Scheduling plan rationality associated factor tracing method and system

Abstract

The invention discloses a scheduling plan rationality associated factor tracing method and system, and belongs to the technical field of power system scheduling control, and the method comprises the steps: reading power grid related data, and constructing a power grid continuous operation data section based on a scheduling plan; establishing a multi-index evaluation system, and fusing subjective and objective weights to calculate a rationality comprehensive score; obtaining a high-value feature subset through a multi-stage screening mechanism; modeling a power grid as an undirected graph, and constructing multi-scale traceability features; and outputting an evaluation result and an associated factor traceability result by adopting a GAT and Bi-GRU fusion model. According to the method, data driving and expert experience are fused, power grid space-time coupling characteristics are accurately captured, scheduling plan rationality beforehand evaluation and associated factor accurate traceability are realized, scheduling plan optimization efficiency and power grid operation safety and economy are improved, and the method is suitable for a novel power system scheduling control scene.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power system dispatching control, and particularly relates to a dispatching plan rationality associated factor tracing method and system. BACKGROUND

[0002] With the in-depth promotion of new power system construction, high penetration rate of intermittent new energy power generation (such as wind power and photovoltaic power) is integrated into the power grid on a large scale, and a large amount of flexible load (such as electric vehicles and energy storage power stations) gradually participates in the response of the power grid, which significantly aggravates the double uncertainty of the power supply side and the load side of the power grid, and promotes the power grid dispatching mode to change from the traditional single static balance mode of hierarchical partition and source following load to the comprehensive dynamic balance mode of fully considering the uncertainty of source and load and the regulation capacity of the whole network. As the core basis for power grid operation, the dispatching plan usually covers maintenance plans, power generation plans and tie line plans, and its formulation needs to consider power reliability, operation economy and environmental protection. At present, the evaluation of the rationality of the dispatching plan in the field is mainly focused on safety and economy, and the evaluation method adopts a single subjective weight (such as relying on expert experience for assignment) or a single objective weight (such as only based on historical data statistics). In the aspect of associated factor analysis, the factors affecting the prediction error of source and load are mainly screened through simple correlation analysis. In the aspect of tracing modeling, a single time sequence model (such as a recurrent neural network) or a traditional machine learning model (such as a support vector machine) is mainly used to process related data to try to identify the correlation between the dispatching plan and the power grid operation state.

[0003] However, the existing technology still has obvious limitations in practical application: Firstly, the rationality evaluation dimension of the dispatching plan is not comprehensive enough, and the single weight method cannot balance the professionalism of expert experience and the objectivity of data internal structure, resulting in low score result differentiation and inability to accurately quantify the advantages and disadvantages of the dispatching plan in different scenarios. Secondly, the screening of the source and load prediction error associated factors lacks a systematic multi-stage verification mechanism, and only through simple correlation analysis may retain irrelevant or weakly correlated variables, causing feature redundancy and reducing subsequent tracing efficiency. Thirdly, the tracing model cannot effectively adapt to the topological characteristics of the power grid and the dynamic evolution characteristics of the dispatching process, the single time sequence model cannot capture the spatial dependence relationship between the nodes of the power grid, and the traditional machine learning model cannot fully depict the time sequence evolution rule of the power grid state guided by the dispatching plan, resulting in insufficient accuracy of the associated factor tracing, difficulty in identifying potential problems of the dispatching plan in advance, and thus possible power curtailment, insufficient system peak regulation capacity, lack of dispatching flexibility and other power grid operation risks, which cannot meet the needs of the new power system for prior evaluation and accurate tracing of the dispatching plan. SUMMARY

[0004] The technical problem solved by the present application is to provide a dispatch plan rationality associated factor tracing method and system to solve the technical problem that in the background of a new power system, due to the intensification of source and load uncertainty and source and load prediction error, the existing technology cannot scientifically and quantitatively make forward-looking evaluation on the dispatch plan, and cannot accurately locate the key associated factors affecting the rationality of the plan before the plan is executed, thereby failing to effectively guide the plan optimization and power grid operation risk prevention and control.

[0005] The present application adopts the following technical solutions: A dispatch plan rationality associated factor tracing method, comprising the following steps: Read the power grid primary equipment model data and historical measurement data to obtain dispatch plan data at the same time scale; after organization processing, generate power grid model data samples, measurement data samples and prediction data samples; replace the historical measurement data with the dispatch plan data to obtain a power grid continuous operation data section based on the dispatch plan; Based on the power grid continuous operation data section, use the measurement data samples and the prediction data samples to establish a multi-index evaluation system, fuse subjective weights and objective weights to obtain comprehensive weights, and calculate the rationality comprehensive score of the dispatch plan to be evaluated through ideal solution and negative ideal solution; Take the rationality comprehensive score as the response variable, construct a candidate feature set based on the error characteristics of the prediction data samples, and remove irrelevant or weakly related variables through a multi-stage screening mechanism to obtain a high-value feature subset composed of key elements strongly associated with source and load prediction error; Based on the power grid model data samples, model the power grid as an undirected graph, based on the high-value feature subset, combine the node historical operation state provided by the measurement data samples, perform feature extraction and encoding operations that fuse global context information and node-level spatiotemporal fine-grained information, construct a multi-scale feature representation system that fuses global context information and node local state, and generate tracing features representing the power grid operation state; A tracing model is constructed by combining a graph attention network with a bidirectional gated recurrent unit to encode and process the tracing features, output a dispatch plan rationality evaluation result, and realize associated factor tracing.

[0006] Preferably, the time interval of the measurement data and the prediction data is unified as 15 minutes; the dispatch plan data includes generation plan, tie line plan and maintenance plan; the organization processing is performed according to the model data service interface data structure of the new generation dispatching technical support system.

[0007] Preferably, the multi-index evaluation system covers the dimensions of safety, economy, environmental protection and dispatchability, and the optimization direction of each index is clear; the subjective weight is calculated by the analytic hierarchy process, the objective weight is calculated by the entropy weight method, and the fusion is weighted fusion.

[0008] Preferably, the rationality comprehensive score of the to-be-evaluated dispatch plan is calculated by ideal solution and negative ideal solution, and specifically includes: According to the optimal value and the worst value of each index in the historical sample, an ideal solution and a negative ideal solution are constructed, and the weighted Euclidean distance between the index value of the to-be-evaluated dispatch plan and the two reference points is calculated And The rationality comprehensive score of the dispatch plan under the 100-point system is calculated by the method of approaching ideal solution .

[0009] Preferably, the multi-stage screening mechanism specifically includes: Spearman rank correlation coefficient is used to evaluate the monotonic correlation between each candidate feature and the response variable, and the features with significant correlation are retained to form a preliminary screening feature subset; An XGBoost regression model is constructed based on the preliminary screening feature subset, and the importance measure of each feature is calculated by using the model; According to the feature importance ranking output by the XGBoost model, combined with the preset cumulative importance proportion or importance threshold, the high-value feature subset is screened from the preliminary screening feature subset.

[0010] Preferably, in the features with significant correlation, the judgment condition of significant correlation is that the p value is less than the preset threshold or the absolute value of the correlation coefficient is higher than the set threshold; the screening also includes the preset cumulative importance proportion or importance threshold.

[0011] Preferably, the multi-scale feature representation system of constructing fusion global context information and node local state specifically includes: Meteorological features, date and holiday type features, and historical load curve statistical features are extracted from the high-value feature subset, then spliced to form a global feature vector, and then mapped to a low-dimensional global context vector by a multilayer perception ; For each node in the undirected graph, its static features are extracted and encoded into static embedding by a multilayer perception , and its dynamic historical time series features are extracted and encoded into dynamic embedding by a gated recurrent unit ; The static embedding and the dynamic embedding are spliced to form a node-level fusion feature , which is used as the traceability feature.

[0012] Preferably, the power grid is modeled as an undirected graph , denotes a set of nodes, denotes a set of edges, denotes an adjacency matrix.

[0013] Preferably, the static features include voltage amplitude, voltage phase angle, unit active / reactive power output, load active / reactive power of the previous day; the dynamic features are composed of historical time series with an interval of 15 minutes in the past 7 days.

[0014] Preferably, the tracing model is constructed by combining the graph attention network with the bidirectional gated recurrent unit, specifically including: fusing the global context vector with the node-level fusion features of each node to form initial features of each node; inputting the initial features into the graph attention network for multi-layer encoding, performing information aggregation by adaptively calculating attention coefficients between nodes, and obtaining a node feature matrix containing spatial dependency information; based on the node feature matrix, fusing it into a global feature vector by using the gated attention mechanism, and constructing a global feature sequence in a period of time ; inputting the global feature sequence into the bidirectional gated recurrent unit for encoding to capture the context dependency in the sequence, and taking the hidden state of the last time step as a scheduling plan feature vector ; inputting the scheduling plan feature vector into a fully connected classifier to output the scheduling plan rationality evaluation result.

[0015] Preferably, the calculation formula of the attention coefficient is as follows:

[0016] wherein, is a learnable weight matrix, is an attention weight vector, and the updated feature of node is obtained by weighted summation, is, is the vector of node in the layer graph attention network, is the vector of node in the layer graph attention network.​ for sum index variable, for all neighbor nodes directly connected with the node for all neighbor nodes directly connected with the node.

[0017] In a second aspect, an embodiment of the present application provides a scheduling plan rationality associated factor tracing system, comprising: A scenario module reads power grid primary equipment model data and historical measurement data, obtains scheduling plan data at the same time scale, and then generates power grid model data samples, measurement data samples and prediction data samples after organization processing. The scheduling plan data is used to replace the historical measurement data to obtain power grid continuous operation data sections based on scheduling plans. A scoring module uses the measurement data samples and the prediction data samples to establish a multi-index evaluation system based on the power grid continuous operation data sections obtained by the scenario module, fuses subjective weights and objective weights to obtain comprehensive weights, and calculates the rationality comprehensive score of the to-be-evaluated scheduling plan through ideal solution and negative ideal solution. A screening module takes the rationality comprehensive score obtained by the scoring module as a response variable, constructs a candidate feature set based on the error characteristics of the prediction data samples, removes irrelevant or weakly related variables through a multi-stage screening mechanism, and obtains a high-value feature subset composed of key elements strongly associated with source load prediction errors. A feature module models the power grid as an undirected graph based on the power grid model data samples, provides node historical operation states based on the high-value feature subset obtained by the screening module and the measurement data samples, performs feature extraction and coding operations that fuse global context information and node-level spatiotemporal fine-grained information, constructs a multi-scale feature representation system that fuses global context information and node local states, and generates tracing features representing power grid operation states. An analysis module adopts a combination of graph attention networks and bidirectional gated recurrent units to construct a tracing model, encodes the tracing features representing power grid operation states obtained by the feature module, outputs scheduling plan rationality evaluation results, and realizes associated factor tracing.

[0018] Preferably, in the scenario module, the time interval of the measurement data and the prediction data is unified as 15 minutes; the scheduling plan data includes generation plan, tie line plan and maintenance plan; and the organization processing is performed according to the model data service interface data structure of the new generation of dispatching technical support system.

[0019] Preferably, in the scoring module, the multi-index evaluation system covers safety, economy, environmental protection and dispatchability dimensions, and clearly defines the optimization direction of each index; the subjective weights are calculated by the analytic hierarchy process, the objective weights are calculated by the entropy weight method, and the fusion is weighted fusion. The rationality comprehensive score of the to-be-evaluated dispatch plan is calculated by the ideal solution and the negative ideal solution, and specifically includes: According to the optimal value and the worst value of each index in the historical sample, an ideal solution and a negative ideal solution are constructed, and the weighted Euclidean distance between the index value of the to-be-evaluated dispatch plan and the two reference points is calculated and ; the rationality comprehensive score of the dispatch plan under the 100-point system is calculated by the approximation ideal solution method .

[0020] Preferably, in the screening module, the multi-stage screening mechanism specifically includes: Spearman rank correlation coefficient is used to evaluate the monotonic correlation between each candidate feature and the response variable, and the features with significant correlation are retained to form a preliminary screening feature subset, and the judgment condition of significant correlation is that the p value is less than a preset threshold or the absolute value of the correlation coefficient is higher than a set threshold; the screening basis also includes a preset cumulative importance ratio or an importance threshold; An XGBoost regression model is constructed based on the preliminary screening feature subset, and the importance measure of each feature is calculated by using the model; According to the feature importance ranking output by the XGBoost model, combined with the preset cumulative importance ratio or the importance threshold, the high-value feature subset is screened from the preliminary screening feature subset.

[0021] Preferably, in the feature module, the multi-scale feature representation system integrating global context information and node local state specifically includes: Meteorological features, date and holiday type features, and historical load curve statistical features are extracted from the high-value feature subset, then spliced to form a global feature vector, and then mapped to a low-dimensional global context vector by a multilayer perception ; For each node in the undirected graph, its static features are extracted and encoded into static embedding by a multilayer perception , and its dynamic historical time series features are extracted and encoded into dynamic embedding by a gated recurrent unit ; The static embedding and the dynamic embedding are spliced to form a node-level fusion feature , which is used as the traceability feature.

[0022] Preferably, the power grid is modeled as an undirected graph , denotes a set of nodes, is a set of edges, is an adjacency matrix; The static characteristics include the voltage amplitude, voltage phase angle, active / reactive power output of the generator unit, and active / reactive power output of the load for the previous day; the dynamic characteristics consist of a historical time series of the past 7 days at 15-minute intervals.

[0023] Preferably, in the analysis module, the method of constructing the source tracing model by combining graph attention networks and bidirectional gated recurrent units specifically includes: The global context vector Node-level fusion features of each node The nodes are then merged to form their initial characteristics. The initial features are input into a graph attention network for multi-layer encoding. Information is aggregated by adaptively calculating attention coefficients between nodes to obtain a node feature matrix containing spatial dependency information. The attention coefficients... The calculation formula is:

[0024] in, The weight matrix is ​​a learnable matrix. For attention weight vectors, nodes The updated features are obtained through weighted summation. for, For the first Nodes in a layered graph attention network The vector, For the first Nodes in a layered graph attention network The vector, For summation index variables, For nodes The set of all directly connected neighboring nodes; Based on the node feature matrix, a gated attention mechanism is used to fuse them into a global feature vector, and a global feature sequence over a period of time is constructed. ; The global feature sequence The input is encoded into a bidirectional gated recurrent unit to capture contextual dependencies in the sequence, and the hidden state at the last time step is taken as the scheduling plan feature vector. ; The scheduling plan feature vector The input is fed into a fully connected classifier, which outputs the rationality evaluation result of the scheduling plan.

[0025] Thirdly, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for tracing the rationality of scheduling plan correlation factors.

[0026] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium comprising a computer program which, when executed by a processor, implements the steps of the above-mentioned dispatch plan rationality associated factor tracing method.

[0027] In a fifth aspect, a chip comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned dispatch plan rationality associated factor tracing method when executing the computer program.

[0028] In a sixth aspect, an embodiment of the present application provides an electronic device comprising a computer program which, when executed by the electronic device, implements the steps of the above-mentioned dispatch plan rationality associated factor tracing method.

[0029] Compared with the prior art, the present application has at least the following beneficial effects: A dispatch plan rationality associated factor tracing method, by constructing a dispatch plan-based data section, converts post-event analysis into pre-event evaluation; by multi-index fusion scoring, converts fuzzy experience judgment into objective quantitative indicators; by multi-stage feature screening, accurately locates key influencing elements from massive factors; and finally by a graph neural network model, deeply reveals the influence mechanism of complex space-time correlation of a power grid on plan rationality. The entire scheme is closely linked, realizing a leap from perceiving a problem to diagnosing a root cause.

[0030] Further, by standardizing data specifications, the uniformity and accuracy of scenario construction are ensured. A uniform time interval ensures data time sequence alignment and analysis accuracy; a clear plan type range (power generation, tie line, maintenance) defines the application boundary of the present application, avoiding an excessively wide protection range; and the interface standard of a "new generation of dispatch technology support system" is referenced, ensuring that the scheme can be seamlessly connected with existing industrial systems, greatly improving the practical value and industrialization landing ability of the scheme.

[0031] Further, a multi-index system covers safety, economy, environmental protection, and dispatchability, comprehensively covering core evaluation dimensions of a dispatch plan, avoiding one-sidedness of single-dimensional evaluation; a clear index optimization direction provides a clear basis for scoring; by obtaining subjective weights through an analytic hierarchy process and obtaining objective weights through an entropy weight method and weighted fusion, both professional experience of dispatch experts and internal structure information of data are fully utilized, solving the drawbacks of traditional single weight methods that either rely on experience or are limited by data, greatly improving the scientificity and reliability of rationality scoring.

[0032] Further, the ideal solution and negative ideal solution are constructed as reference benchmarks, the weighted Euclidean distance between the to-be-evaluated plan and the benchmarks is converted into a score of 100 points, and the scoring result is directly quantized. The calculation method takes into account the relative advantages and disadvantages of each index and the proportion of the weight, can accurately distinguish the rationality difference of different scheduling plans, is especially suitable for complex scenes such as new energy high penetration and temporary maintenance, provides accurate and quantifiable response variables for subsequent associated element screening, and solves the problems of traditional scoring methods, such as fuzzification and low discrimination.

[0033] Further, the two-stage screening of Spearman rank correlation coefficient preliminary screening + XGBoost regression model precise screening is adopted, irrelevant or weakly correlated variables are removed first, and then the feature importance is quantified, taking into account the monotonic correlation identification and nonlinear interaction effect capture. The traditional feature screening has the problems of high redundancy and weak generalization ability, and the screened high-value feature subset has clear physical meaning and strong discrimination ability, which reduces the calculation complexity of the subsequent tracing model and improves the tracing accuracy, providing reliable support for accurately locating the key factors of source and load prediction error.

[0034] Further, the quantified judgment standard and the supplementary screening condition are used to make the feature screening process quantifiable and repeatable, avoiding the screening deviation caused by subjective judgment. The design improves the rigor and flexibility of feature screening, and can dynamically adjust the parameters according to different power grid scenes, ensuring that the core associated elements can be screened under different source and load characteristics, further enhancing the reliability of the high-value feature subset and providing stable input for the tracing model.

[0035] Further, the global context information and the node local state are fused, the global features are encoded by MLP, and the dynamic features are encoded by GRU, forming node-level fusion features as tracing features. The design solves the problem that the traditional feature extraction either ignores global macro factors or lacks node fine-grained information, and captures global influencing factors such as weather and date, as well as node static steady-state information and dynamic time evolution law, providing high-quality input rich in spatio-temporal coupling features for the tracing model, laying the feature foundation for accurate tracing.

[0036] Further, the power grid is abstracted as a standard undirected graph, and the power grid topology relationship is accurately represented by node set, edge set and adjacency matrix, making the power grid spatial structure mathematical and standardized. The expression conforms to the processing requirements of graph neural network, provides a clear mathematical basis for GAT model to capture the electrical connection dependence between nodes, solves the problem that traditional methods are difficult to quantify the power grid topology association, ensures the effective extraction of spatial dependence information, and improves the adaptability of the tracing model to the power grid structure.

[0037] Further, the static feature selects the steady-state measurement information of the previous day, and the dynamic feature selects the 15-minute interval time series data of the past 7 days, which not only ensures the stability of the static feature, but also covers the time sequence evolution law of the dynamic feature. The feature definition is in line with the actual power grid operation, the static feature reflects the basic operation state of the node, the dynamic feature captures the short-term fluctuation trend, the combination of the two comprehensively describes the node operation characteristics, solves the problem of single dimension and insufficient timeliness of traditional feature selection, and provides rich and accurate original data for fusion feature generation.

[0038] Further, the global context and node features are first fused through GAT to capture spatial dependency, then the global feature sequence is encoded through Bi-GRU to capture temporal context dependency, and finally the evaluation result is output through a full-connection classifier. The model architecture is suitable for both the spatial characteristics of the power grid topology and the dynamic evolution characteristics of the scheduling plan, solves the limitations of traditional single models that cannot handle non-Euclidean topology data or cannot capture temporal dynamics, greatly improves the accuracy of rationality evaluation and correlation factor tracing, and realizes dual optimization in space and time dimensions.

[0039] Further, by adaptively calculating the attention coefficients between nodes, the GAT model can allocate weights according to the node feature content, realizing asymmetric and content-aware information aggregation. The correlation strength between the node and its neighbors is accurately quantified, avoiding the one-sidedness of traditional graph model equal weight aggregation, highlighting the influence of key neighbor nodes, improving the accuracy of spatial dependency information extraction, providing a high-quality node feature matrix for subsequent time sequence encoding, and further strengthening the core performance of the tracing model.

[0040] It can be understood that the beneficial effects of the above-mentioned second aspect to the sixth aspect can be referred to the related description in the first aspect, which will not be repeated here.

[0041] In summary, the scheduling plan rationality evaluation and tracing method of the present application adopts data-driven and expert experience fusion, and constructs an analysis framework integrating comprehensive scoring, key factor screening, spatio-temporal feature modeling and deep learning tracing, which helps to accurately identify potential problems and influencing factors of the scheduling plan before its execution, and can more scientifically and efficiently guide the optimization of the scheduling plan and the prevention and control of power grid operation risks.

[0042] The technical solutions of the present application will be further described in detail below with the aid of drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 is the flowchart of the present application; Figure 2 is a schematic diagram of a computer device provided by an embodiment of the present application; Figure 3A block diagram of a chip according to an embodiment of the present application.

[0044] In the description of the present application, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present application will be clearly and completely described in connection with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort fall within the protection scope of the present application.

[0046] In the description of the present application, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0047] It should also be understood that the terms used in the specification of the present application are only for the purpose of describing the specific embodiments and are not intended to limit the present application. As used in the specification and the appended claims of the present application, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0048] It should be further understood that the term "and / or" used in the specification and the appended claims of the present application means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations, for example, A and / or B can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " in the present application generally represents an "or" relationship between the front and rear associated objects.

[0049] It should be understood that although the terms first, second, third, etc. may be used in the embodiments of the present application to describe the preset ranges, etc., these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from each other. For example, the first preset range can also be referred to as the second preset range, and similarly, the second preset range can also be referred to as the first preset range without departing from the scope of the embodiments of the present application.

[0050] Depending on the context, the word "if" as used herein can be interpreted to mean "when" or "while" or "in response to determining" or "in response to detecting." Similarly, the phrase "if it is determined" or "if [a stated condition or event] is detected" can be interpreted to mean "when it is determined" or "in response to determining" or "when [a stated condition or event] is detected" or "in response to detecting [a stated condition or event]."

[0051] Various structural diagrams according to the disclosed embodiments of the present application are shown in the drawings. These diagrams are not drawn to scale, in which certain details are exaggerated for clarity of presentation and may omit certain details. The shapes of various regions, layers shown in the drawings and their relative sizes, positional relationships are only exemplary, and in actuality, they can deviate due to manufacturing tolerances or technical limitations, and a person skilled in the art can additionally design regions / layers with different shapes, sizes, relative positions according to actual needs.

[0052] The present application provides a dispatch plan rationality associated factor tracing method, which considers multiple indicators such as power grid safety margin, balance and regulation capacity, economic low carbon, etc., can calculate single indicators, and can also combine dispatch professional expert experience, select important indicators for combination and weighted calculation, realize rationality comprehensive evaluation, for the influencing factors of dispatch plan rationality, through analyzing the influence degree of various types of data inside and outside the power grid on the rationality of the dispatch plan, realizing the quantization and error positioning of the data correlation elements of the power grid dispatch plan, providing a technical support method for power grid dispatch operation personnel, and helping to improve the accuracy of dispatch plan arrangement.

[0053] Referring to Figure 1 The dispatch plan rationality associated factor tracing method of the present application comprises the following steps: S1, dispatch plan rationality tracing scenario construction; Read the primary equipment model data and historical measurement data of a certain power grid in the north, organize and generate power grid model data samples, measurement data samples and prediction data samples according to the model data service interface data structure of the new generation of dispatch technical support system, wherein the time interval of the measurement data and the prediction data is unified to 15 minutes; at the same time, obtain the dispatch plan data at the same time scale, including the generation plan, the tie line plan and the maintenance plan, which are used to replace the historical measurement data with the dispatch plan data in the subsequent steps to construct the power grid continuous operation data section based on the dispatch plan.

[0054] S2, dispatch plan rationality scoring; Based on the constructed dispatch plan data section, the rationality of the dispatch plan is quantitatively scored. The scoring process is as follows: ​S201, a multi-index evaluation system covering safety, economy, environmental protection, and dispatchability is established, and the optimization direction of each index is determined; S202, the analytic hierarchy process and entropy weight method are used to obtain the subjective weight and objective weight of the index respectively, and the comprehensive weight is formed by weighted fusion to give consideration to expert experience and data internal structure information; S203, the ideal solution and negative ideal solution are constructed according to the optimal value and the worst value of each index in the historical sample, and the weighted Euclidean distance between the index value of the to-be-evaluated dispatch plan and the two reference points is calculated and ; S204, the rationality comprehensive score of the dispatch plan under the 100-point system is calculated by the approximation ideal solution method , the overall advantages and disadvantages of the dispatch plan are intuitively and quantitatively evaluated, and reliable scoring basis is provided for subsequent traceability analysis.

[0055] S3, source load prediction error related element screening; In view of the complex nonlinear relationship between source load prediction error and its potential influencing factors, a related element identification method based on a multi-stage screening mechanism is proposed.

[0056] S301, taking the dispatch plan rationality score as the response variable, a candidate feature set including weather conditions, historical load / output characteristics, prediction length, regional type, and equipment state is constructed; S302, the Spearman Correlation Coefficient is used to preliminarily evaluate the monotonic correlation between each candidate feature and the target variable. The coefficient measures the correlation degree by calculating the rank difference of the feature and the response variable in the sample ranking. The features with significant correlation (such as p value less than a preset threshold or correlation coefficient absolute value higher than a set threshold) are reserved, and irrelevant or weakly correlated variables are effectively removed, realizing the first-stage dimension reduction and redundancy suppression; S303, based on the feature subset screened in the first stage, an XGBoost (eXtreme Gradient Boosting) regression model is constructed. The cumulative contribution of each feature to the loss function reduction during the tree structure splitting process is used as the feature importance measure, which quantifies the strength of each feature's explanation ability for the prediction error; S304, according to the feature importance ranking output by XGBoost, combined with the preset cumulative importance proportion or importance threshold, further screening out key related elements with strong discriminant ability, excellent generalization performance and clear physical meaning, forming a high-value feature subset for subsequent traceability modeling.

[0057] The method combines the feature discrimination ability of non-parametric correlation analysis and ensemble learning, considers both monotonic correlation identification and nonlinear interaction effect capture, and significantly improves the accuracy and robustness of source-load prediction error correlation element screening.

[0058] S4, scheduling plan rationality traceability feature selection; To fully depict the spatio-temporal coupling features that affect the rationality of the scheduling plan, the power grid is modeled as an undirected graph , wherein, represents a set containing nodes (such as generator units, load nodes, or substations), is a set of edges, is an adjacency matrix representing the electrical connection topological relationship between nodes. Based on this graph structure, a multi-scale feature representation system is constructed that integrates global context information and node local state.

[0059] S401, for the macro factors affecting the overall operation state of the power grid, a global key factor encoder is designed: the meteorological features (such as temperature, humidity, etc.) , date and holiday type features , and historical load curve statistical features selected in step S3 are spliced to form a global feature vector , and a multi-layer perception (MLP) is used to map it to a low-dimensional global context vector .

[0060] S402, for each node in the graph, its static features and dynamic features are extracted respectively: static features include voltage amplitude, voltage phase angle, unit active / reactive power, load active / reactive, and other steady-state measurement information of the previous day, which are encoded into static embeddings by independent MLPs; S403, dynamic features are composed of 7-day history time series with 15-minute intervals, and a gated recurrent unit (GRU) network is used to model their time evolution rules, outputting dynamic embeddings .

[0061] S404, the static embeddings and dynamic embeddings are spliced by channel to form node-level fusion features , which are used as the core input to represent the operating state of each node in the subsequent traceability model. This feature selection mechanism effectively integrates global macro factors and node-level spatio-temporal fine-grained information, laying a data foundation for accurately identifying the influence path of the scheduling plan rationality.

[0062] S5, scheduling plan rationality traceability model construction.

[0063] To capture the complex spatial dependencies between power grid nodes and analyze the dynamic process of scheduling plans guiding the evolution of system state, a method for constructing a scheduling plan rationality tracing model based on the combination of Graph Attention Network (GAT) and Bidirectional Gated Recurrent Unit (Bi-GRU) is proposed.

[0064] S501. First, for the undirected graph after modeling... It uses GAT as a feature encoder and achieves asymmetric, content-aware information aggregation by adaptively allocating attention weights to neighboring nodes.

[0065] Specifically, at each node In the initial stage, the global context vector Through broadcasting mechanism and node characteristics Fusion, forming initial fusion characteristics In the first In layer GAT, for nodes and his neighbors Calculate the attention coefficient as follows:

[0066] in, The weight matrix is ​​a learnable matrix. For attention weight vectors, nodes The updated features are obtained by weighted summation:

[0067] S502. After multi-layer GAT encoding, a node feature matrix containing rich spatial dependency information is obtained. To further analyze the system's dynamic characteristics, a system state sequence over a period of time needs to be constructed. Therefore, the node features output by GAT are fused into a single global feature vector using a gated attention mechanism. By repeating this process at different time steps, a global feature sequence is constructed. To effectively capture contextual dependencies in global feature sequences, a bidirectional gated recurrent unit (Bi-GRU) is used to process the sequences. Encoding is performed. The Bi-GRU processes the sequence through two processes: forward and backward.

[0068]

[0069] The forward and backward hidden states at each time step are concatenated to obtain the final hidden state. ; Retrieve the hidden state at the last time step As a condensed representation of the dynamic characteristics of the entire dispatch plan, i.e., a dispatch plan feature vector .

[0070] S503, finally, the dispatch plan feature vector is input into the full connection classifier for rationality evaluation, and the calculation formula is as follows:

[0071]

[0072] Wherein, , , , is the classifier parameter, is the output class probability distribution.

[0073] During the model training process, a weighted cross-entropy loss function is used to deal with the class imbalance problem, so as to ensure that the model can accurately identify the rationality of the dispatch plan and support the optimization decision of power grid operation.

[0074] In another embodiment of the present application, a dispatch plan rationality associated factor tracing system is provided, which can be used to realize the above-mentioned dispatch plan rationality associated factor tracing method. Specifically, the dispatch plan rationality associated factor tracing system comprises a scene module, a scoring module, a screening module, a feature module and an analysis module.

[0075] The scene module reads the power grid primary equipment model data and historical measurement data, obtains the dispatch plan data of the same time scale, and then generates power grid model data samples, measurement data samples and prediction data samples after organization processing. The dispatch plan data is used to replace the historical measurement data to obtain the power grid continuous operation data section based on the dispatch plan. The scoring module establishes a multi-index evaluation system based on the power grid continuous operation data section obtained by the scene module, obtains comprehensive weights by fusing subjective weights and objective weights, and calculates the rationality comprehensive score of the dispatch plan to be evaluated through ideal solution and negative ideal solution. The screening module takes the rationality comprehensive score obtained by the scoring module as the response variable, constructs a candidate feature set, removes irrelevant or weakly related variables through a multi-stage screening mechanism, and obtains a high-value feature subset composed of key elements strongly associated with source load prediction error. The feature module models the power grid as an undirected graph, performs feature extraction and encoding operations based on the high-value feature subset obtained by the screening module, constructs a multi-scale feature representation system that fuses global context information and node-level spatiotemporal fine-grained information, and generates tracing features representing the power grid operation state. The analysis module adopts a mode of combining a graph attention network with a bidirectional gated recurrent unit to construct a traceability model, encodes and processes traceability features of the power grid operation state obtained by the feature module, outputs a scheduling plan rationality evaluation result, and realizes traceability of associated factors.

[0076] In the scenario module, the time interval of the measurement data and the prediction data is unified as 15 minutes; the scheduling plan data includes a generation plan, a tie line plan, and a maintenance plan; and the organization processing is performed according to a model data service interface data structure of a new generation of dispatching technical support system.

[0077] In the scoring module, the multi-index evaluation system covers safety, economy, environmental protection, and dispatchability dimensions, and the optimization direction of each index is clear; the subjective weight is calculated by an analytic hierarchy process, the objective weight is calculated by an entropy weight method, and the fusion is weighted fusion. The rationality comprehensive score of the to-be-evaluated scheduling plan calculated by the ideal solution and the negative ideal solution specifically includes: The ideal solution and the negative ideal solution are respectively constructed according to the optimal value and the worst value of each index in the historical sample, and the weighted Euclidean distance between the index value of the to-be-evaluated scheduling plan and the two reference points is calculated and The rationality comprehensive score of the scheduling plan under a 100-point system is calculated by the approximation ideal solution method .

[0078] In the screening module, the multi-stage screening mechanism specifically includes: The Spearman rank correlation coefficient is used to perform monotonic correlation evaluation on each candidate feature and the response variable, and the features with significant correlation are retained to form a preliminary screening feature subset, and the significant correlation is determined according to a condition that a p value is less than a preset threshold or an absolute value of a correlation coefficient is higher than a set threshold; the screening criteria also include a preset cumulative importance ratio or an importance threshold; An XGBoost regression model is constructed based on the preliminary screening feature subset, and the importance measure of each feature is calculated by using the model; According to the feature importance ranking output by the XGBoost model, the high-value feature subset is screened from the preliminary screening feature subset in combination with the preset cumulative importance ratio or the importance threshold.

[0079] In the feature module, the multi-scale feature representation system fusing global context information and node local state is constructed, specifically including: The meteorological features, the date and holiday type features, and the historical load curve statistical features screened by the screening module are spliced to form a global feature vector, and the global feature vector is mapped to a low-dimensional global context vector by a multilayer perceptron . For each node in the undirected graph, extract its static features and encode them into static embedding through multi-layer perception , and extract its dynamic historical time series features and encode them into dynamic embedding through gated recurrent unit ; Splice the static embedding and the dynamic embedding to form node-level fusion features as the traceability features.

[0080] Model the power grid as an undirected graph , denote a set of nodes, a set of edges, an adjacency matrix. The static features include voltage amplitude, voltage phase angle, unit active / reactive power output, load active / reactive power of the previous day; the dynamic features consist of 7-day historical time series with 15-minute intervals.

[0081] In the analysis module, the traceability model is constructed by combining graph attention network with bidirectional gated recurrent unit, specifically including: Fuse the global context vector with the node-level fusion features of each node to form the initial features of each node; Input the initial features into the graph attention network for multi-layer encoding, aggregate information by adaptively calculating the attention coefficients between nodes, obtain the node feature matrix containing spatial dependency information, and the calculation formula of the attention coefficient is as follows:

[0082] wherein, is a learnable weight matrix, is an attention weight vector, and the updated feature of node is obtained by weighted summation, is is the vector of node in the layer graph attention network, is the vector of node in the layer graph attention network, is a summation index variable, is the set of all neighbor nodes directly connected to node ; Based on the node feature matrix, a global feature vector is fused by using a gated attention mechanism, and a global feature sequence in a period of time is constructed ; The global feature sequence is input into a bidirectional gated recurrent unit for encoding to capture context dependence in the sequence, and a hidden state at the last time step is taken as a scheduling plan feature vector ; The scheduling plan feature vector is input into a fully connected classifier, and a scheduling plan rationality evaluation result is output.

[0083] The application provides a terminal device, which comprises a processor and a memory, the memory is used for storing a computer program, the computer program comprises program instructions, and the processor is used for executing the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components and the like, which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement a corresponding method flow or a corresponding function. The processor in the embodiments of the application can be used for the operation of the scheduling plan rationality correlation factor tracing method, which comprises the following steps: The power grid primary equipment model data and historical measurement data are read, and scheduling plan data of the same time scale is obtained. Then, after organization processing, power grid model data samples, measurement data samples and prediction data samples are generated. The scheduling plan data is used to replace the historical measurement data to obtain power grid continuous operation data sections based on scheduling plans. Based on the obtained power grid continuous operation data sections, a multi-index evaluation system is established, a comprehensive weight is obtained by fusing subjective weight and objective weight, and a rationality comprehensive score of the to-be-evaluated scheduling plan is calculated through ideal solution and negative ideal solution. The obtained rationality comprehensive score is taken as a response variable, a candidate feature set is constructed, irrelevant or weakly related variables are removed through a multi-stage screening mechanism, and a high-value feature subset composed of key elements with strong correlation with source load prediction error is obtained. The power grid is modeled as an undirected graph, based on the obtained high-value feature subset, feature extraction and coding operations of fusing global context information and node-level spatio-temporal fine-grained information are performed, a multi-scale feature representation system fusing global context information and node local state is constructed, and traceability features representing the power grid operation state are generated. A traceability model is constructed by combining a graph attention network and a bidirectional gated recurrent unit, the obtained traceability features representing the power grid operation state are coded, and scheduling plan rationality evaluation results are output, realizing correlation factor traceability.

[0084] Please refer to Figure 2 , the terminal device is a computer device, the computer device 60 of the embodiment includes: a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61, and the computer program 63 realizes the scheduling plan rationality correlation factor traceability method in the embodiment when executed by the processor 61. To avoid repetition, this will not be repeated here. Alternatively, the computer program 63 realizes the functions of each model / unit in the scheduling plan rationality correlation factor traceability system of the embodiment when executed by the processor 61. To avoid repetition, this will not be repeated here.

[0085] The computer device 60 can be a desktop computer, a notebook computer, a palm computer, and a cloud server, etc. The computer device 60 can include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art can understand that Figure 2 is only an example of the computer device 60 and does not constitute a limitation on the computer device 60, and can include more or fewer components than shown, or combine certain components, or different components, for example, the computer device can also include an input / output device, a network access device, a bus, etc.

[0086] The processor 61 can be a central processing unit (CPU), and can also be other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0087] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or a memory of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device 60.

[0088] Further, the memory 62 can include both an internal storage unit and an external storage device of the computer device 60. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0089] Please refer to Figure 3 , the terminal device is an electronic device 600, and the electronic device 600 is in the form of a general-purpose computing device. The components of the electronic device can include but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0090] Among them, the storage unit stores program codes, and the program codes can be executed by the processing unit 610, so that the processing unit 610 executes the steps of various exemplary embodiments according to the present application described in the method part of the present specification. For example, the processing unit 610 can execute the steps as shown in Figure 1 .

[0091] Storage 620 can include a readable medium in the form of volatile memory, such as random access memory (RAM) 6201 and / or cache memory 6202, and can further include non-volatile memory, such as read only memory (ROM) 6203.

[0092] Storage 620 can also include program / utility 6204 having a set of programs / modules 6205, including operating system, one or more application programs, other program modules, and program data, each of which or a combination thereof, can include implementation of a network environment, as in each of the examples or some combination thereof.

[0093] Bus 630 can represent one or more of several types of bus structures, including a storage bus or bus controller, peripheral bus, graphics bus, processor or local bus using any of a variety of bus architectures.

[0094] Electronic device 600 can also communicate with one or more external devices 700 such as a keyboard or pointing device, using one or more communication interfaces 650. Communication interfaces 650 can also enable communication with one or more devices that enable a user to interact with electronic device 600, such as a

[0095] Example 4 The present application further provides a storage medium, specifically a computer readable storage medium, which is a memory device in the terminal device, and is used to store programs and data. It can be understood that the computer readable storage medium herein can include the built-in storage medium in the terminal device, and of course can include the expansion storage medium supported by the terminal device, and can be any tangible medium containing or storing programs, which can be used by or in combination with an instruction execution system, device or apparatus. The computer readable storage medium provides a storage space, which stores the operating system of the terminal. Moreover, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more computer programs (including program codes). It should be noted that more specific examples of the computer readable storage medium include an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0096] The computer readable storage medium further includes a data signal carried in baseband or propagated as a carrier wave, in which readable program codes are borne. Such a propagated data signal can take various forms, including but not limited to electro-magnetic signal, optical signal or any suitable combination of the above. The readable storage medium can also be any readable medium other than the readable storage medium, which can send, propagate or transmit programs for use by or in combination with an instruction execution system, device or apparatus. The program codes contained in the readable storage medium can be transmitted by any suitable medium, including but not limited to wireless, wired, optical cable, radio frequency, etc., or any suitable combination of the above.

[0097] The program codes for executing the operation of the present application can be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and conventional procedural programming languages such as "C" language or similar programming languages. The program codes can be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on the remote computing device, or entirely on the remote computing device or server. In the case involving remote computing device, the remote computing device can be connected to the user computing device through any kind of network, including local area network or wide area network, or can be connected to external computing device (for example, connected through the Internet by using an Internet service provider).

[0098] The one or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the method for tracing the associated factors of the dispatch plan rationality in the above embodiments; the one or more instructions stored in the computer-readable storage medium are loaded and executed by the processor to implement the following steps: The power grid primary equipment model data and historical measurement data are read to obtain dispatch plan data at the same time scale. After organization processing, power grid model data samples, measurement data samples and prediction data samples are generated. The dispatch plan data is used to replace the historical measurement data to obtain power grid continuous operation data sections based on the dispatch plan. Based on the obtained power grid continuous operation data sections, a multi-index evaluation system is established, subjective weights and objective weights are fused to obtain comprehensive weights, and the rationality comprehensive score of the to-be-evaluated dispatch plan is calculated through ideal solution and negative ideal solution. The obtained rationality comprehensive score is used as a response variable to construct a candidate feature set. Through a multi-stage screening mechanism, irrelevant or weakly related variables are removed to obtain a high-value feature subset composed of key elements strongly associated with source load prediction error. The power grid is modeled as an undirected graph. Based on the obtained high-value feature subset, feature extraction and coding operations are performed to fuse global context information and node-level spatio-temporal fine-grained information, a multi-scale feature representation system that fuses global context information and node local state is constructed, and traceability features representing the power grid operation state are generated. A traceability model is constructed by combining a graph attention network and a bidirectional gated recurrent unit to encode and process the obtained traceability features representing the power grid operation state, and output the dispatch plan rationality evaluation result, thereby realizing the associated factor tracing.

[0099] The database involved in each embodiment provided by the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, and the like, without being limited thereto. The processor involved in each embodiment provided by the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, and the like, without being limited thereto.

[0100] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0101] The operation data of a provincial power grid in the north from March 1, 2024 to March 31, 2024 are selected, containing 100 groups of dispatching plan samples (covering working days, holidays, high penetration rate days of new energy, temporary maintenance days and the like), the effect of the present application is verified, and the traditional method (single weight score + traditional machine learning tracing) is compared, and the results are as follows:

[0102] The experimental data show that the present application greatly improves the efficiency through multi-stage feature screening, and the fusion model significantly improves the tracing accuracy, and can accurately identify potential risks in advance. After application, the problem of wind and light curtailment of the power grid is effectively alleviated, the frequency of dispatching plan adjustment is reduced by 35%, the economic efficiency of power grid operation is improved by 15 million yuan / month, and the significant advantages of the present application in practical application are fully verified.

[0103] In summary, the dispatching plan rationality associated factor tracing method and system of the present application, aiming at the problem of intensified uncertainty of source and load of new power system, realizes accurate quantification of dispatching plan rationality and efficient tracing of associated factors through scene construction, multi-dimensional scoring, multi-stage feature screening, spatio-temporal fusion feature extraction and GAT+Bi-GRU tracing model. Compared with the traditional method, the scoring result is more distinguishable and scientific, the feature screening efficiency is improved by more than 30%, the tracing accuracy is more than 90%, and the potential risks of dispatching plan can be identified 24 hours in advance. The present application effectively solves the problems of wind and light curtailment, insufficient peak shaving and the like, reduces the operation risk of power grid, improves the accuracy of dispatching plan arrangement and the safe and economic operation level of power grid, and provides reliable technical support for dispatching optimization of new power system.

[0104] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be realized in the form of hardware or software. In addition, the specific name of each functional unit and module is only for the convenience of mutual distinction, and does not limit the protection scope of the present application. The specific working process of the unit and module in the above system can be referred to the corresponding process in the foregoing method embodiment, which will not be described here.

[0105] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can be referred to the related description of other embodiments.

[0106] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present application can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0107] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / terminal and method can be implemented by other ways. For example, the above-mentioned apparatus / terminal embodiments are only schematic, and the division of the modules or units is only a logical function division, and there can be another division way in actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0108] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0109] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0110] The integrated module / unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer-readable storage medium. The computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms. The computer-readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the computer-readable medium can include or exclude contents according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0111] The present application is described with reference to flowcharts and / or block diagrams of methods, devices, and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks

[0112] These computer program instructions can also be stored in a computer-readable storage medium that can guide the computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable storage medium produce a manufactured product including instruction devices that implement the functions specified in the flowcharts and / or block diagrams.Figure 1 one or more processes and / or functions described in one or more blocks. Figure 1 one or more blocks.

[0113] These computer program instructions can also be loaded into computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable data processing devices to generate a computer-implemented process, so that the instructions executed on the computer or other programmable data processing devices provide steps for implementing the function described in one or more processes and / or functions described in one or more blocks. Figure 1 one or more processes and / or functions described in one or more blocks. Figure 1 one or more blocks.

[0114] The above is only to illustrate the technical idea of the present application, and cannot limit the protection scope of the present application. Any modification made according to the technical idea of the present application on the basis of the technical scheme falls within the protection scope of the claims of the present application.

Claims

1. A method for tracing the sources of factors related to the rationality of a scheduling plan, characterized in that, include: Read the primary equipment model data and historical measurement data of the power grid to obtain the scheduling plan data at the same time scale; After processing, power grid model data samples, measurement data samples, and prediction data samples are generated. By replacing the historical measurement data with the scheduling plan data, a continuous power grid operation data profile based on the scheduling plan is obtained; Based on the continuous operation data section of the power grid, a multi-index evaluation system is established using the measured data sample and the predicted data sample. Subjective weights and objective weights are integrated to obtain a comprehensive weight. The comprehensive score of the rationality of the dispatch plan to be evaluated is calculated through the ideal solution and the negative ideal solution. Using the comprehensive rationality score as the response variable, a candidate feature set is constructed based on the error characteristics of the predicted data sample. Irrelevant or weakly correlated variables are eliminated through a multi-stage screening mechanism to obtain a high-value feature subset consisting of key elements strongly correlated with the source load prediction error. Based on the power grid model data samples, the power grid is modeled as an undirected graph. Based on the high-value feature subset and combined with the node historical operating status provided by the measurement data samples, feature extraction and encoding operations that integrate global context information and node-level spatiotemporal fine-grained information are performed to construct a multi-scale feature representation system that integrates global context information and node local states, and generate traceability features that characterize the power grid operating status. A source tracing model is constructed by combining graph attention network and bidirectional gated recurrent unit. The source tracing features are encoded and processed, and the rationality evaluation result of the scheduling plan is output to realize the source tracing of related factors.

2. The method for tracing the causes of factors related to the rationality of scheduling plans according to claim 1, characterized in that, The time interval between the measured data and the predicted data is uniformly set to 15 minutes; the scheduling plan data includes power generation plan, tie line plan and maintenance plan; the organization and processing are performed according to the model data service interface data structure of the new generation scheduling technology support system.

3. The method for tracing the sources of factors related to the rationality of scheduling plans according to claim 1, characterized in that, The multi-indicator evaluation system covers dimensions of safety, economy, environmental protection and dispatchability, and clarifies the optimization direction of each indicator; the subjective weights are calculated by the analytic hierarchy process, the objective weights are calculated by the entropy weight method, and the fusion is a weighted fusion.

4. The method for tracing the sources of factors related to the rationality of scheduling plans according to claim 1, characterized in that, The comprehensive score for the rationality of the scheduling plan under evaluation is calculated by using both the ideal solution and the negative ideal solution. This score includes: Based on the best and worst values ​​of each indicator in the historical samples, ideal and negative ideal solutions are constructed respectively. The weighted Euclidean distance between the values ​​of each indicator of the scheduling plan to be evaluated and these two reference points is calculated. and The approximation method for ideal solutions is used to calculate the overall rationality score of the scheduling plan on a 100-point scale. .

5. The method for tracing the sources of factors related to the rationality of scheduling plans according to claim 1, characterized in that, The multi-stage screening mechanism specifically includes: The Spearman rank correlation coefficient was used to evaluate the monotonic correlation between each candidate feature and the response variable, and features with significant correlation were retained to form a preliminary feature subset. An XGBoost regression model is constructed based on the initial screening feature subset, and the importance measure of each feature is calculated using the model. Based on the feature importance ranking output by the XGBoost model, and combined with a preset cumulative importance ratio or importance threshold, the high-value feature subset is selected from the initial feature subset.

6. The method for tracing the sources of factors related to the rationality of scheduling plans according to claim 5, characterized in that, Among the features that retain significant correlation, the criteria for determining significant correlation are that the p-value is less than a preset threshold or the absolute value of the correlation coefficient is higher than a set threshold; the screening criteria also include preset cumulative importance ratios or importance thresholds.

7. The method for tracing the sources of factors related to the rationality of scheduling plans according to claim 1, characterized in that, The construction of a multi-scale feature representation system that integrates global context information and node local states specifically includes: Meteorological features, date and holiday type features, and historical load curve statistical features are extracted from the high-value feature subset. These features are then concatenated to form a global feature vector, which is then mapped into a low-dimensional global context vector using a multilayer perceptron. ; For each node in the undirected graph, its static features are extracted and encoded into static embeddings using a multilayer perceptron. It extracts dynamic historical time series features and encodes them into dynamic embeddings using gated recurrent units. ; The static embedding With the dynamic embedding By splicing the data, node-level fusion features are formed. , as the traceability feature.

8. The method for tracing the sources of factors related to the rationality of scheduling plans according to claim 7, characterized in that, Model the power grid as an undirected graph. , Indicates inclusion A set of nodes, Let be the set of edges. It is an adjacency matrix.

9. The method for tracing the sources of factors related to the rationality of scheduling plans according to claim 7, characterized in that, The static characteristics include the voltage amplitude, voltage phase angle, active / reactive power output of the generator unit, and active / reactive power output of the load for the previous day; the dynamic characteristics consist of a historical time series of the past 7 days at 15-minute intervals.

10. The method for tracing the sources of factors related to the rationality of scheduling plans according to claim 1, characterized in that, The source tracing model is constructed by combining graph attention networks with bidirectional gated recurrent units, specifically including: The global context vector Node-level fusion features with each node The nodes are then merged to form their initial characteristics. The initial features are input into a graph attention network for multi-layer encoding. Information is aggregated by adaptively calculating the attention coefficients between nodes to obtain a node feature matrix containing spatial dependency information. Based on the node feature matrix, a gated attention mechanism is used to fuse them into a global feature vector, and a global feature sequence over a period of time is constructed. ; The global feature sequence The input is encoded into a bidirectional gated recurrent unit to capture contextual dependencies in the sequence, and the hidden state at the last time step is taken as the scheduling plan feature vector. ; The scheduling plan feature vector The input is fed into a fully connected classifier, which outputs the rationality evaluation result of the scheduling plan.

11. The method for tracing the sources of factors related to the rationality of scheduling plans according to claim 10, characterized in that, The attention coefficient The calculation formula is: in, The weight matrix is ​​a learnable matrix. For attention weight vectors, nodes The updated features are obtained through weighted summation. for, For the first Nodes in a layered graph attention network The vector, For the first Nodes in a layered graph attention network The vector, For summation index variables, For nodes The set of all directly connected neighboring nodes.

12. A system for tracing the sources of factors related to the rationality of a scheduling plan, characterized in that, include: The scenario module reads primary equipment model data and historical measurement data from the power grid to obtain scheduling plan data at the same time scale; then, after processing, it generates power grid model data samples, measurement data samples, and prediction data samples. By replacing the historical measurement data with the scheduling plan data, a continuous power grid operation data profile based on the scheduling plan is obtained; The scoring module, based on the continuous operation data profile of the power grid obtained by the scenario module, establishes a multi-index evaluation system using the measured data sample and the predicted data sample, integrates subjective weights and objective weights to obtain a comprehensive weight, and calculates the comprehensive score of the rationality of the dispatch plan to be evaluated through the ideal solution and the negative ideal solution. The screening module uses the comprehensive rationality score obtained by the scoring module as the response variable, constructs a candidate feature set based on the error characteristics of the predicted data sample, and eliminates irrelevant or weakly correlated variables through a multi-stage screening mechanism to obtain a high-value feature subset composed of key elements strongly correlated with the source load prediction error. The feature module models the power grid as an undirected graph based on the power grid model data samples. Based on the high-value feature subset obtained by the screening module and the historical operating status of the nodes provided by the measurement data samples, it performs feature extraction and encoding operations that integrate global context information and node-level spatiotemporal fine-grained information, constructs a multi-scale feature representation system that integrates global context information and node local states, and generates traceability features that characterize the operating status of the power grid. The analysis module uses a combination of graph attention network and bidirectional gated cyclic unit to construct a source tracing model. It encodes the source tracing features that represent the power grid operation status obtained by the feature module, outputs the rationality evaluation results of the scheduling plan, and realizes the source tracing of related factors.

13. The scheduling plan rationality correlation factor tracing system according to claim 12, characterized in that, In the scenario module, the time interval between the measured data and the predicted data is uniformly set to 15 minutes; the scheduling plan data includes power generation plan, tie line plan, and maintenance plan; the organization processing is executed based on the model data service interface data structure of the new generation scheduling technology support system.

14. The scheduling plan rationality correlation factor tracing system according to claim 12, characterized in that, In the scoring module, the multi-indicator evaluation system covers the dimensions of safety, economy, environmental protection, and schedulability, and clarifies the optimization direction of each indicator; the subjective weights are calculated using the analytic hierarchy process, the objective weights are calculated using the entropy weight method, and the fusion is a weighted fusion. The comprehensive score for the rationality of the scheduling plan under evaluation is calculated by using both the ideal solution and the negative ideal solution. This score includes: Based on the best and worst values ​​of each indicator in the historical samples, ideal and negative ideal solutions are constructed respectively. The weighted Euclidean distance between the values ​​of each indicator of the scheduling plan to be evaluated and these two reference points is calculated. and ; The rationality score of the scheduling plan on a 100-point scale is calculated using the approximation ideal solution method. .

15. The scheduling plan rationality correlation factor tracing system according to claim 12, characterized in that, In the filtering module, the multi-stage filtering mechanism specifically includes: Spearman's rank correlation coefficient is used to evaluate the monotonic correlation between each candidate feature and the response variable. Features with significant correlation are retained to form a preliminary feature subset. The criteria for significant correlation are that the p-value is less than a preset threshold or the absolute value of the correlation coefficient is higher than a set threshold. The screening criteria also include preset cumulative importance ratio or importance threshold. An XGBoost regression model is constructed based on the initial screening feature subset, and the importance measure of each feature is calculated using the model. Based on the feature importance ranking output by the XGBoost model, and combined with a preset cumulative importance ratio or importance threshold, the high-value feature subset is selected from the initial feature subset.

16. The scheduling plan rationality correlation factor tracing system according to claim 12, characterized in that, In the feature module, the construction of a multi-scale feature representation system that integrates global context information and node local states specifically includes: Meteorological features, date and holiday type features, and historical load curve statistical features are extracted from the high-value feature subset. These features are then concatenated to form a global feature vector, which is then mapped into a low-dimensional global context vector using a multilayer perceptron. ; For each node in the undirected graph, its static features are extracted and encoded into static embeddings using a multilayer perceptron. It extracts dynamic historical time series features and encodes them into dynamic embeddings using gated recurrent units. ; The static embedding With the dynamic embedding By splicing the data, node-level fusion features are formed. , as the traceability feature.

17. The scheduling plan rationality correlation factor tracing system according to claim 16, characterized in that, Model the power grid as an undirected graph. , Indicates inclusion A set of nodes, Let be the set of edges. It is an adjacency matrix; The static characteristics include the voltage amplitude, voltage phase angle, active / reactive power output of the generator unit, and active / reactive power output of the load for the previous day; the dynamic characteristics consist of a historical time series of the past 7 days at 15-minute intervals.

18. The scheduling plan rationality correlation factor tracing system according to claim 12, characterized in that, In the analysis module, the source tracing model is constructed by combining graph attention networks with bidirectional gated recurrent units, specifically including: The global context vector Node-level fusion features with each node The nodes are then merged to form their initial characteristics. The initial features are input into a graph attention network for multi-layer encoding. Information is aggregated by adaptively calculating attention coefficients between nodes to obtain a node feature matrix containing spatial dependency information. The attention coefficients... The calculation formula is: in, The weight matrix is ​​a learnable matrix. For attention weight vectors, nodes The updated features are obtained through weighted summation. for, For the first Nodes in a layered graph attention network The vector, For the first Nodes in a layered graph attention network The vector, For summation index variables, For nodes The set of all directly connected neighboring nodes; Based on the node feature matrix, a gated attention mechanism is used to fuse them into a global feature vector, and a global feature sequence over a period of time is constructed. ; The global feature sequence The input is encoded into a bidirectional gated recurrent unit to capture contextual dependencies in the sequence, and the hidden state at the last time step is taken as the scheduling plan feature vector. ; The scheduling plan feature vector The input is fed into a fully connected classifier, which outputs the rationality evaluation result of the scheduling plan.

19. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform the method of any one of claims 1 to 11.

20. A computing device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including steps for performing the method of any one of claims 1 to 11.

Images