Intelligent filling method for power flow data based on space-time coupling

Abstract

The invention belongs to the technical field of power system application, and particularly relates to an intelligent power flow data filling method based on space-time coupling. The invention provides an intelligent power flow data filling method based on space-time coupling, which comprises the following steps: firstly, carrying out data preprocessing on original data through Hampel filtering, phase alignment and normalization, constructing a space-time encoder through joint coding of adaptive graph convolution (Adaptive GCN) and attention LSTM, capturing space-time coupling characteristics, and carrying out data processing on the space-time coupling characteristics; a multi-physical constraint generative adversarial network (Pry-GAN) is designed through generator physical constraints and a multi-scale discriminator to realize data filling, and it is ensured that a filling result has both numerical reasonability and physical feasibility.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system application, and in particular relates to a method for intelligently filling power flow data based on time-space coupling. Background Art

[0002] With the accelerated advancement of the construction of new power systems, the power grid is undergoing a profound transformation from a traditional centralized architecture to a collaborative interactive mode of "source-grid-load-storage". The high proportion of renewable energy access, large-scale deployment of distributed power sources, and widespread penetration of power electronic equipment have made the power grid operation conditions present characteristics such as strong randomness, multi-time scale coupling, and complex spatiotemporal correlation. In this context, power flow data, as the core carrier reflecting the real-time operating status of the power grid, has a direct impact on the reliability of key functions such as state estimation, safety assessment, and optimal scheduling. However, the problem of data loss in actual systems due to factors such as measurement device failure, communication link interruption, and network attacks has become increasingly prominent.

[0003] Traditional filling methods based on statistical principles (such as mean filling and time series interpolation) are ineffective in dealing with complex nonlinear spatiotemporal correlations. Although pure data-driven deep learning models can capture complex patterns, they often ignore the physical laws of the power grid, resulting in filling results that violate basic constraints such as circuit principles, which may cause chain errors in subsequent analysis. Summary of the invention

[0004] In view of the technical problems existing in filling power flow data, the present invention proposes an intelligent filling method for power flow data based on time-space coupling, which has reasonable design, simple method, strong theoretical basis and can realize the intelligence and rationality of filling power flow data.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: a method for intelligent filling of power flow data based on time-space coupling, comprising the following steps:

[0006] S1, preprocessing the raw data by Hampel filtering, phase alignment and normalization;

[0007] S2, through the joint encoding of adaptive graph convolution Adaptive GCN and attention LSTM, a spatiotemporal encoder is constructed to capture the spatiotemporal coupling feature Z;

[0008] S3. Design a multi-physics constraint generative adversarial network Phy-GAN through generator physical constraints and multi-scale discriminator;

[0009] S4. Expand the feature dimensions of the missing mask M and the covariate C so that they have the same spatial dimension as the spatiotemporal coupling feature Z. Then concatenate the expanded M, C and Z as the generator input of the generative adversarial network Phy-GAN, and achieve data filling through the discriminator and adversarial training.

[0010] Preferably, the step S1 performs phase alignment on the original data, and the formula is: ,in, represents the continuous phase after expansion, Indicates the current phase. represents the phase angle at the previous moment, Represents the rounding function.

[0011] Preferably, in step S2, the attention LSTM introduces a temporal attention mechanism into the LSTM, and the spatiotemporal encoder performs gated feature fusion on the spatial feature matrix obtained by the adaptive graph convolution Adaptive GCN and the temporal feature matrix obtained by the attention LSTM, sets the gating weight, and obtains the spatiotemporal coupling feature: ,in, is the spatiotemporal coupling feature, is the gating weight, is the trainable fusion weight matrix, is the spatial feature matrix, is the time feature matrix, Indicates the multiplication of corresponding positions of matrices, Represents the Sigmoid activation function;

[0012] Preferably, in step S3, the multi-physical constraint generative adversarial network adds a physical constraint layer after the generator output layer, including dynamic constraints and algebraic constraints.

[0013] Preferably, in step S3, the multi-scale discriminator of the multi-physics constraint generative adversarial network is: ,in, Score the authenticity of the data, X represents the input data, s represents the downsampling scale, , represents the discriminant subnetwork of the corresponding scale, represents average pooling.

[0014] Preferably, the missing mask M in step S4 is a binary matrix identifying the location of missing data, and the covariate C is a vector of external variables, circuit breaker status and weather encoding.

[0015] Preferably, the dynamic constraints and algebraic constraints are: , ,in, represents the node voltage vector, represents the branch current vector, represents the node admittance matrix, represents the conjugate complex number of the branch current vector, represents taking the real part of a complex number, Indicates the actual measured active power value.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: the present invention proposes a method for intelligent filling of power flow data based on spatiotemporal coupling. First, the original data is preprocessed by Hampel filtering, phase alignment and normalization. A spatiotemporal encoder is constructed through the joint encoding of adaptive graph convolution Adaptive GCN and attention LSTM to capture the spatiotemporal coupling characteristics. The multi-physical constraint generative adversarial network Phy-GAN is designed through the generator physical constraints and multi-scale discriminators to achieve data filling, ensuring that the filling results have both numerical rationality and physical feasibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A schematic diagram of a flow chart of a method for intelligently filling power flow data based on time-space coupling provided by an embodiment of the present invention;

[0019] Figure 2 A schematic diagram of a data preprocessing process provided by an embodiment of the present invention;

[0020] Figure 3 A schematic diagram of the structure of a multi-physics constraint generative adversarial network provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.

[0023] Examples, such as Figure 1 , 2 As shown in 3, considering that traditional filling methods based on statistical principles (such as mean filling and time series interpolation) are weak in dealing with complex nonlinear spatiotemporal correlations, and pure data-driven deep learning models can capture complex patterns, they often ignore the physical laws of the power grid, resulting in the filling results violating basic constraints such as circuit principles, which may cause chain errors in subsequent analysis. For this reason, the present invention provides an intelligent filling method for power flow data based on spatiotemporal coupling. Taking into account the existence of timestamp offset, sensor noise and dimensional differences in the original power data, directly inputting the model will lead to unstable training. Therefore, this step eliminates data heterogeneity and improves model convergence efficiency through filtering, alignment and normalization operations.

[0024] Considering the statistical characteristics of Hampel filtering based on median and absolute deviation, it can effectively identify and replace impulse noise (such as outliers caused by communication interference). Compared with mean filtering, it is more resistant to outlier interference. The sliding window mechanism can adapt to the noise level in different time periods, avoid misjudgment of fixed thresholds during load fluctuations, and only replace points that significantly deviate from the statistical characteristics of the window, retaining real power events (such as sudden changes caused by faults). Traditional low-pass filtering will smooth real fluctuations, while Hampel filtering will retain key dynamic information while removing noise.

[0025] Phase alignment eliminates phase jumps, which are caused by the periodicity of the voltage / current phasor angle modulo 2π, ensuring that the time series is continuous and differentiable, and avoiding the model from misjudging it as a sudden change. Continuous phase changes are more consistent with the modeling assumptions of timing networks such as LSTM, reducing gradient anomalies. The aligned phase can be directly used for complex operations (such as power calculation) without additional processing.

[0026] Normalization eliminates dimensional differences and unifies parameters of different dimensions (such as kV, MW) into per-unit values, avoiding the model's bias towards large-scale features (such as power suppressing voltage) due to differences in numerical ranges. The generalization ability is enhanced, and the same model can be adapted to power grids of different voltage levels (such as 10kV distribution network and 500kV main network) without retraining. Numerical stability is particularly effective in high-dimensional parameter space.

[0027] Therefore, first of all, the raw data is preprocessed by Hampel filtering, phase alignment and normalization. Specifically, the formula for phase alignment of the raw data is: ,in, represents the continuous phase after expansion, Indicates the current phase. represents the phase angle at the previous moment, Represents the rounding function.

[0028] Considering that the power of power grid nodes is jointly affected by topological connection (space) and load fluctuation (time). Traditional methods independently process the space-time dimension, resulting in information fragmentation. This module captures the space-time coupling characteristics through the joint encoding of adaptive graph convolution and attention LSTM. Therefore, through the joint encoding of adaptive graph convolution Adaptive GCN and attention LSTM, a space-time encoder is constructed to capture the space-time coupling feature Z. Specifically, the attention LSTM introduces the time attention mechanism in LSTM. The space-time encoder performs gated feature fusion on the spatial feature matrix obtained by the adaptive graph convolution Adaptive GCN and the temporal feature matrix obtained by the attention LSTM, sets the gating weight, and obtains the space-time coupling feature: ,in, is the spatiotemporal coupling feature, is the gating weight, is the trainable fusion weight matrix, is the spatial feature matrix, is the time feature matrix, Indicates the multiplication of corresponding positions of matrices, Represents the Sigmoid activation function. Since this fusion is not ordinary addition, but selective fusion using a matrix with a gating mechanism, it can retain more useful information, remove noise information, and reduce information loss. The joint encoding of adaptive graph convolution and attention LSTM generates asymmetric adjacency relationships through a learnable embedding matrix, captures actual electrical coupling (such as the strong correlation of high-load lines), and transcends the limitations of fixed topology. Balance computational efficiency and receptive field, covering multi-hop neighbors (such as capturing the indirect impact of adjacent substations). Attention LSTM introduces the temporal attention mechanism in LSTM, automatically focusing on key periods (such as load peaks), and alleviating the problem of insufficient modeling of long-distance dependencies in traditional LSTM.

[0029] Considering that the generative adversarial network can synthesize data distribution, but the power data must strictly satisfy the laws of physics. This module ensures that the filling results have both numerical rationality and physical feasibility through differential algebraic equation constraints and multi-scale discriminator design. Therefore, the multi-physics constraint generative adversarial network Phy-GAN is designed through the generator physical constraints and multi-scale discriminators. The multi-physics constraint generative adversarial network adds a physical constraint layer after the generator output layer, including dynamic constraints and algebraic constraints, which are: , ,in, represents the node voltage vector, represents the branch current vector, represents the node admittance matrix, represents the conjugate complex number of the branch current vector, represents taking the real part of a complex number, Represents the actual measured active power value. The multi-scale discriminator of the multi-physics constraint generative adversarial network is: ,in, Score the authenticity of the data, X represents the input data, s represents the downsampling scale, , represents the discriminant subnetwork of the corresponding scale, Represents average pooling. Multi-scale joint training forces the generator to cover the full-dimensional feature distribution and avoid generating single-mode data.

[0030] Finally, the missing mask M and covariate C are expanded in feature dimension to have the same spatial dimension as the spatiotemporal coupling feature Z, and then the expanded M, C and Z are spliced ​​as the generator input of the generative adversarial network Phy-GAN, and data filling is achieved through the discriminator and adversarial training. The missing mask M is a binary matrix that identifies the missing position of the data, and the covariate C is the vector of the external variable circuit breaker state and weather encoding. The mask M (missing position) and covariate C (such as circuit breaker state) are expanded along the spatial dimension (number of nodes N) to the same dimension N×T as the spatiotemporal feature Z to ensure that the missing state and external factors of each node are independently encoded. This application forms a closed loop from data cleaning to physical constraints, taking into account both data quality and physical laws.

[0031] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for intelligent filling of power flow data based on time-space coupling, characterized in that: The steps include: S1. Preprocess the raw data by Hampel filtering, phase alignment and normalization; S2, through the joint encoding of adaptive graph convolution Adaptive GCN and attention LSTM, a spatiotemporal encoder is constructed to capture the spatiotemporal coupling feature Z; S3. Design a multi-physics constraint generative adversarial network Phy-GAN through generator physical constraints and multi-scale discriminator; S4. Expand the feature dimensions of the missing mask M and the covariate C so that they have the same spatial dimension as the spatiotemporal coupling feature Z. Then concatenate the expanded M, C and Z as the generator input of the generative adversarial network Phy-GAN, and achieve data filling through the discriminator and adversarial training.

2. According to claim 1, a method for intelligently filling in power flow data based on time-space coupling is characterized in that: The S1 step performs phase alignment on the original data, and the formula is: ,in, represents the continuous phase after expansion, Indicates the current phase. represents the phase angle at the previous moment, Represents the rounding function.

3. The method for intelligently filling in power flow data based on time-space coupling according to claim 1 is characterized in that: In the step S2, the attention LSTM introduces the temporal attention mechanism into the LSTM. The spatiotemporal encoder performs gated feature fusion on the spatial feature matrix obtained by the adaptive graph convolution Adaptive GCN and the temporal feature matrix obtained by the attention LSTM, sets the gating weights, and obtains the spatiotemporal coupling features: ,in, is the spatiotemporal coupling feature, is the gating weight, is the trainable fusion weight matrix, is the spatial feature matrix, is the time feature matrix, Indicates the multiplication of corresponding positions of matrices, Represents the Sigmoid activation function.

4. The method for intelligently filling in power flow data based on time-space coupling according to claim 1 is characterized in that: In the S3 step, the multi-physics constraint generative adversarial network adds a physical constraint layer after the generator output layer, including dynamic constraints and algebraic constraints.

5. The method for intelligently filling in power flow data based on time-space coupling according to claim 1 is characterized in that: In the step S3, the multi-scale discriminator of the multi-physics constraint generative adversarial network is: ,in, Score the authenticity of the data, X represents the input data, s represents the downsampling scale, , represents the discriminant subnetwork of the corresponding scale, represents average pooling.

6. The method for intelligently filling in power flow data based on time-space coupling according to claim 1 is characterized in that: The missing mask M in step S4 is a binary matrix that identifies the location of missing data, and the covariate C is a vector of external variables, the circuit breaker status, and the weather encoding.

7. The method for intelligently filling in power flow data based on time-space coupling according to claim 4 is characterized in that: The dynamic constraints and algebraic constraints are: , ,in, represents the node voltage vector, represents the branch current vector, represents the node admittance matrix, represents the conjugate complex number of the branch current vector, represents taking the real part of a complex number, Indicates the actual measured active power value.

Images