Dynamic state estimation method, system and equipment based on power distribution network and medium

Abstract

The invention discloses a dynamic state estimation method, system and device based on a power distribution network and a medium. The method comprises the steps of obtaining initial state variable data of power distribution network equipment; performing time-frequency analysis and feature extraction on the initial state variable data to obtain state variable data, constructing an objective function based on the state variable data, performing initial state prediction of the power distribution network based on the objective function, and obtaining a first prediction value of the state of the power distribution network; and according to the first predicted value, calculating a noise parameter of a dynamic process, judging whether the noise parameter exceeds a preset threshold value, and dynamically updating the noise parameter to obtain a second predicted value of the state of the power distribution network. The method comprises the following steps: performing data fusion on preprocessed state variable data; performing initial dynamic state prediction on the fused data; calculating a noise parameter of a dynamic process based on the first predicted value, judging whether the noise parameter exceeds a preset threshold, and dynamically updating the noise parameter to obtain a second predicted value; and according to the first prediction value, the noise parameter in the dynamic process is calculated, quantitative evaluation of the prediction model error and dynamic updating of the noise parameter are realized, and the precision and reliability of the dynamic state estimation value are improved.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, and in particular to a method, system, device and medium for dynamic state estimation based on distribution networks. Background Technology

[0002] Distribution network state estimation is a key step in achieving distribution network situational awareness. By utilizing the correlation and redundancy of measurement data and combining it with computer data processing technology, it aims to improve the reliability and integrity of data, thereby effectively estimating system operating parameters and obtaining the real-time operating status of the system. Traditional state estimation methods are inadequate when facing the complex and ever-changing operating environment of modern distribution networks. Traditional measurement methods suffer from long sampling times and a lack of phasor measurements, which leads to numerous state estimation iterations and slow result updates, making it difficult to meet the requirements of rapidly changing distribution network situational awareness. Traditional methods often assume that the system model is known and invariant, ignoring the correlation between data from different sources, resulting in poor performance in handling uncertainties and noise interference in dynamic processes.

[0003] Despite significant advancements in existing technologies, some shortcomings remain. Traditional methods have limited capabilities in handling nonlinear dynamic processes and cannot fully capture the complex dynamic characteristics of distribution networks. For large-scale distribution networks, the sheer volume of computation may prevent existing technologies from meeting real-time requirements. In terms of noise parameter estimation, traditional methods typically employ fixed or static noise models that fail to adapt to actual dynamic changes, thus affecting the accuracy of state estimation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a dynamic state estimation method based on distribution networks to address the limitations of traditional methods in handling nonlinear dynamic processes, which cannot fully capture the complex dynamic characteristics of distribution networks. For large-scale distribution networks, due to the huge amount of computation, existing technologies may not be able to meet the real-time requirements. In terms of noise parameter estimation, traditional methods usually use fixed or static noise models, which fail to adjust according to actual dynamic changes, affecting the accuracy of state estimation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a dynamic state estimation method based on a distribution network, comprising:

[0008] Obtain initial state variable data of power distribution network equipment;

[0009] Time-frequency analysis and feature extraction are performed on the initial state variable data to obtain state variable data. Based on the state variable data, an objective function is constructed. Based on the objective function, the initial state of the distribution network is predicted to obtain the first predicted value of the distribution network state.

[0010] Based on the first predicted value, the noise parameters of the dynamic process are calculated, it is determined whether the noise parameters exceed a preset threshold, and the noise parameters are dynamically updated to obtain the second predicted value of the distribution network status.

[0011] As a preferred embodiment of the dynamic state estimation method based on distribution networks described in this invention, the method involves: performing time-frequency analysis and feature extraction on the initial state variable data to obtain state variable data, including:

[0012] By performing Fourier transform and differentiation on the initial state variable data, the frequency range abrupt change characteristics of the initial state variable data are obtained;

[0013] Based on the adjusted frequency range and abrupt change characteristics, time-frequency analysis and feature extraction are performed to obtain the fused state variable data.

[0014] As a preferred embodiment of the dynamic state estimation method based on distribution networks described in this invention, the method includes: constructing an objective function based on the state variable data, predicting the initial state of the distribution network based on the objective function, and obtaining a first predicted value of the distribution network state, comprising:

[0015] Calculate the weights of different state variables based on the state variable data, construct the objective function, solve and update the objective function to obtain the dynamic state variables;

[0016] Based on the initial state variable data of the distribution network equipment, a dynamic characteristic matrix is ​​generated. Combined with the dynamic state variables, a distribution network state prediction model is constructed. The covariance matrix and the first predicted value of the distribution network state are obtained by making predictions through the distribution network state prediction model.

[0017] The beneficial effects of this preferred technical solution are as follows: by performing Fourier transform and differentiation on the initial state variable data, the frequency range and abrupt change characteristics are accurately captured. Combined with time-frequency analysis and feature extraction, data fusion is achieved, effectively integrating key information in multi-source heterogeneous data, reducing noise interference and preserving dynamic characteristics.

[0018] As a preferred embodiment of the dynamic state estimation method based on a distribution network according to the present invention, the following steps are included: calculating noise parameters of the dynamic process based on the first predicted value, and determining whether the noise parameters exceed a preset threshold:

[0019] Based on the first predicted value of the distribution network state and the predicted covariance matrix, cubic points are generated, and the cubic points are predicted by the distribution network state prediction model. The state mean and the predicted covariance matrix are calculated.

[0020] Substitute the first predicted value into the measurement function to obtain the actual observed value, and calculate the measurement residual by the difference method;

[0021] Based on the state mean, the predicted covariance matrix, and the measurement residuals, the process noise covariance matrix is ​​dynamically adjusted, and the noise parameters of the dynamic process are calculated.

[0022] A semi-qualitative determination is performed on the adjusted process noise covariance matrix: if the semi-qualitative condition is met, it is output to the prediction covariance matrix; if not, the measurement error covariance matrix is ​​calculated based on historical measurement errors, the process noise covariance matrix is ​​further adjusted, and then output to the prediction covariance matrix.

[0023] The beneficial effects of this preferred technical solution are as follows: Based on the first predicted value and the covariance matrix, cubic points are generated. Through the nonlinear mapping of the cubic points by the prediction model, comprehensive sampling of the state space is achieved, improving the accuracy of dynamic process simulation. By dynamically adjusting the process noise covariance matrix using measurement residuals and ensuring the mathematical stability of the matrix through semi-qualitative judgment, the shortcomings of traditional static noise models in adapting to dynamic changes in the distribution network are solved. This makes the noise parameter estimation more closely match actual operating conditions and reduces prediction bias caused by a fixed noise model.

[0024] As a preferred embodiment of the dynamic state estimation method based on distribution networks described in this invention, the method includes: dynamically updating the noise parameters to obtain a second predicted value of the distribution network state, comprising:

[0025] Based on the measurement residuals, calculate the residual influence term caused by the measurement residuals; based on the prediction covariance matrix, obtain the change in the prediction covariance; based on the change and the residual influence term, update the process noise covariance matrix;

[0026] When the residual influence term is less than or equal to the preset threshold, it is determined to be a stable state, and the second predicted value of the distribution network state is output.

[0027] When the residual influence term exceeds the preset threshold, it is determined to be an unstable state and enters the dynamic update stage. The updated process noise covariance matrix is ​​output and fed back to the prediction covariance matrix until the residual influence term is less than or equal to the preset threshold. Then, the second predicted value of the distribution network state is output.

[0028] The beneficial effects of this preferred technical solution are as follows: By dynamically updating the noise covariance matrix through residual influence terms and predicted covariance changes, real-time response to non-stationary states is achieved. When system fluctuations exceed a threshold, iterative optimization ensures that the state estimate always converges to the true value, guaranteeing estimation efficiency under stationary conditions while improving robustness under complex dynamic scenarios. This enhances the accuracy and reliability of distribution network state estimation, meeting the dual requirements of real-time performance and accuracy for large-scale distribution networks.

[0029] As a preferred embodiment of the dynamic state estimation method based on distribution networks described in this invention, the second predicted value of the distribution network state includes:

[0030] Build a visual interface to display and store the second predicted value of the distribution network status;

[0031] Users who have passed real-name verification are allowed to view this information;

[0032] The state variable data, the first predicted value of the distribution network state, and the second predicted value of the distribution network state are stored in the central database. The central database is sorted in chronological order, labeled with corresponding tags, and the integrity of the backup data is checked periodically.

[0033] As a preferred embodiment of the dynamic state estimation method based on distribution networks described in this invention, the initial state variable data of distribution network equipment is obtained, including:

[0034] The topology information and initial state variable data of the distribution network are collected using smart sensors and SCADA systems. The SCADA system includes RTU and FTU units, the sensors include PMU and AMI sensors, and the state variable data includes current and voltage data.

[0035] The collected initial state variable data is preprocessed, including using statistical methods to identify and remove outliers, performing time alignment on the state variable data, and normalizing the state variable data.

[0036] Secondly, the present invention provides a dynamic state estimation system based on a power distribution network, comprising: a data acquisition module for acquiring initial state variable data of power distribution network equipment;

[0037] The initial prediction module is used to perform time-frequency analysis and feature extraction on the initial state variable data to obtain state variable data, construct an objective function based on the state variable data, predict the initial state of the distribution network based on the objective function, and obtain the first predicted value of the distribution network state.

[0038] The prediction and optimization module is used to calculate the noise parameters of the dynamic process based on the first prediction value, determine whether the noise parameters exceed a preset threshold, and dynamically update the noise parameters to obtain the second prediction value of the distribution network status.

[0039] Thirdly, the present invention provides an electronic device, comprising:

[0040] Memory and processor;

[0041] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of a dynamic state estimation method based on a power distribution network.

[0042] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the dynamic state estimation method based on the power distribution network.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention collects data and preprocesses it, then fuses the preprocessed state variable data; performs initial dynamic state prediction on the fused data; calculates noise parameters of the dynamic process based on the first predicted value, determines whether the noise parameters exceed a preset threshold, and dynamically updates the noise parameters to obtain a second predicted value; calculates noise parameters in the dynamic process based on the first predicted value, thereby realizing a quantitative assessment of the prediction model error and a dynamic update of the noise parameters, reducing prediction errors and improving the accuracy and reliability of dynamic state estimates. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a schematic diagram of the overall process of a dynamic state estimation method based on a power distribution network according to an embodiment of the present invention. Detailed Implementation

[0046] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0047] Example 1, referring to Figure 1 As one embodiment of the present invention, a dynamic state estimation method based on a distribution network is provided, comprising:

[0048] S100: Obtain initial state variable data of power distribution network equipment;

[0049] S101: Perform time-frequency analysis and feature extraction on the initial state variable data to obtain the state variable data. Based on the state variable data, construct an objective function. Based on the objective function, predict the initial state of the distribution network and obtain the first predicted value of the distribution network state.

[0050] S102: Based on the first predicted value, calculate the noise parameters of the dynamic process, determine whether the noise parameters exceed a preset threshold, and dynamically update the noise parameters to obtain the second predicted value of the distribution network status.

[0051] It should be noted that this invention obtains the final predicted value by acquiring initial state variable data, data fusion and initial prediction, and calculating and dynamically updating noise parameters. Traditional methods have limited ability to handle nonlinear dynamic processes in distribution networks, struggle to capture complex dynamic characteristics, and suffer from high computational demands, making them unsuitable for real-time performance. Furthermore, static noise models cannot dynamically adjust with actual conditions, affecting estimation accuracy. This invention integrates key information through data fusion, combining initial prediction with dynamic noise parameter updates. This solves the problems of feature redundancy or loss in traditional data processing and allows noise parameters to adapt to system changes, ultimately improving the accuracy, real-time performance, and reliability of distribution network state estimation, enabling a more precise reflection of the real-time operating status of the distribution network.

[0052] Example 2, refer to Figure 1 This is one embodiment of the present invention. Based on the above embodiment, a dynamic state estimation method for power distribution networks is provided.

[0053] In this embodiment of the invention, obtaining the initial state variable data of the power distribution network equipment in step S100 specifically includes:

[0054] Data collection and preprocessing refers to the use of smart sensors and SCADA systems to collect topology information and state variable data of the distribution network; SCADA systems include RTU and FTU units; sensors include PMU and AMI sensors; state variable data includes current and voltage data;

[0055] The collected state variable data is preprocessed; the preprocessing includes using statistical methods to identify and remove outliers, aligning the state variable data to time, and normalizing the state variable data.

[0056] It should be noted that by deploying smart sensors, such as PMUs and AMIs, and SCADA systems including RTUs and FTUs, comprehensive monitoring of the distribution network structure and operating status is achieved. This step enables the real-time acquisition of key parameters such as current and voltage with high precision. By applying advanced statistical analysis techniques to the collected state variable data to identify and eliminate outliers, data quality is effectively improved, removing inaccurate data points caused by sensor errors or sudden events. By implementing a time alignment algorithm, data from different sources are adjusted to a unified time reference, achieving seamless integration of multi-source heterogeneous data. This step eliminates timing deviations caused by differences in sampling frequencies, ensuring consistency and correlation between different types of data. By performing normalization operations on all state variable data, the numerical range is standardized, enabling physical quantities of different magnitudes to be compared on the same scale. This simplifies the mathematical modeling process and improves model training efficiency, ultimately enhancing the model's generalization ability and improving prediction accuracy.

[0057] In this embodiment of the invention, step S101 involves performing time-frequency analysis and feature extraction on the initial state variable data to obtain state variable data, constructing an objective function based on the state variable data, predicting the initial state of the distribution network based on the objective function, and obtaining the first predicted value of the distribution network state. It also includes sub-steps A1-A2:

[0058] A1: Calculate the weights of different state variables based on the state variable data, construct the objective function, solve and update the objective function to obtain the dynamic state variables;

[0059] A2: Based on the initial state variable data of the distribution network equipment, a dynamic characteristic matrix is ​​generated. Combined with the dynamic state variables, a distribution network state prediction model is constructed. The covariance matrix and the first predicted value of the distribution network state are obtained through the distribution network state prediction model.

[0060] In an embodiment of the present invention, the process of obtaining state variable data includes:

[0061] The initial state variable data after preprocessing is subjected to Fourier transform to obtain the feature spectrum. The main frequency features in the feature spectrum are identified by a bandpass filter to obtain the frequency range of the state variable data.

[0062] Based on the topology information of the distribution network, the three-phase admittance matrix Y is constructed using the node analysis method, and the frequency threshold α is set using the spectrum analysis method. Frequencies higher than or equal to the threshold α are high frequencies, and frequencies lower than the threshold α are low frequencies.

[0063] Based on the three-phase admittance matrix Y and the threshold α, the high-frequency and low-frequency signals are identified using the Fourier transform method. The results of identifying the high-frequency and low-frequency signals are compared with the characteristic spectrum, and the frequency range is supplemented by the union of the comparison results.

[0064] By utilizing the distribution network topology information, a three-phase admittance matrix Y was constructed using the node analysis method. This not only integrated the network structure information but also considered the electrical connection characteristics between each node. Based on the three-phase admittance matrix Y, Fourier transform technology was used again to distinguish between high-frequency and low-frequency components, further refining the previously obtained spectrum information. By comparing the old and new spectra, the frequency range was adjusted to better adapt to actual operating conditions, ensuring the rationality of the frequency range setting.

[0065] The abrupt change characteristics are obtained by differentiating the preprocessed state variable data; based on the adjusted frequency range and abrupt change characteristics, the wavelet basis function DB4 is selected, including the input layer, decomposition layer and output layer.

[0066] The input layer of the wavelet basis function DB4 is set to the preprocessed state variable data.

[0067] The decomposition layer uses the wavelet transform tool PyWavelets to perform wavelet decomposition on the voltage and current data in each preprocessed state variable data, as shown in the following formula:

[0068]

[0069]

[0070] in, For voltage data, For the wavelet detail coefficients of the j-th layer of voltage data, These are the approximation coefficients for the J-th layer of voltage data. For current data, For the wavelet detail coefficients of the j-th layer of the current data, These are the approximation coefficients for the J-th layer of the current data;

[0071] Record the wavelet detail coefficients and approximation coefficients of the voltage and current data of the decomposition layer.

[0072] The output layer uses a linear combination method to merge the wavelet detail coefficients of voltage and current into a single detail coefficient. ;

[0073] For example, based on the wavelet detail coefficients as Calculate data energy The formula is as follows:

[0074]

[0075] For example, noise energy is estimated using the difference method based on wavelet detail coefficients. The formula is as follows:

[0076]

[0077] in, To estimate wavelet detail coefficients;

[0078] For example, based on data energy and noise energy The signal-to-noise ratio (SNR) of state variable data is calculated using the direct energy ratio method, as shown in the following formula:

[0079]

[0080] For example, based on state variable data, the signal-to-noise ratio (SNR) of the objective function is constructed, and the weights of the state variable data are calculated using the energy ratio method. The formula is as follows:

[0081]

[0082] in, This represents the total signal-to-noise ratio. Let be the signal-to-noise ratio of the i-th signal.

[0083] Construct the weights of the objective function based on the state variable data. The detail coefficients of all state variable data are weighted and fused to obtain the weighted fused detail coefficients. The weighted fusion detail coefficients were processed using the inverse wavelet transform tool PyWavelets. Perform an inverse transformation to obtain the fused state variable data R(t), where R(t) contains voltage data and current data;

[0084] The main frequency characteristics include fundamental frequency, harmonics, peak frequency, bandwidth and noise frequency, and abrupt change characteristics include slope change, spike and abrupt change amplitude.

[0085] In one alternative implementation, the empirical mode decomposition (EMD) method can be used to process the state variable data. Specifically, the preprocessed voltage and current data are subjected to EMD decomposition to obtain several intrinsic mode functions (IMFs) and residual components. IMF components containing major frequency characteristics (such as fundamental frequency and harmonics) are selected by energy proportion. Combined with abrupt change characteristics (such as the instantaneous energy change of IMF corresponding to a peak), the selected components are weighted and reconstructed (the weights are calculated based on the signal-to-noise ratio of each component) to obtain the fused state variable data.

[0086] In another alternative implementation, Kalman filtering combined with multi-sensor data fusion can be used to process state variable data. Specifically, a multi-sensor observation model is constructed based on the distribution network topology, and voltage and current data collected by sensors at different locations are used as observations. The multi-source data is fused through the prediction-update process of Kalman filtering, wherein the observation noise covariance matrix is ​​dynamically adjusted according to the signal-to-noise ratio of each sensor (the higher the signal-to-noise ratio, the greater the weight). Finally, the fused state variable data and the corresponding covariance matrix are output.

[0087] It should be noted that the present invention selects the suitable wavelet basis function DB4, which provides a flexible and efficient data processing framework suitable for various application scenarios, ultimately improving the flexibility and efficiency of data processing and laying the foundation for subsequent advanced analysis.

[0088] It should also be noted that the inverse wavelet transform tool PyWavelets is used to perform an inverse transform on the weighted fusion detail coefficients to obtain the fused state variable data R(t). Through inverse wavelet transform, the weighted fusion detail coefficients are restored to the state variable data R(t) in the time domain. This process preserves the main features of the original data while removing unnecessary noise interference. Its role is to achieve a perfect regression from the frequency domain to the time domain, ensuring the consistency and coherence of the data before and after processing, and ultimately achieving the effect of restoring high-quality data, providing reliable data support for practical applications.

[0089] In this embodiment of the invention, obtaining the covariance matrix and the first predicted value of the distribution network state includes:

[0090] Define the merged state variable data R(t) as the static state variable. Based on the relationship between the voltage across the terminals and the branch current in the topological information, the nonlinear relationship between the branch current and the voltage across the terminals is linearized by Taylor expansion to obtain the voltage and current scaling factors a and b.

[0091] For example, the measurement function is obtained based on circuit theory. The formula is as follows:

[0092]

[0093] in, For voltage, For current;

[0094] Measurement functions are defined based on circuit theory. This has enabled the establishment of a bridge between physical measurements and mathematical models;

[0095] For example, based on measurement functions Construct the static state estimation equation, as shown in the following formula:

[0096]

[0097] in, For observation purposes, This is the measurement error vector;

[0098] Based on the static state estimation equation, the objective function is constructed using the least squares method. The formula is as follows:

[0099]

[0100] Where i is the index of the sample, and L is the total number of samples. For observations The variance;

[0101] The objective function was constructed using the least squares method, achieving the best fit to the observed data. The power P of the state variable data R(t) was calculated using the physical power formula, as follows:

[0102]

[0103] Calculate the joint probability density function f(V) based on the power P. I P), the formula is as follows:

[0104]

[0105] in, The normalization constant is The average voltage. The average value of the current. The average power. This is the square of the standard deviation of the voltage. This is the square of the standard deviation of the current. It is the square of the power standard deviation.

[0106] According to the joint probability density function f(V) I P), calculate the weighting factor of the observation. The formula is as follows:

[0107]

[0108] Based on weighting factors Construct the update objective function The formula is as follows:

[0109]

[0110] Solve iteratively using the Gauss-Newton method. ,Will The solution is used as pseudo-measurement data and fused into the state variable data R(t) using the Kalman filter method to obtain the dynamic state variable. Based on the topology information of the distribution network, the dynamic characteristic matrix is ​​generated using the KVL dynamic equation method, and the state transition matrix is ​​obtained by discretizing the dynamic characteristic matrix using the matrix exponentiation method. .

[0111] Based on dynamic state variable x and state transition matrix A power distribution network state prediction model is constructed, and the formula is as follows:

[0112]

[0113] in, Let k be the state variable at time k. For process noise;

[0114] Based on process noise Define the covariance matrix =E[ ], where E[.] is the expected value operation.

[0115] In one optional implementation, the Extended Kalman Filter (EKF) method can be used to obtain the first predicted value of the distribution network state. Specifically, based on the fused state variable data R(t), a nonlinear state equation is directly constructed (without Taylor expansion linearization). Through the prediction step (calculating the prior state estimate based on the state transition matrix Φ) and update step (correcting the state using the residual between the measurement function h(∙) and the observation z), the dynamic state variable x is iteratively solved. The initial value of the process noise covariance matrix is ​​set based on historical data statistics, and the observation weights are adaptively adjusted through the error covariance matrix. Finally, the first predicted value of the distribution network state and the corresponding covariance matrix are output.

[0116] In another alternative implementation, the particle filtering (PF) method can be used to obtain the first predicted value of the distribution network state. Specifically, a large number of particles (representing possible state variable values) are generated based on the distribution network topology information. The particles are updated based on the state transition matrix Φ. The weight of each particle is calculated by combining the measurement function h (∙) (the weight is positively correlated with the joint probability density function f (V,I,P)). High-weight particles are retained by resampling. After iteration, the weighted mean of the particle set is obtained as the first predicted value of the distribution network state, and the variance matrix of the particle set is used as the covariance matrix.

[0117] It should be noted that by constructing a joint probability density function f(V, I, P), a comprehensive characterization of the statistical relationship between voltage, current, and power is achieved. By calculating the weighting factors of the observed quantities, the importance of different observed data is assessed, enhancing the model's robustness to outliers and improving the precision of data analysis. By constructing an update objective function and using the Gauss-Newton method for iterative solution, the dynamic response speed is improved while ensuring prediction accuracy. By using the solution results as pseudo-measurement data and integrating them into the state variable data R(t) using the Kalman filter method, the original dataset is dynamically expanded, promoting continuous improvement and optimization. Ultimately, this strengthens the data update mechanism and improves the system's adaptive capability. By constructing a state prediction equation, a forward-looking estimate of the future system state is achieved. By defining the covariance matrix to express the statistical characteristics of process noise, the uncertainty is quantified, providing the necessary input for the state prediction equation.

[0118] In this embodiment of the invention, step S102, which calculates the noise parameters of the dynamic process based on the first predicted value, determines whether the noise parameters exceed a preset threshold, and dynamically updates the noise parameters to obtain the second predicted value of the distribution network status, further includes sub-steps B1-B5:

[0119] B1: Based on the first predicted value of the distribution network state and the predicted covariance matrix, generate cubic points, predict the cubic points through the distribution network state prediction model, and calculate the state mean and the predicted covariance matrix.

[0120] B2: Substitute the first predicted value into the measurement function to obtain the actual observed value, and calculate the measurement residual by the difference method;

[0121] B3: Based on the state mean, the predicted covariance matrix, and the measurement residuals, dynamically adjust the process noise covariance matrix and calculate the noise parameters of the dynamic process; perform a semi-qualitative judgment on the adjusted process noise covariance matrix: if it meets the semi-qualitative requirements, output it to the predicted covariance matrix; if it does not meet the requirements, calculate the measurement error covariance matrix based on the historical measurement errors, further adjust the process noise covariance matrix, and output it to the predicted covariance matrix.

[0122] B4: Based on the measurement residuals, calculate the residual influence term caused by the measurement residuals; based on the prediction covariance matrix, obtain the change in the prediction covariance; based on the change and the residual influence term, update the process noise covariance matrix;

[0123] B5: When the residual influence term is less than or equal to the preset threshold, it is determined to be a stable state, and the second predicted value of the distribution network state is output; when the residual influence term is greater than the preset threshold, it is determined to be a non-stable state, and enters the dynamic update stage, outputting the updated process noise covariance matrix and feeding it back into the prediction covariance matrix, until the residual influence term is less than or equal to the preset threshold, and the second predicted value of the distribution network state is output.

[0124] In this embodiment of the invention, calculating the noise parameters of the dynamic process includes:

[0125] Current covariance matrix based on state prediction and current state prediction value Cubic points are generated through Cholesky decomposition. ;

[0126] Cholesky decomposition based on the current covariance matrix and the current state prediction value generates cubic points, which realizes effective sampling of the uncertainty distribution and constructs a set of sample points representing different possibilities in the state space. This ensures the accuracy and reliability of the subsequent nonlinear mapping process. These cubic points are then used to simulate the state change at the next moment, providing a multi-angle observation perspective for the dynamic process. Ultimately, this improves the accuracy of state estimation and enhances the robustness of the model.

[0127] Define the initial process noise covariance matrix Given the values, use the state equation for the cubic points. Perform a nonlinear mapping to predict the state at the next time step. and cubic points State prediction values ​​are labeled as historical data, based on all cube points in the historical data. Calculation of state prediction values ​​and state mean and the predicted covariance matrix The formula is as follows:

[0128]

[0129]

[0130] in, = The uniform weights of the cube points are n, the dimension of the state variables are n, and the time steps are k.

[0131] Predict the current state value Substitute into the measurement function Obtain actual observation values The measurement residual is obtained by subtracting the predicted value and the actual observed value of the current state using the difference method. The formula is as follows:

[0132]

[0133] The process noise covariance matrix is ​​dynamically adjusted using the robust unbiased NSE formula. The formula is as follows:

[0134]

[0135] in, To provide the robust, unbiased NSE dynamically adjusted process noise covariance, Forgetting factor, To measure the residual, This is the Kalman gain.

[0136] when = It was determined to satisfy the semi-qualitative condition, and... Output to the prediction covariance matrix ;

[0137] when ≠ The result was determined to be non-qualitative, and the measurement error covariance matrix of the measurement error in the historical data was calculated using MATLAB.

[0138] For example, a biased NSE formula is used to dynamically adjust the process noise covariance. The formula is as follows:

[0139]

[0140] in, The process noise covariance after biased NSE dynamic adjustment. The measurement error covariance matrix, To measure the correction term for process noise, Predicting covariance matrix .

[0141] It should be noted that the robust unbiased NSE formula is introduced, and the process noise covariance matrix is ​​dynamically adjusted. Combined with parameters such as forgetting factor, measurement residual and Kalman gain, it aims to adapt to changes in the internal and external systems, and ensure good performance under different operating conditions. This achieves the effect of enhancing model adaptability and reducing prediction risk. By performing a semi-qualitative judgment on the adjusted process noise covariance matrix to determine whether it meets the semi-qualitative conditions, the mathematical properties of the covariance matrix are ensured, and numerical instability is avoided. At the same time, a backup scheme of the biased NSE formula is provided to deal with complex situations. Ultimately, the effect of ensuring numerical stability and improving the robustness of the algorithm is achieved.

[0142] It should also be noted that the predicted state cube points Substitute into the measurement function Mapping is performed to obtain the predicted measurement value. Predicted measurement values Marked as historical forecast data, based on all forecast measurements in the historical forecast data. Calculate the state mean and measurement error covariance matrix ;

[0143] Calculate the cross-covariance matrix between state and measurement. The formula is as follows:

[0144]

[0145] based on and Calculate the new Kalman gain The formula is as follows:

[0146]

[0147] Bundle The updated formulas for the state equation and predicted covariance are as follows:

[0148]

[0149]

[0150] in, This is the updated dynamic state estimate. This is the updated predicted covariance.

[0151] In this embodiment of the invention, obtaining the second predicted value of the distribution network status by dynamically updating noise parameters includes:

[0152] Define the dynamic update phase, including based on Using measurement equations The updated observation values ​​were obtained. '; Using the predicted state estimates and updated predicted measurements 'Calculate measurement residuals' ';

[0153] Measuring residuals As input, calculate the residual effect term caused by the measurement residual. The formula is as follows:

[0154]

[0155] Using the updated predicted covariance Covariance of predictions from the previous time step Calculate the change in predicted covariance Based on the change in predicted covariance Δ and residual influence terms The process noise covariance matrix is ​​updated using the robust unbiased noise statistical estimator NSE formula. The formula is as follows:

[0156]

[0157] A threshold τ is set for the predicted covariance using the statistical distribution method, and the change Δ in the predicted covariance is compared. Compare with the threshold τ: when If the value is less than or equal to τ, the state is considered stationary, and the dynamic state estimate is output. ;when If the value is greater than τ, the system is determined to be in a non-stationary state and enters the dynamic update phase, outputting the updated process noise covariance matrix. Feeding it back into the prediction covariance matrix until satisfied. If ≤τ, output the second predicted value.

[0158] In one optional implementation, an adaptive Kalman filter (AKF) method can be used to obtain the second predicted value of the distribution network state. Specifically, the process noise covariance matrix Q and the measurement noise covariance matrix R of the Kalman filter are dynamically adjusted based on the statistical characteristics (such as mean and variance) of the measurement residual ε_(k+1)'. When the residual fluctuation is small, the weight of Q is reduced to enhance the filter stability; when the residual changes abruptly, the weight of Q is increased to improve the response speed to dynamic changes. By updating Q and R in real time, the state estimate is iteratively optimized, and finally, the second predicted value of the distribution network state is output when the residual influence term is ≤ the threshold τ.

[0159] In another alternative implementation, a second predicted value of the distribution network state can be obtained using an unscented Kalman filter (UKF) combined with a sliding window method. Specifically, a sliding window (with a window size set based on the dynamic characteristics of the distribution network) is introduced when generating cubic points. Only recent historical data within the window is used to calculate the cubic point weights, enhancing the ability to capture the latest state changes. The predicted value is obtained through the nonlinear mapping of the cubic points using UKF. The process noise covariance matrix is ​​updated by combining the residual influence term and the covariance change within the window. When the residual influence term within the window stabilizes within a threshold τ, the second predicted value of the distribution network state is output.

[0160] In this embodiment of the invention, a visual interface is constructed to display and store the dynamic state estimation values;

[0161] In one alternative implementation, a visual interface is built using the front-end framework React.js, and a data visualization tool is used to display the dynamic status data changes of the power distribution network in real time, allowing users who have been verified with their real names to access the data; the collected status data, dynamic predictions and estimated data are stored in a central database, and the central database is sorted in chronological order, labeled with corresponding tags, and the integrity of the backup data is checked periodically.

[0162] Example 3 illustrates a schematic scheme for a dynamic state estimation method based on a distribution network. It should be noted that the technical solution of this distribution network-based dynamic state estimation system belongs to the same concept as the aforementioned distribution network-based dynamic state estimation method. Details not described in detail in this embodiment can be found in the description of the aforementioned distribution network-based dynamic state estimation method.

[0163] This embodiment also provides a dynamic state estimation system based on a distribution network, including:

[0164] The data acquisition module is used to acquire the initial state variable data of the power distribution network equipment;

[0165] The initial prediction module is used to perform time-frequency analysis and feature extraction on the initial state variable data to obtain the state variable data. Based on the state variable data, an objective function is constructed, and the initial state of the distribution network is predicted based on the objective function to obtain the first predicted value of the distribution network state.

[0166] The prediction and optimization module is used to calculate the noise parameters of the dynamic process based on the first prediction value, determine whether the noise parameters exceed a preset threshold, and dynamically update the noise parameters to obtain the second prediction value of the distribution network status.

[0167] This embodiment also provides an electronic device suitable for dynamic state estimation based on a power distribution network, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the dynamic state estimation method based on a power distribution network as proposed in the above embodiment.

[0168] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the dynamic state estimation method based on the power distribution network as proposed in the above embodiments.

[0169] The storage medium proposed in this embodiment and the method for implementing dynamic state estimation based on power distribution network proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0170] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0171] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A dynamic state estimation method based on a distribution network, characterized in that, include: Obtain initial state variable data of power distribution network equipment; Time-frequency analysis and feature extraction are performed on the initial state variable data to obtain state variable data. Based on the state variable data, an objective function is constructed. Based on the objective function, the initial state of the distribution network is predicted to obtain the first predicted value of the distribution network state. Based on the first predicted value, the noise parameters of the dynamic process are calculated, it is determined whether the noise parameters exceed a preset threshold, and the noise parameters are dynamically updated to obtain the second predicted value of the distribution network status.

2. The dynamic state estimation method based on distribution network as described in claim 1, characterized in that, Time-frequency analysis and feature extraction are performed on the initial state variable data to obtain state variable data, including: By performing Fourier transform and differentiation on the initial state variable data, the frequency range abrupt change characteristics of the initial state variable data are obtained; Based on the adjusted frequency range and abrupt change characteristics, time-frequency analysis and feature extraction are performed to obtain the fused state variable data.

3. The dynamic state estimation method based on distribution network as described in claim 2, characterized in that, Based on the state variable data, an objective function is constructed, and the initial state of the distribution network is predicted based on the objective function to obtain the first predicted value of the distribution network state, including: Calculate the weights of different state variables based on the state variable data, construct the objective function, solve and update the objective function to obtain the dynamic state variables; Based on the initial state variable data of the distribution network equipment, a dynamic characteristic matrix is ​​generated. Combined with the dynamic state variables, a distribution network state prediction model is constructed. The covariance matrix and the first predicted value of the distribution network state are obtained by making predictions through the distribution network state prediction model.

4. The dynamic state estimation method based on distribution network as described in claim 3, characterized in that, Based on the first predicted value, calculate the noise parameters of the dynamic process, and determine whether the noise parameters exceed a preset threshold, including: Based on the first predicted value of the distribution network state and the predicted covariance matrix, cubic points are generated, and the cubic points are predicted by the distribution network state prediction model. The state mean and the predicted covariance matrix are calculated. Substitute the first predicted value into the measurement function to obtain the actual observed value, and calculate the measurement residual by the difference method; Based on the state mean, the predicted covariance matrix, and the measurement residuals, the process noise covariance matrix is ​​dynamically adjusted, and the noise parameters of the dynamic process are calculated. A semi-qualitative determination is performed on the adjusted process noise covariance matrix: if the semi-qualitative condition is met, it is output to the prediction covariance matrix; if not, the measurement error covariance matrix is ​​calculated based on historical measurement errors, the process noise covariance matrix is ​​further adjusted, and then output to the prediction covariance matrix.

5. The dynamic state estimation method based on distribution network as described in claim 4, characterized in that, And dynamically update the noise parameters to obtain the second predicted value of the distribution network status, including: Based on the measurement residuals, calculate the residual influence term caused by the measurement residuals; based on the prediction covariance matrix, obtain the change in the prediction covariance; based on the change and the residual influence term, update the process noise covariance matrix; When the residual influence term is less than or equal to the preset threshold, it is determined to be a stable state, and the second predicted value of the distribution network state is output. When the residual influence term exceeds the preset threshold, it is determined to be an unstable state and enters the dynamic update stage. The updated process noise covariance matrix is ​​output and fed back to the prediction covariance matrix until the residual influence term is less than or equal to the preset threshold. Then, the second predicted value of the distribution network state is output.

6. The dynamic state estimation method based on distribution network as described in claim 5, characterized in that, The second predicted value of the distribution network status includes: Build a visual interface to display and store the second predicted value of the distribution network status; Users who have passed real-name verification are allowed to view this information; The state variable data, the first predicted value of the distribution network state, and the second predicted value of the distribution network state are stored in the central database. The central database is sorted in chronological order, labeled with corresponding tags, and the integrity of the backup data is checked periodically.

7. The dynamic state estimation method based on distribution network as described in claim 5, characterized in that, Obtain initial state variable data for power distribution network equipment, including: The topology information and initial state variable data of the distribution network are collected using smart sensors and SCADA systems. The SCADA system includes RTU and FTU units, the sensors include PMU and AMI sensors, and the state variable data includes current and voltage data. The collected initial state variable data is preprocessed, including using statistical methods to identify and remove outliers, performing time alignment on the state variable data, and normalizing the state variable data.

8. A dynamic state estimation system based on a distribution network, using the method described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to acquire the initial state variable data of the power distribution network equipment; The initial prediction module is used to perform time-frequency analysis and feature extraction on the initial state variable data to obtain state variable data, construct an objective function based on the state variable data, predict the initial state of the distribution network based on the objective function, and obtain the first predicted value of the distribution network state. The prediction and optimization module is used to calculate the noise parameters of the dynamic process based on the first prediction value, determine whether the noise parameters exceed a preset threshold, and dynamically update the noise parameters to obtain the second prediction value of the distribution network status.

9. An electronic device, characterized in that, include: A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the dynamic state estimation method based on the power distribution network according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the steps of the dynamic state estimation method based on the power distribution network as described in any one of claims 1 to 7.

Images