Power system current fluctuation prediction method based on multivariate data and feature selection

Abstract

The invention relates to the technical field of power system data modeling and intelligent prediction, and discloses a power system current fluctuation prediction method based on multivariate data and feature selection, and the method comprises the following steps: S1, data disturbance enhancement processing: collecting current, voltage, load and meteorological multi-source heterogeneous power system original data, constructing a feature matrix, and carrying out data disturbance enhancement processing; introducing a disturbance term to the data to form a disturbance sample matrix for enhancing the robustness of the model to data missing and noise interference; and S2, disturbance sample repair and reconstruction: inputting the disturbance sample matrix into a self-encoding structure, and obtaining a repaired feature matrix for subsequent modeling through minimizing a reconstruction error and encoding sparsity constraint. By introducing disturbance enhancement, self-coding reconstruction, graph structure embedding and feature scoring mechanisms, an end-to-end current fluctuation prediction framework is constructed, and the adaptability of the model to data exception and system structure information is effectively enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of power system data modeling and intelligent prediction, and specifically to a power system current fluctuation prediction method based on multi-source data and feature selection. Background Art

[0002] With the continuous expansion of the scale of the power system and the rapid increase in the proportion of new energy access, the monitoring and prediction of current fluctuations have become an important link to ensure the stable operation of the power grid and assist dispatching decisions. Especially in the context of high proportion of intermittent power sources access, the short-term fluctuations of current present more complex non-linear and random characteristics, posing higher requirements for the stability and response ability of prediction methods.

[0003] Traditional current fluctuation prediction methods mainly rely on time series modeling of historical data or regression analysis of single statistical features. Such methods usually assume that the input data is complete and stable, lacking the systematic modeling ability for measurement errors, data missing and multi-source disturbances in actual operation. In real applications, the monitoring data collected by the power system often has inevitable noise interference, packet loss phenomenon and multi-source feature heterogeneity, which will seriously affect the reliability and generalization ability of the prediction model.

[0004] On the other hand, existing methods also have certain limitations in dealing with the structural relationships between multi-dimensional features. Due to the failure to fully consider the complex dependence relationships formed by the grid topology structure, physical coupling relationships or operation logics between features, the prediction model often lacks structural awareness ability and is difficult to effectively capture the potential spatial correlations and information propagation paths between high-dimensional features, resulting in expression deviation of the model when dealing with multi-variable inputs.

[0005] In addition, in high-dimensional input scenarios, the feature selection mechanism usually relies on static statistical evaluation or built-in weight analysis of the model, lacking a systematic feature evaluation method for disturbance response. This leads to the model may contain a large number of redundant or low-value features, further increasing the computational burden of the model and reducing its stability and interpretability in different operation scenarios. Summary of the Invention

[0006] Aiming at the deficiencies of the existing technology, the present invention provides a power system current fluctuation prediction method based on multi-source data and feature selection, which solves the problem of realizing efficient, stable and structurally aware prediction of current fluctuation trends under the conditions of disturbances, incompleteness and high-dimensional redundant features in multi-source data of the power system in the existing technology.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: A power system current fluctuation prediction method based on multi-source data and feature selection, including the following steps: S1. Data perturbation enhancement processing: Collect the original power system data of multi-source heterogeneity including current, voltage, load, meteorology, etc., construct a feature matrix, introduce a perturbation term to the data to form a perturbed sample matrix, which is used to enhance the robustness of the model to data missing and noise interference; S2. Perturbed sample repair and reconstruction: Input the perturbed sample matrix into an auto-encoding structure, and obtain the repaired feature matrix by minimizing the reconstruction error and the coding sparsity constraint for subsequent modeling; S3. Graph structure embedding modeling: Based on the power grid nodes and their connection relationships, construct an adjacency matrix, perform graph convolution processing on the repaired feature matrix to generate an embedded feature representation, so as to introduce the ability of the power grid topology structure to perceive the propagation of current fluctuations; S4. Feature importance scoring and screening: Use a collaborative filtering mechanism to construct a feature-sample relationship matrix based on the prediction error, extract the feature scores through non-negative matrix factorization, and select a feature subset according to a set threshold. This subset will be used as the input features for the final prediction model training; S5. Multi-objective loss modeling and prediction training: Based on the feature subset, construct a current fluctuation prediction model, and optimize the multi-objective loss function including the prediction error term, the perturbation repair stability term, and the graph structure consistency term during the training process to improve the accuracy of the model and the system coordination; S6. Module feedback linkage adjustment: During the training process, feedback the feature scoring results to the perturbation reconstruction module to focus on strengthening the repair accuracy of highly sensitive features. At the same time, use the structure indicators in the graph embedding to optimize the feature screening weights to achieve the information closed-loop adjustment between functional modules; S7. Prediction output: Use the trained prediction model to perform current fluctuation prediction on new input data and output the prediction results.

[0008] Preferably, the perturbation enhancement processing in step S1 is: adding a noise term that follows a Gaussian distribution to the original feature matrix to construct a perturbed sample ; wherein, is the original feature matrix, is the input sample matrix, is a noise matrix that follows a normal distribution with a mean of 0 and a covariance of and is the identity matrix.

[0009] Preferably, the reconstruction loss function of the auto-encoding structure in step S2 is defined as: ; wherein, is the input perturbed sample matrix, is the encoder function, is a decoder function, is the weight factor of the sparse regularization term.

[0010] Preferably, in step S3, the graph structure embedding adopts a normalized adjacency matrix ; wherein, is the original adjacency matrix, is the degree matrix, which is a diagonal matrix; and calculate the graph embedding feature representation ; wherein, is the repaired feature matrix, is the trainable weight matrix, is the activation function.

[0011] Preferably, the element of the feature-sample error matrix constructed in step S4 is defined as: ; wherein, is the true value of the th sample, is the prediction model after removing the th feature, is the corresponding input.

[0012] Preferably, in step S4, the non-negative matrix factorization M≈UV is performed on the feature-sample relationship matrix, and the feature score is the L2 norm of each row of the matrix U and the feature subset Sf={j|sj≥τ} is obtained according to the set threshold τ.

[0013] Preferably, the feedback mechanism in step S6 includes feeding back the high-scoring features identified in step S4 to the auto-encoding structure in step S2 and performing weighted processing on its reconstruction process.

[0014] Preferably, the feedback mechanism in step S6 further includes using the node centrality index obtained in the graph embedding to participate in the weighted operation of the feature score.

[0015] Preferably, the parameter optimization process of the prediction model in step S5 adopts a Bayesian optimization strategy to minimize the multi-objective loss function.

[0016] The present invention provides a power system current fluctuation prediction method based on multivariate data and feature selection. It has the following beneficial effects: 1. The present invention simulates the data uncertainty in the operation of the power system by introducing a perturbation enhancement mechanism, and combines an autoencoder to jointly reconstruct the missing and noisy data, effectively improving the adaptability of the model to input anomalies, providing a more complete and credible feature representation for predictive modeling, and having significant engineering applicability in the current fluctuation prediction scenario.

[0017] 2. The present invention combines the implicit dependence relationship between the physical structure or characteristics of the power grid, constructs a graph structure embedding module, and introduces graph convolution to perform structure-aware re-representation on multi-dimensional features, thereby effectively modeling the spatial topology and logical connection between features, providing an input with both semantic and structural consistency for downstream tasks, and helping to improve the model's ability to depict system-level correlation patterns.

[0018] 3. The present invention quantitatively analyzes the impact of perturbation samples on the prediction performance, and constructs a global feature importance evaluation mechanism through non-negative matrix factorization, solving the problems of lack of robustness and interpretability in traditional feature selection methods, and ensuring the effectiveness of the selected features in actual prediction tasks.

[0019] 4. The present invention designs a multi-objective loss function including prediction error, perturbation stability and graph structure consistency, and establishes a feedback linkage mechanism, realizing the system optimization from feature construction to model training, making the parameters of each sub-module depend on each other and dynamically adjust, constructing a unified and highly collaborative end-to-end prediction process, and improving the comprehensive adaptability of the method under complex working conditions of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a method step diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0022] Please refer to the attached Figure 1 , the embodiment of the present invention provides a power system current fluctuation prediction method based on multi-source data and feature selection, including the following steps: S1. Data perturbation enhancement processing: Collect the original data of the power system including multi-source heterogeneous data such as current, voltage, load and meteorology, construct a feature matrix, introduce a perturbation term into the data to form a perturbation sample matrix, and use it to enhance the robustness of the model to data missing and noise interference; In this embodiment, aiming at the practical engineering problems faced in the current fluctuation prediction of the power system, such as unstable quality of original data, data loss, and noise pollution, a data processing strategy based on perturbation enhancement is proposed. As the first step of the overall method of the present invention, it lays a reliable data foundation for subsequent feature reconstruction, structure modeling, and prediction training.

[0023] In an actually operating power system, the collected multi-source data often comes from various sensors and measurement systems, such as smart meters, substation monitoring equipment, environmental monitoring modules, etc. These data usually include but are not limited to observed values with time series and multi-dimensional attributes such as current, voltage, active power, reactive power, load, frequency, temperature, humidity, wind speed, etc. Due to factors such as sensing accuracy, electromagnetic interference, and communication link failures, these original data often have varying degrees of missing, inconsistent, or noise perturbation phenomena.

[0024] To improve the generalization ability and anti-noise performance of the subsequent model, in this embodiment, a perturbation enhancement mechanism is preferably introduced to perform perturbation sampling on the original feature matrix to construct robust input samples and simulate the data perturbation behavior during the operation of the power system.

[0025] Let the collected original feature data set be a real number matrix , where represents the number of samples, represents the feature dimension. Apply a perturbation function based on a probability distribution to this original matrix to construct a perturbation sample matrix . The form of this perturbation function is preferably a Gaussian distribution, that is: ; where, is the embedded feature matrix output by the graph convolution, represents a noise matrix that follows a normal distribution with zero mean and covariance , is an identity matrix with dimension , is an adjustable noise intensity parameter, which is usually tuned as a model hyperparameter during training.

[0026] The design purpose of this perturbation operation is not only to simulate the possible perturbation patterns of the input data during actual acquisition and transmission, but more importantly, to guide the subsequent model to learn the implicit redundant structure and dependency relationship between the input features through the random perturbation mechanism, thereby enhancing the model's adaptability to incomplete data and offset data.

[0027] It should be noted that the perturbation process should not be simply understood as a data augmentation operation. It plays an important role as a "detector" in the overall framework of the present invention, stimulating the structure recovery and feature selection processes. In the subsequent reconstruction stage, the model needs to perform adaptive reconstruction based on the perturbed samples to identify the internal structure among the data. Meanwhile, in the feature selection stage, this perturbation also provides a premise for evaluating the sensitivity of features to prediction errors.

[0028] In addition, to ensure that the perturbed samples still retain the statistical distribution characteristics of the original data and can be used for training and optimization, in the embodiment, the perturbed samples are further normalized or standardized to keep them consistent with the original data distribution.

[0029] Preferably, this step can also be combined with the time sliding window strategy to batch process multi-time point data to construct a data perturbation cluster within the time series segment, further enhancing the model's ability to identify local time perturbations.

[0030] The perturbed sample matrix will be input as an input into the autoencoder structure in the next step, participating in the subsequent feature reconstruction and missing value repair processes, and constituting the first important link in the data processing chain.

[0031] In summary, the data perturbation enhancement processing in this embodiment injects noise into the original feature matrix by simulating actual system noise, missing values, and abnormal fluctuations, thereby realizing the basic support for the model's robustness and generalization ability. It has a clear engineering background and theoretical feasibility, and can effectively support the subsequent deep reconstruction, structure modeling, and feature evaluation processes.

[0032] S2. Repair and reconstruction of perturbed samples: Input the perturbed sample matrix into the autoencoder structure, and obtain the repaired feature matrix by minimizing the reconstruction error and the coding sparsity constraint for subsequent modeling. In this embodiment, based on the perturbed sample matrix constructed in step S1 , an autoencoder reconstruction method integrating compressive learning and sparse constraint is proposed for jointly modeling and repairing the potential missing information and noise pollution in the perturbed data.

[0033] To effectively restore the expression integrity of the original features and avoid information redundancy or structural deformation introduced by the perturbation, in this embodiment, an autoencoder structure is preferably used as a repair tool for the input perturbed samples. The autoencoder is a non-linear dimensionality reduction and reconstruction structure that can learn the low-dimensional latent representation in the sample data and adaptively reconstruct the input data based on this.

[0034] The autoencoder includes a set of encoder functions and decoder functions , where: Encoder function , which is responsible for mapping the input samples to a low-dimensional latent space; Decoder function , which is responsible for restoring the latent representation to the original feature space.

[0035] Let the perturbed sample matrix be , then the encoder output is the latent variable matrix , corresponding to the low-dimensional latent representation of each sample. Subsequently, the decoder will reconstruct it into , that is, the restored matrix , with the goal of making the reconstruction result as close as possible to the original unperturbed sample.

[0036] To ensure that the model can not only restore the surface features during the reconstruction process, but also extract potential correlation information, while suppressing the interference caused by the input perturbation term to the output, in this embodiment, a dual loss constraint is introduced into the training process, including a reconstruction error term and a sparse regularization term. The overall loss function is in the following form: ; where: The first term is the average reconstruction error (in the form of the square of the L2 norm), which measures the difference between the decoded output and the original sample; The second term is the L1 regularization term of the latent coding vector, which introduces a sparsity constraint and encourages the model to extract concise and representative feature expressions; is the regularization weight coefficient, which is used to balance the relative importance between reconstruction accuracy and sparse representation.

[0037] During the training process, the reconstruction mechanism adjusts the parameters through backpropagation and gradient descent optimization. Preferably, activation functions (such as ReLU, LeakyReLU, etc.) can be used to improve the non-linear modeling ability, and combined with batch normalization operations to accelerate the convergence speed and improve the model stability.

[0038] It should be noted that the design goal of the auto-encoding structure in this embodiment is not limited to reconstructing the surface features of the input data. More importantly, it aims to jointly model the perturbed samples, identify the relevant structures and internal mapping relationships between the input features, so as to achieve automatic completion of possible missing items, active correction of outliers, and provide a higher-quality input feature matrix for subsequent graph structure modeling and feature selection.

[0039] The reconstructed output data matrix is denoted as in this embodiment. As the structure recovery result of the method of the present invention, this matrix has removed certain perturbation interferences in structure and filled in some missing or non-ideal observed features through latent space learning.

[0040] This step is not only a preliminary repair of the input samples, but also a fundamental link in the prediction framework of the present invention to achieve feature reconstruction and capture context correlations. In subsequent steps, the repaired data will be used as the direct input to the graph embedding module and the feature scoring mechanism, affecting the final feature screening and prediction performance.

[0041] In summary, in this embodiment, by introducing a structured sparse autoencoder, joint reconstruction of the perturbed data is performed to establish a non-linear mapping relationship between the perturbed samples and the original data, thereby effectively completing data missing repair and feature noise reduction processing, and forming a reconstructed feature matrix with structural integrity and embeddability. , which provides a solid foundation for data-driven modeling in the overall method of the present invention.

[0042] S3. Graph structure embedding modeling: Based on the power grid nodes and their connection relationships, construct an adjacency matrix, perform graph convolution processing on the repaired feature matrix, and generate an embedded feature representation to introduce the ability of the power grid topology structure to perceive the propagation of current fluctuations. In this embodiment, in order to more accurately depict the structural correlation relationships between various feature variables in the power system and make full use of their potential physical constraint information, before subsequent modeling of the feature matrix repaired in step S2 , preferably introduce a graph neural network modeling mechanism to perform structure-aware feature embedding processing on it.

[0043] In an actual power system, the power quantity features are not independent of each other, and there are often physical transmission paths and logical dependency relationships. For example, the current fluctuations of adjacent substation nodes have topological connections in space, and there are coupled evolution characteristics between voltage and load in time. Such complex structural relationships are difficult to directly model through traditional feature engineering. Therefore, in this embodiment, a graph structure modeling method is introduced to express and learn the mutual dependencies between features from the structural level.

[0044] Specifically, first construct a graph structure based on the physical or logical connection relationships between features , where: represents the feature set; represents the edge set between features, reflecting the correlation or adjacency relationship between each feature variable.

[0045] According to the constructed graph structure, define the adjacency matrix of the feature graph , where represents the feature is connected to the feature , represents no connection relationship.

[0046] To eliminate the imbalance caused by the difference in feature connectivity, preferably, the adjacency matrix is normalized to obtain a symmetric normalized adjacency matrix. , and its calculation method is as follows: ; Among them, is the original adjacency matrix, is the degree matrix, which is a diagonal matrix; Among them, is the degree matrix, , and the rest of the elements are 0, indicating the number of connections of each node in the graph. This normalization operation helps to maintain the weight balance of each node during the information propagation process and prevent the high-connectivity features from having a disproportionate impact on the embedding result.

[0047] After obtaining the normalized graph structure, a Graph Convolutional Network (GCN) is used to perform graph embedding operations on the repaired feature matrix . The graph convolution process is essentially an information aggregation process based on the node adjacency structure, and its embedding expression form in the present invention is: ; Among them: represents the feature representation matrix after graph embedding; is the input repaired feature matrix; is the graph convolution weight matrix, which needs to be optimized through training; is the activation function, and preferably, non-linear functions such as ReLU and tanh can be selected; is the aforementioned normalized adjacency matrix, ensuring the effective transmission of the graph structure during convolution propagation.

[0048] The core idea of the above graph convolution operation is that each feature representation is not only composed of its own information but also integrates the information of its adjacent feature nodes in the graph structure. This mechanism enables the feature expression to have the ability of structural context awareness, especially suitable for modeling the power linkage changes caused by the network topology in the power system.

[0049] In the method of the present invention, the feature matrix after graph embedding will be used in the subsequent feature importance evaluation and selection process. The embedded features retain the semantic information of the original features and at the same time introduce the dependency constraints of the structural hierarchy, which helps to more accurately evaluate the contribution degree of each feature in the entire network structure during the feature screening stage.

[0050] Preferably, in the implementation process, the graph structure can be constructed based on the physical topology of the power grid (such as node connection graph, electrical distance matrix, etc.), or a similarity graph can be constructed according to the statistical correlation between historical data, so as to realize the fusion between physical prior and data-driven.

[0051] In summary, in this embodiment, by introducing the graph structure embedding method, the connection relationship between various characteristic variables in the power system is incorporated into the modeling process, and the effective transmission and fusion of structural information are realized through graph convolution operations, generating a characteristic embedding result with structural expression ability, providing structural support for subsequent characteristic evaluation, screening and prediction modeling in the method of the present invention.

[0052] S4. Feature importance scoring and screening: Use the collaborative filtering mechanism to construct a feature-sample relationship matrix based on the prediction error, extract the feature scores through non-negative matrix factorization, and select the feature subset according to the set threshold. This subset will be used as the input features for the final prediction model training. In this embodiment, based on the graph structure embedding feature result obtained in step S3 and the feature matrix repaired in step S2 a feature importance scoring and screening method that combines the collaborative filtering mechanism and matrix factorization technology is proposed to identify a feature subset that significantly contributes to the current fluctuation prediction task from the original high-dimensional features, so as to reduce input redundancy, improve the generalization ability of the model, and provide an optimized feature representation for subsequent prediction modeling.

[0053] In the actual operation data of the power system, the influence degrees of different features on the prediction target are significantly different. Some features may have a weak dynamic correlation with the current fluctuation and may even introduce noise interference. Traditional feature selection methods based on a single statistical index or model weight often cannot fully consider the sensitivity of features to the prediction error in a perturbed environment. Therefore, in the solution of the present invention, the feature selection step not only focuses on the static correlation between features and the prediction target, but also introduces a feature scoring idea based on the "perturbation-response" mechanism, that is, by analyzing the influence of removing a certain feature on the prediction performance under the condition of perturbed samples to measure the prediction contribution degree of this feature.

[0054] Specifically, let the total number of samples be n and the total feature dimension be d. For each sample i and each feature j, construct a feature sensitivity evaluation matrix where each element represents the error change generated by the th sample in the prediction stage after removing the th feature, and is defined as follows: ; where: is the The true label value corresponding to a sample (i.e., the true current fluctuation amount); Indicates removing the sample feature vector obtained after the th feature; prediction model retrained in the sample space not containing the

[0055] This error matrix M reflects the influence degree of each feature on the prediction result under the condition of perturbed samples. If a significant increase in error occurs after a certain feature is removed, it can be considered that this feature has strong predictive value in the model.

[0056] To further extract the global feature importance scores from this error matrix, in this embodiment, the non-negative matrix factorization (Non-negative Matrix Factorization, NMF) method is preferably used to factorize the error matrix M: ; Where: is the feature factor matrix; is the sample factor matrix; is the set latent factor dimension, usually much smaller than and ; All matrix elements satisfy the non-negative constraint to maintain interpretability and sparsity.

[0057] Through the above decomposition process, the feature scores can be calculated based on the row norms (or row averages) of matrix U. Preferably, in this embodiment, the L2 norm is used as the feature scoring index, that is, for each feature j, its score is defined as: ; Where, represents the th row of matrix . This score represents the contribution degree of feature

[0058] In the error matrix, the higher the score, the more significant the role of this feature in error reconstruction, and thus it can be considered more predictive. After obtaining the scoring results of all features , a threshold can be set for screening, and the features with scores not lower than this threshold are retained to form a feature subset ; Preferably, the threshold can be determined through cross-validation, stability analysis, or a feedback mechanism based on structural metrics. Additionally, the feature scores can be weighted by combining the node centrality metrics in the graph embedding stage to improve the structural consistency and system interpretability of the screening results.

[0059] This feature subset will ultimately serve as the effective input dimension in the training stage of the prediction model of the present invention, directly affecting the subsequent learning efficiency and performance of the model. It should be noted that this screening method not only considers the influence degree of features on the prediction error, but also models the contribution structure of each feature to the prediction residual space under the perturbation environment through NMF factorization, possessing strong global adaptability and module interpretability.

[0060] In summary, in this embodiment, by constructing a perturbation error matrix and introducing a non-negative matrix factorization mechanism, the importance measurement and effective screening of input features in the prediction task are achieved, forming a feature subset that not only has data-driven basis but also integrates system structure perception, providing refined and high-quality input support for subsequent model training.

[0061] S5. Multi-objective loss modeling and prediction training: Based on the feature subset, construct a current fluctuation prediction model. During the training process, optimize the multi-objective loss function including the prediction error term, perturbation repair stability term, and graph structure consistency term to improve the accuracy and system coordination of the model. In this embodiment, after completing the selection of the feature subset in step S4, the embedded features corresponding to this feature subset are used as the main input of the prediction model of the present invention to construct a learning model for predicting the current fluctuation in the power system, and a multi-objective loss function is proposed to jointly optimize the prediction accuracy, perturbation robustness, and structure consistency, thereby improving the robustness and generalization ability of the model in practical applications.

[0062] First, let the input feature matrix finally used for training be , which is composed of the features selected by step S4 from the original feature matrix and can be a part of the structural representation after graph embedding or the original reconstructed feature representation, and the specific selection depends on the application strategy.

[0063] In the present invention, the prediction target is the current fluctuation at a certain time point or a certain time period in the power system, denoted as , where represents the prediction model, is the sample input feature vector, is the prediction output result, is the true label corresponding to this sample.

[0064] To achieve the systematic optimization of the prediction model, in this embodiment, a multi-objective joint loss function is designed to simultaneously constrain the prediction error of the model, the stability of the perturbed samples, and the graph structure preservation ability. The form of the multi-objective loss function is as follows: ; Where: is the prediction error term, which is used to measure the distance between the model output and the true value. Preferably, it can be expressed in the form of mean squared error (MSE) as: ; This term is the basic error index in supervised learning and directly reflects the fitting degree of the model to the target variable.

[0065] is the perturbation stability constraint term, which is used to ensure that the model is robust to input perturbations and is defined as the difference in the prediction outputs between the original sample and the perturbed sample. Let the input of the perturbed sample be , and its predicted value be , then this term can be expressed as: ; This term ensures that the model remains stable in the face of input noise, missing values, or perturbations, thereby enhancing the anti-interference ability during the operation of the system.

[0066] is the structure consistency constraint term, which is mainly used to maintain the consistency of the structural features before and after graph embedding. Preferably, it can be achieved through graph Laplacian constraint, feature reconstruction error, or graph contrast loss. Let the feature before embedding be , and the graph embedding feature be , then the consistency loss term can be simplified as: ; Among them, is the structure preservation mapping matrix, represents the Frobenius norm, which measures the deviation of the embedded feature from the original feature in the overall representation.

[0067] In the above loss function, α and β are adjustable weight coefficients, which are used to balance the importance of each sub-task objective. The loss function can be jointly optimized through the backpropagation mechanism to drive the model to achieve coordination and balance among different tasks.

[0068] Preferably, the model structure can be a deep feedforward neural network (such as a multi-layer perceptron), a time series network (such as LSTM) or a graph neural network (such as GCN / GAT). The specific architecture can be flexibly selected according to the target prediction cycle and task characteristics. The model training process can use a gradient descent optimization algorithm, such as Adam or RMSProp, with an appropriate learning rate adjustment strategy to achieve rapid convergence.

[0069] After the training is completed, the obtained prediction model parameters will be used in the subsequent online prediction process to estimate the current fluctuation trend of new input samples. The prediction results can be used as auxiliary decision-making information in the power grid dispatching system, or in business modules such as load anomaly detection and system stability assessment.

[0070] In summary, this embodiment introduces a structured multi-objective loss optimization mechanism in the current fluctuation prediction model training process. By jointly considering prediction accuracy, disturbance robustness and graph structure preservation, the adaptability of the model in the complex environment of the power system is improved, and a complete and implementable training strategy and optimization framework are provided for achieving high-reliability prediction tasks.

[0071] S6, module feedback linkage adjustment: During the training process, the feature scoring results are fed back to the perturbation reconstruction module to focus on strengthening the restoration accuracy of highly sensitive features. At the same time, the structural indicators in the graph embedding are used to optimize the feature screening weights to achieve information closed-loop adjustment between the functional modules; In this embodiment, during the process of completing the training and construction of the prediction model, a module-to-module feedback linkage mechanism is further introduced to open up the information path between the functional sub-modules of the present invention, and realize the coordinated adjustment between feature evaluation, graph structure modeling, disturbance repair and prediction optimization, thereby improving the stability and learning consistency of the overall model in the scenario of systematic data disturbance.

[0072] In the overall process of the present invention, although each module has a clear division of labor in terms of function, there is a close logical dependency between its input and output. For example, the reconstruction accuracy of the perturbed sample repair module in step S2 directly affects the subsequent graph embedding expression effect and feature scoring stability; and the result of feature selection in step S4 determines the final input dimension of the model, which further affects the model prediction performance and the consistency of graph structure embedding.

[0073] Therefore, in order to avoid local optimality or decision deviation caused by information fragmentation between modules, this embodiment introduces the following two types of feedback mechanisms during the training process: In this embodiment, firstly based on the feature scoring result obtained in step S4 , a set of highly sensitive features is identified, that is, the features that cause a significant increase in the prediction error during the feature elimination process. This set is fed back to the autoencoder reconstruction module in step S2 to guide the feature weighting strategy during the reconstruction process. Specifically, a feature weighting term is introduced into the original reconstruction loss function, so that the model pays more attention to the reconstruction accuracy of highly sensitive features during the reconstruction stage. The modified loss function is as follows: ; Where: is the reconstruction weight of feature , and preferably can be determined according to its normalized score in the feature scoring; is the reconstruction value of the -th feature of sample ; The remaining symbols are consistent with the definitions in step S2.

[0074] In this way, the model can perform differential modeling according to the feedback of feature importance in the data repair stage, effectively improving the information consistency of the entire process.

[0075] Furthermore, to enhance the fusion between graph structure modeling and feature selection, in this embodiment, the structural indicators obtained in the graph embedding stage are also introduced as feedback signals to participate in the weighted operation of feature importance scoring. Specifically, in step S3, after graph convolution, the structural centrality measure of each feature node can be calculated, such as node degree, PageRank value, or average attention weight based on the embedding space, denoted as . This structural indicator can be combined with the feature score to construct a weighted scoring function: ; Where: , are adjustable weight coefficients used to control the influence ratio of prediction error sensitivity and graph structure weight; is the comprehensive score finally used for feature screening.

[0076] The above structural feedback mechanism ensures that feature selection not only considers the contribution of prediction error, but also incorporates the key degree of features in the graph structure, thus being more conducive to constructing a feature subset with reasonable structure and effective prediction.

[0077] Preferably, in this embodiment, the graph embedding weight matrix W can be further optimized according to the final training error feedback, and parameter cascade update can be achieved through an end-to-end optimization mechanism. If a graph neural network is used as the main model, the above graph embedding weights can be jointly trained directly under the drive of the prediction loss to form an overall unified optimization closed-loop.

[0078] In summary, in this embodiment, by constructing two types of feedback paths, namely "feature scoring → reconstruction weighting" and "graph structure index → feature weighting", the parameter dependence and information coupling between each processing module are unblocked, and an adaptive linkage learning mechanism is constructed, thereby improving the overall modeling coordination and stability of the prediction model of the present invention in a complex disturbance environment.

[0079] S7. Prediction output: Use the trained prediction model to predict the current fluctuation of new input data and output the prediction result. In this embodiment, after completing the model training and multi-objective joint optimization, based on the prediction model obtained in step S5 , and the feature subset selected through step S4 , perform the prediction operation on the future current fluctuation amount in the power system and achieve the final prediction output result.

[0080] To ensure the consistency and stability of the input of the prediction model, in this step, it is preferably to preprocess the new sample data to be predicted according to the same feature processing flow as the training data. Specifically, it includes: constraining the dimension of the input feature matrix according to the feature subset ; normalizing or standardizing the features to match the feature distribution in the training stage; in the case of structural embedding requirements, synchronously constructing the graph structure information and calculating the embedding representation.

[0081] Let the set of samples to be predicted be denoted as , where each sample represents the feature input after feature screening and processing. Input it into the trained prediction model , and the corresponding prediction output can be obtained: ; The above model preferably has an internal structure of a feedforward regression model in this embodiment. Its training process has been completed in step S5, and the parameters have been optimized and learned based on the joint loss function. If the model used includes a graph neural network or a time series structure, the prediction function can also perform forward calculation through the corresponding graph convolution layer or sequence propagation mechanism.

[0082] In the case of graph embedding requirements, the graph representation of the prediction sample needs to be synchronously constructed through the graph structure generation method defined in step S3, and then the input features are subjected to graph convolution operation using the normalized adjacency matrix and the weight matrix W learned in the training stage. Its graph embedding representation can be expressed as: ; Among them, represents the original prediction sample feature matrix, It is the graph embedding result, which can be used as the model input or concatenated with the original features and then input into the model.

[0083] Prediction result It can be expressed as the current change amount of a certain node or multiple nodes in the power system at a future moment or the average fluctuation amount within the prediction interval range. This output form can be adjusted according to actual business needs, including but not limited to single-point prediction, sliding window prediction, or multi-step prediction, etc.

[0084] In the solution of the present invention, the prediction result can be directly used in the power dispatching center, the operation control system, or the power grid load prediction module, and can also be linked with the alarm module or the load response control system to identify potential load abnormalities, system overloads, or prediction deviations.

[0085] Preferably, to improve the interpretability and traceability of the prediction output in the business system, this embodiment can also synchronously output intermediate quantity information such as feature response weights and node embedding activation intensities to assist the subsequent decision-making module in analyzing the prediction formation process.

[0086] In summary, through the forward inference process based on the final trained model in this embodiment, the prediction of the current fluctuation amount of the newly input sample is completed, forming the prediction output mechanism of the last link in the method of the present invention. On the basis of maintaining consistency with the training stage, by introducing the collaborative mechanism of structure embedding and feature screening, it ensures that the output is reasonable, stable, and practical, constituting a deployable and scalable prediction execution scheme for the actual scenario.

[0087] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting current fluctuations in power systems based on multivariate data and feature selection, characterized in that: The steps include: S1. Data disturbance enhancement processing: Collecting heterogeneous original data of the power system including current, voltage, load and meteorology, constructing a feature matrix, introducing disturbance terms into the data, and forming a disturbance sample matrix to enhance the robustness of the model against data missing and noise interference; S2. Restoration and reconstruction of perturbation samples: Input the perturbation sample matrix into the autoencoder structure, and obtain the repaired feature matrix by minimizing the reconstruction error and the coding sparsity constraint for subsequent modeling; S3, graph structure embedding modeling: construct an adjacency matrix based on the grid nodes and their connection relationships, perform graph convolution processing on the repaired feature matrix, and generate an embedded feature representation to introduce the grid topology structure's ability to perceive the propagation of current fluctuations; S4. Feature importance scoring and screening: Use collaborative filtering mechanism to build a feature-sample relationship matrix based on prediction error, extract feature scores through non-negative matrix decomposition, and select feature subsets based on set thresholds. This subset will be used as the input features for prediction model training. S5. Multi-objective loss modeling and prediction training: A current fluctuation prediction model is constructed based on the feature subset. During the training process, a multi-objective loss function including a prediction error term, a disturbance repair stability term, and a graph structure consistency term is optimized to improve the accuracy of the model and the coordination of the system. S6, module feedback linkage adjustment: During the training process, the feature scoring results are fed back to the perturbation reconstruction module to focus on strengthening the restoration accuracy of highly sensitive features. At the same time, the structural indicators in the graph embedding are used to optimize the feature screening weights to achieve information closed-loop adjustment between the functional modules; S7. Prediction output: Use the trained prediction model to predict the current fluctuation of new input data and output the prediction results.

2. The method for predicting electric power system current fluctuation based on multivariate data and feature selection according to claim 1, characterized in that: The disturbance enhancement process in step S1 is: adding a noise term that obeys Gaussian distribution to the original feature matrix to construct a disturbance sample ； in, is the original feature matrix, is the input sample matrix, is subject to the mean of 0 and the covariance of The normally distributed noise matrix is is the identity matrix.

3. The method for predicting electric power system current fluctuation based on multivariate data and feature selection according to claim 1, characterized in that: The reconstruction loss function of the autoencoder structure in step S2 is defined as: ； in, is the input perturbation sample matrix, is the encoder function, is the decoder function, is the weight factor of the sparse regularization term.

4. The method for predicting electric power system current fluctuation based on multivariate data and feature selection according to claim 1, characterized in that: The graph structure embedding in step S3 adopts the normalized adjacency matrix ； in, is the original adjacency matrix, is the degree matrix, which is a diagonal matrix; And calculate the graph embedding feature representation ； in, is the embedding feature matrix output by graph convolution, is the repaired feature matrix, is the trainable weight matrix, is the activation function.

5. The method for predicting electric power system current fluctuation based on multivariate data and feature selection according to claim 1, characterized in that: The feature-sample error matrix constructed in step S4 The elements of are defined as: ； in, For the The true value of the samples, To remove The prediction model after features, For the corresponding input.

6. The method for predicting electric power system current fluctuation based on multivariate data and feature selection according to claim 1, characterized in that: In step S4, non-negative matrix decomposition M≈UV is performed on the feature-sample relationship matrix, and the feature score is the L2 norm of each row of the matrix U And according to the set threshold τ, the feature subset Sf={j|sj≥τ} is obtained.

7. The method for predicting electric power system current fluctuation based on multivariate data and feature selection according to claim 1, characterized in that: The feedback mechanism in step S6 includes feeding back the high-scoring features identified in step S4 to the autoencoder structure in step S2, and performing a weighted processing on the reconstruction process thereof.

8. The method for predicting electric power system current fluctuation based on multivariate data and feature selection according to claim 1, characterized in that: The feedback mechanism in step S6 also includes using the node centrality index obtained in the graph embedding to participate in the weighted calculation of feature scoring.

9. The method for predicting electric power system current fluctuation based on multivariate data and feature selection according to claim 1, characterized in that: The parameter optimization process of the prediction model in step S5 adopts a Bayesian optimization strategy to minimize the multi-objective loss function.

Images