A power system state estimation method

Abstract

The application discloses a power system state estimation method, and relates to the technical field of power system regulation and control, and aims to solve the problem of low precision and robustness of power system state estimation. The method comprises the following steps: acquiring multi-section measurement data of a power grid model; generating a first mask matrix for data anomalies caused by missing measurements; generating a second mask matrix for data anomalies caused by power grid topology changes; assigning a learnable embedding vector to each measurement point and fusing the measurement data of the measurement point through a gating mechanism to obtain fused measurement features; performing missing data filling processing and zeroing processing on the fused measurement features, and performing channel weighting to obtain mixed measurement features of each measurement point; and inputting the mixed measurement features into a multilayer perception network to output power system state estimation results corresponding to each measurement point. The application has the effect of improving the precision and robustness of power system state estimation.

Description

Technical Field

[0001] This invention relates to the field of power system control technology, and in particular to a power system state estimation method based on learnable embedded vectors and gating mechanisms. Background Technology

[0002] Power system state estimation is a fundamental aspect of power dispatching and operation control. Its purpose is to accurately estimate the operating state of the power grid in the presence of measurement errors and incomplete data. Traditional state estimation methods are typically based on model-driven methods such as weighted least squares (WLS), which are highly dependent on the completeness of measurement data and the accuracy of the power grid topology.

[0003] In real-world operating environments, power system measurement data commonly suffers from missing measurements and disconnection issues due to grid topology changes, caused by communication failures, equipment malfunctions, and maintenance adjustments. Existing methods typically handle these abnormal data by filling in fixed values ​​or using simple interpolation, which can easily lead to outliers being mistakenly included in calculations as actual measurements, thereby reducing the accuracy and robustness of state estimation.

[0004] In recent years, deep learning-based methods have been introduced into the field of power system state estimation. However, existing models often simply treat missing and unconnected data as input features, lacking effective modeling of the reliability differences of measurement points, and making it difficult to obtain stable estimation results under complex operating conditions and high missing data scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide a power system state estimation method to improve the accuracy, robustness and generalization ability of state estimation.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A power system state estimation method, comprising:

[0008] Acquire multi-section measurement data of the power grid model;

[0009] For data anomalies caused by missing measurements in the measurement data, a first mask matrix is ​​generated to identify the missing measurements; for data anomalies caused by changes in the power grid topology that result in non-connection, a second mask matrix is ​​generated to identify the non-connection of the topology.

[0010] A corresponding learnable embedding vector is assigned to each measurement point in the power grid model, and the learnable embedding vector is fused with the measurement data of that measurement point through a gating mechanism to obtain fused measurement features;

[0011] Under the constraints of the first mask matrix and the second mask matrix, missing data filling processing and zeroing processing of topology-unconnected measurements are performed on the fused measurement features, and channel weighting is performed on the processed fused measurement features to obtain the mixed measurement features of each measurement point.

[0012] The hybrid measurement features are input into a multilayer perceptron network, which outputs the power system state estimation results corresponding to each measurement point.

[0013] More specifically, the first mask matrix and the second mask matrix are mutually exclusive at the same measurement point.

[0014] More specifically, the process of fusing the learnable embedding vector with the measurement data of the measurement point through a gating mechanism to obtain the fused measurement features is expressed by the following formula:

[0015] x hi = g i · x i +(1- g i )· e i ;

[0016] In the formula, x hi This represents the fused measurement characteristics of a certain measurement point. x i This represents the original measurement value of the measurement point. e i This represents the corresponding learnable embedding vector. g i This represents the gating parameters.

[0017] More specifically, the application of channel weights to the processed fused measurement features is expressed by the following formula:

[0018] ;

[0019] in, s j Indicates the first j The channel weights and scaling factors corresponding to each feature channel; This represents the fused measurement feature of the i-th measurement point on the j-th channel; This indicates the characteristics of mixed measurement.

[0020] More specifically, the multilayer perceptron network is replaced with a graph neural network or an attention network.

[0021] Compared with existing technologies, the power system state estimation method provided by this invention has the following advantages:

[0022] By constructing learnable embedding vectors to represent measurement points and power grid structural features, and introducing a gating mechanism to adaptively filter and weightedly fuse different measurement information, robust estimation of power system state is achieved without relying on manual rules, and the accuracy of power system state estimation is improved. Attached Figure Description

[0023] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0024] Figure 1 A flowchart of a method provided in an embodiment of the present invention. Detailed Implementation

[0025] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

[0026] To facilitate a clear description of the technical solutions in the embodiments of the present invention, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. For example, the first threshold and the second threshold are merely used to distinguish different thresholds and do not limit their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0027] It should be noted that in this invention, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0028] The embodiments of the present invention will now be described in further detail with reference to the accompanying drawings.

[0029] like Figure 1 As shown in the figure, this embodiment of the invention provides a power system state estimation method, which specifically includes the following steps.

[0030] Step S1: Obtain multi-section measurement data of the power grid model.

[0031] The power grid model contains multiple measurement points. The measurement data for each measurement point includes a measurement value, a measurement missing flag, and a topology disconnection flag. Based on these measurement values, preset outlier detection statistical rules can be used to detect outliers in the measurement data of each measurement point. These statistical rules include outlier detection methods based on historical statistical characteristics, such as three... σ Principle (3) σ in principle), Z-score Outlier detection, etc.; when the measured value of a certain measurement point does not meet the reasonable range or deviation threshold corresponding to the statistical rule, the measurement data corresponding to the measurement point is determined to be an abnormal measurement, thereby completing the identification of abnormal measurements.

[0032] Step S2: For data anomalies caused by missing measurements in the measurement data, generate a first mask matrix to identify missing measurements; for data anomalies caused by changes in the power grid topology that result in unconnected data in the measurement data, generate a second mask matrix to identify unconnected topology data.

[0033] Specifically, when the measurement data of a certain measurement point does not meet the statistical rules or is missing, the measurement point is determined to be an abnormal measurement point, and the corresponding element in the measurement data is set to 1. For measurement points that are not determined to be abnormal, the corresponding element is set to 0. The measurement data of each measurement point that is abnormal due to missing measurement are summarized to form a complete first mask matrix.

[0034] Based on the current power grid topology information, it is determined whether each measurement point maintains an effective connection with the power grid model under the current operating state. When a measurement point is not involved in the current power grid topology due to a change in switch status, equipment shutdown, or network structure adjustment, it is determined that the measurement point is an unconnected measurement point, and the corresponding element in the measurement data of the measurement point is set to 1. The corresponding element of the measurement point in the normal connection state is set to 0. All the measurement data that are not connected due to changes in power grid topology are summarized to form a complete second mask matrix.

[0035] For the same measurement point, since only one case of measurement missing or topology not connected will occur in the measurement data, only one of the corresponding elements in the first mask matrix and the second mask matrix is ​​allowed to be 1. That is, the identification relationship between the first mask matrix and the second mask matrix at this measurement point is always mutually exclusive.

[0036] Step S3: Assign a corresponding learnable embedding vector to each measurement point in the power grid model, and fuse the learnable embedding vector with the measurement data of the measurement point through a gating mechanism to obtain fused measurement features.

[0037] For any measurement point in the power grid model, its fused measurement characteristics satisfy the following relationship:

[0038] x hi = g i × x i +(1- g i )× e i ;

[0039] In the formula, x hi This indicates the fusion measurement characteristics of the measurement points; x i This represents the original measurement value of the measurement point; e i This represents the corresponding learnable embedding vector; g i This represents the gating parameter, used to characterize the reliability of the measurement value at this measurement point.

[0040] Among them, the gating mechanism is used to adaptively weight and fuse measurement data with learnable embedding vectors to adjust the contribution of different measurement information in the fused measurement features.

[0041] Step S4: Under the constraints of the first mask matrix and the second mask matrix, perform missing data filling processing and zeroing processing for topology-unconnected measurements on the fused measurement features, and apply channel weights and scaling factors to the processed fused measurement features to obtain the mixed measurement features of each measurement point.

[0042] For measurement points with missing measurements, the missing measurements are estimated and filled using the effective measurement values ​​of adjacent nodes based on the power grid topology and Kirchhoff's current law and voltage law. For measurement points not connected to the topology, their measurement values ​​are set to zero according to the physical constraints of the power system, since their corresponding branches or nodes do not participate in power transmission in the current operating topology.

[0043] Based on this, the hybrid measurement features corresponding to each measurement point are regarded as an independent measurement channel, and learnable channel weights and scaling coefficients are applied to each measurement channel to enhance the ability to express key measurement information, adjust the contribution of features corresponding to different measurement points in state estimation, and thus enhance the discriminability of different measurement points in state estimation.

[0044] Specifically, let the fusion measurement features obtained after gated fusion, missing fill, and zeroing out of topology be as follows: ,in i Indicates the measurement point index. j This represents the feature channel index. A learnable scaling vector is introduced. s =[ s 1 ,s 2 ,s 3 , … ,sd ]∈ R d , for fusion measurement features Global weighting is performed on each feature channel to obtain the hybrid measurement feature. Its mathematical expression is:

[0045] ;

[0046] in, s j Indicates the first j The channel weights and scaling factors corresponding to each feature channel are used to characterize the relative importance of different measurement features in the power system state estimation task; This represents the fused measurement feature of the i-th measurement point on the j-th channel; This represents the mixed measurement features. In the scaling vector, R represents the set of real numbers, d represents the vector consisting of d real numbers, and the entire state space is a d-dimensional continuous space. By performing missing data imputation and topology disconnection zeroing under the constraints of the first and second mask matrices, the interference of abnormal measurements and invalid topology information on the model input can be avoided. Furthermore, by combining channel weighting and scaling operations, the model adaptively strengthens its modeling ability for key measurement features during training, suppresses the influence of redundant or low-confidence features, and thus improves the accuracy and robustness of power system state estimation results.

[0047] Step S5: Input the hybrid measurement features into the multilayer perceptron network and output the power system state estimation results corresponding to each measurement point.

[0048] The multilayer perceptron network is a fully connected neural network with three or more layers. In some embodiments, the multilayer perceptron network can be replaced with a graph neural network or an attention network. Using multilayer perceptron networks, graph neural networks, or attention networks to analyze the state estimation results of a power system is prior art and will not be elaborated upon here.

[0049] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely exemplary descriptions of the invention as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A power system state estimation method, characterized in that, include: Acquire multi-section measurement data of the power grid model; For data anomalies caused by missing measurements in the measurement data, a first mask matrix is ​​generated to identify the missing measurements; for data anomalies caused by changes in the power grid topology that result in non-connection, a second mask matrix is ​​generated to identify the non-connection of the topology. A learnable embedding vector is assigned to each measurement point in the power grid model, and the learnable embedding vector is fused with the measurement data of that measurement point through a gating mechanism to obtain the fused measurement features, expressed by the following formula: x hi = g i · x i +(1- g i )· e i ； In the formula, x hi Indicates the fusion measurement characteristics of measurement points ，x i This represents the original measurement value of the measurement point. e i This represents the corresponding learnable embedding vector. g i Indicates the gating parameters; Under the constraints of the first and second mask matrices, missing data imputation and zeroing of topology-unconnected measurements are performed on the fused measurement features. Channel weights and scaling factors are then applied to the processed fused measurement features to obtain the mixed measurement features of each measurement point, as expressed by the following formula: ； In the formula, s j This represents the channel weight and scaling factor corresponding to the j-th feature channel; This represents the fused measurement feature of the i-th measurement point on the j-th channel; Indicates mixed measurement characteristics; The hybrid measurement features are input into a multilayer perceptron network, which outputs the power system state estimation results corresponding to each measurement point.

2. The power system state estimation method according to claim 1, characterized in that, The first mask matrix and the second mask matrix are mutually exclusive at the same measurement point.

3. The power system state estimation method according to claim 1, characterized in that, Replace the multilayer perceptron network with a graph neural network or an attention network.

Images