and a data timeliness index; perform a quality assessment on the historical data of power grid planning based on the data quality scoring index system, and generate a quality assessment report; According to the evaluation results in the quality evaluation report, the historical data of the power grid planning is graded and divided into valid data, data to be repaired and invalid data; missing values ​​in the data to be repaired are repaired by using an interpolation method based on time series correlation, and abnormal values ​​in the data to be repaired are identified and corrected by using statistical features and expert rules to generate repaired data; the valid data and the repaired data are merged to form preprocessed data; Design a unified data model, which includes a data identification field, an attribute description field, and an association relationship field; map the preprocessed data to the unified data model through ETL transformation rules to form standardized data; establish a version identification for the standardized data, record data change information, and generate a data traceability chain; Extract basic terms for power grid planning from normative documents and use the basic terms as initial entries; use a word frequency-inverse document frequency algorithm to analyze the standardized data and extract characteristic words; submit the characteristic words to domain experts for review, and merge the reviewed characteristic words with the initial entries to form an initial vocabulary; A multi-level semantic annotation system is constructed, which includes a part-of-speech tagging layer, a concept category tagging layer and a relationship tagging layer; the initial vocabulary is tagged with parts of speech using a sequence tagging model based on conditional random fields to generate part-of-speech tagging results; the part-of-speech tagging results are tagged with concept categories based on domain ontology to generate concept tagging results; the concept tagging results are tagged with relationships through dependency syntactic analysis to extract semantic relationships between words and generate a power grid planning domain vocabulary.

3. The method according to claim 1, characterized in that A semantic vector space is constructed based on the grid planning domain vocabulary, and the grid planning historical data is mapped into the semantic vector space to generate basic semantic features of grid planning; according to grid planning business rules, the basic semantic features of grid planning are decomposed into multiple dimensions to form a multi-dimensional semantic feature library including time dimension features, space dimension features and business dimension features, including: Establish a semantic vector model in the field of power grid planning, wherein the semantic vector model in the field of power grid planning is constructed using the word2vec algorithm, and the training corpus of the word2vec algorithm includes entries in the vocabulary of the field of power grid planning and their context information; obtain a word vector mapping matrix based on the word2vec algorithm training, wherein the word vector mapping matrix records the high-dimensional vector representation corresponding to each entry; and construct a semantic vector space using the word vector mapping matrix; Performing word segmentation processing on the power grid planning historical data, identifying professional terms based on the power grid planning field vocabulary, and generating a word segmentation sequence; mapping the terms in the word segmentation sequence to the semantic vector space through the word vector mapping matrix to obtain the term semantic vector; performing weighted combination on the term semantic vector to generate basic semantic features of power grid planning; Acquire a power grid planning business rule base, the power grid planning business rule base includes a time dimension rule set, a space dimension rule set and a business dimension rule set; extract a time series correlation feature from the power grid planning basic semantic feature based on the time dimension rule set, the time series correlation feature describes the evolution law of power grid planning elements over time; Extracting spatial distribution features from the basic semantic features of power grid planning based on the spatial dimension rule set, wherein the spatial distribution features describe the geographical location relationship and topological connection relationship of power grid planning elements; extracting business logic features from the basic semantic features of power grid planning based on the business dimension rule set, wherein the business logic features describe the constraints and decision-making basis in the power grid planning process; The temporal correlation feature, the spatial distribution feature and the business logic feature are feature-fused to establish a multi-dimensional semantic feature index structure; and a multi-dimensional semantic feature library is constructed based on the multi-dimensional semantic feature index structure.

4. The method according to claim 1, characterized in that In the feature fusion layer, an attention mechanism is used to perform adaptive weight allocation and feature fusion on features of different dimensions; A multi-task learning module is set in the decision output layer to perform model training for power grid load prediction, grid structure optimization and investment benefit evaluation respectively, including: In the feature fusion layer, adaptive weight allocation is performed on the multi-dimensional features of the power grid planning, wherein the multi-dimensional features include power grid load features, grid structure features, and investment benefit features; an attention submodule is constructed for each dimensional feature, and each of the attention submodules includes a query vector generation unit, a key vector generation unit, and a value vector generation unit; The multi-dimensional features are respectively input into the corresponding attention submodules, and the query vector, key vector and value vector of the multi-dimensional features are generated by the query vector generation unit, the key vector generation unit and the value vector generation unit; the query vector and the key vector are subjected to dot product operation and scaled to obtain an attention score; the attention score is subjected to Softmax normalization to obtain the attention weight of the dimensional feature; The attention weight of the dimensional feature and the value vector are weightedly summed to obtain a weighted representation of the dimensional feature; the weighted representations of different dimensional features are fused, and the correlation strength between dimensions is calculated using a cross-dimensional attention mechanism; the weights of different dimensional features are adaptively assigned according to the correlation strength between dimensions to generate a fused feature vector; A multi-task learning module is set at the decision output layer, wherein the multi-task learning module includes a shared feature extraction layer and a task-specific layer; the fused feature vector is input into the shared feature extraction layer to extract high-level semantic features shared between tasks; the high-level semantic features are respectively input into three parallel task-specific layers for model training; The three parallel task-specific layers include a load forecasting training layer, a grid structure optimization training layer and an investment benefit evaluation training layer; the load forecasting training layer uses a time series deep learning model to model the grid load characteristics; the grid structure optimization training layer uses a graph neural network to analyze the grid topology characteristics; the investment benefit evaluation training layer uses a deep reinforcement learning model to train the investment decision-making process.

5. The method according to claim 1, characterized in that Receiving real-time power grid planning demand data, and inputting the real-time power grid planning demand data into the trained deep learning model; Outputting an initial plan for power grid planning based on the deep learning model, the initial plan for power grid planning includes load distribution prediction results, grid structure configuration plan and investment scale recommendations, including: Receive real-time power grid planning demand data through a data receiving module, wherein the real-time power grid planning demand data includes power grid operation status data, historical load distribution data, current grid topology data and investment constraint condition data; Preprocessing the real-time power grid planning demand data, cleaning the real-time power grid planning demand data, and removing abnormal values; performing time series alignment on the real-time power grid planning demand data; normalizing the real-time power grid planning demand data to obtain preprocessed power grid planning demand data; Input the preprocessed power grid planning demand data into a trained deep learning model, wherein the deep learning model includes a data encoding unit and a solution generation unit; the data encoding unit uses a long short-term memory network to process time series data features, uses a graph convolutional network to process network topology features, and uses a multi-layer perceptron to process investment constraint features; the solution generation unit generates an initial power grid planning solution based on the output features of the data encoding unit; The deep learning model generates the initial grid planning plan through a parallel processing mechanism, and the initial grid planning plan includes: using a probabilistic prediction method to generate load distribution prediction results, and the load distribution prediction results include a mean prediction and a confidence interval of future loads; using a graph optimization algorithm to generate a grid structure configuration plan, and the grid structure configuration plan includes a substation site selection plan and a line corridor planning plan; based on cost-benefit analysis, an investment scale proposal is generated, and the investment scale proposal includes a construction cost budget and a phased investment proposal.

6. The method according to claim 1, characterized in that The reinforcement learning algorithm is used to iteratively optimize the initial power grid planning scheme, and the optimization objective function is constructed according to the power grid reliability index, economic index and environmental protection index. The output of the optimal power grid planning scheme includes: The initial plan of the power grid planning is subjected to data standardization processing to obtain standardized planning data including substation layout data, line corridor data and investment configuration data; a reinforcement learning optimization model is established based on the standardized planning data to construct a state space, wherein the state space records the current substation operation status, line load status and investment execution status; an action space is constructed, wherein the action space includes a substation capacity adjustment amount, a line corridor adjustment amount and an investment scale adjustment amount; Based on the operation data in the state space, a power grid reliability index is constructed, and the power supply reliability rate, voltage qualification rate and network loss rate are obtained through power flow calculation and reliability analysis; based on the adjustment data in the action space, an economic index is constructed, and the project construction cost, operation and maintenance cost and benefit payback period are obtained through investment benefit analysis; based on the combined data of the state space and the action space, an environmental protection index is constructed, and the land occupation area, electromagnetic radiation intensity and carbon emissions are obtained through environmental impact assessment; the power grid reliability index, economic index and environmental protection index are combined into an optimization objective function, and the objective function value is calculated by weighted summation, and the objective function value is used as a reward signal for reinforcement learning; A deep Q learning network is constructed based on the state space, action space and optimization objective function, wherein the deep Q learning network maps the current state to the optimal action selection; the state transition sequence after the action is executed and the corresponding reward signal are stored in an experience replay pool; training samples are selected from the experience replay pool, and the time difference error of each sample is calculated; the samples are sorted according to the time difference error, and samples with larger errors are preferentially selected for network training; The deep Q learning network is updated based on the selected training samples, and the loss between the network prediction value and the target value is calculated using the optimization objective function; the network parameters are optimized by the back propagation algorithm so that the action selection output by the network is closer to the optimal strategy; the action selection, experience replay and network update processes are repeated, and when the change in the optimization objective function value of consecutive preset rounds is less than the convergence threshold, the optimal power grid planning scheme is output.

7. A semantic-based integrated modeling system for power grid planning, construction, operation and maintenance, used to implement the method described in any one of claims 1 to 6, characterized in that: include: The first unit is used to obtain historical data of power grid planning, wherein the historical data of power grid planning includes load forecast data, power grid topology data and equipment parameter data; Performing semantic annotation on the historical data of power grid planning and establishing a vocabulary in the field of power grid planning; Constructing a semantic vector space based on the grid planning domain vocabulary, mapping the grid planning historical data into the semantic vector space, and generating basic semantic features of the grid planning; performing multi-dimensional decomposition of the basic semantic features of the grid planning according to grid planning business rules, and forming a multi-dimensional semantic feature library including time dimension features, space dimension features, and business dimension features; The second unit is used to construct a multi-layer neural network structure, which includes a feature extraction layer, a feature fusion layer and a decision output layer; the feature data in the multi-dimensional semantic feature library is input into the feature extraction layer, and local features are extracted by convolution operation; the attention mechanism is used in the feature fusion layer to perform adaptive weight allocation and feature fusion on features of different dimensions; a multi-task learning module is set in the decision output layer to perform model training for power grid load prediction, grid structure optimization and investment benefit evaluation respectively; A third unit is used to receive real-time power grid planning demand data, and input the real-time power grid planning demand data into the trained deep learning model; Outputting an initial plan for power grid planning based on the deep learning model, wherein the initial plan for power grid planning includes load distribution prediction results, grid structure configuration plan and investment scale recommendations; The reinforcement learning algorithm is used to iteratively optimize the initial power grid planning plan, and an optimization objective function is constructed according to power grid reliability indicators, economic indicators and environmental protection indicators to output the optimal power grid planning plan.

8. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images