CI and Bar-Shalog-Campo fused spatio-temporal data characterization method

Abstract

The invention belongs to the technical field of electric power, and relates to a spatio-temporal data characterization method based on CI and Bar-Shilom-Campo fusion, which lays a foundation for subsequent analysis by collecting and preprocessing data of a sensor, a PMU and an SCADA system; the dynamic change of the power grid state is captured by extracting time features (including a time sequence, a frequency domain and a time frequency feature); spatial features are extracted based on power grid topology and geographic position data, and understanding of power grid structure and environmental influence is enhanced. Time and space features are converted into evidences by utilizing a Dempster-Shafer theory, comprehensive belief measurement is calculated, the measurement integrates confidence intervals and probability distribution of all data sources, and the uncertainty of data is effectively processed; and finally, fusing the comprehensive belief metrics through a Bar-Shalog-Campo algorithm to realize accurate estimation of the state of the power grid. According to the method, the accuracy and the reliability of power grid state estimation are effectively improved by integrating spatial-temporal characteristics and evidence synthesis.

Description

Technical Field

[0001] This application belongs to the field of power technology, and more specifically, relates to a spatio-temporal data characterization method that integrates CI and Bar-Shalom-Campo. Background Art

[0002] In modern power grid systems, accurate state estimation and monitoring are crucial for ensuring the stability and safety of the power grid. To achieve this goal, the power grid system integrates a variety of data collection and monitoring technologies, mainly including:

[0003] Sensor network: Used to monitor parameters such as voltage, current, and power in real time. Sensor data provides the basic data for the operation of the power grid.

[0004] PMU (Phasor Measurement Unit): A high-precision device that provides phasor information (phase angle and amplitude) at various points in the power grid and can perform real-time measurements with high time resolution.

[0005] SCADA system: Provides the operating status of power grid equipment, switch actions, historical data, etc., for controlling and monitoring the operation of the power grid.

[0006] However, the core technical problem in power grid state estimation is how to effectively integrate information from these different data sources. Each data source provides information with different characteristics and uncertainties. How to combine these data to obtain an accurate power grid state estimation is a technical challenge. Summary of the Invention

[0007] The present invention provides a spatio-temporal data characterization method that integrates CI and Bar-Shalom-Campo, aiming to integrate different data to obtain an accurate power grid state estimation.

[0008] The spatio-temporal data characterization method that integrates CI and Bar-Shalom-Campo includes the following steps:

[0009] Step 1: Collect sensor data, PMU data, and SCADA system data in the power grid, and preprocess the collected data;

[0010] Step 2: Extract time features based on the preprocessed data to obtain time features, frequency domain features, and time-frequency features;

[0011] Step 3: Extract spatial features based on the power grid topology information, preprocessed sensor data, and geographical location data to obtain topology features and geographical features;

[0012] Step 4: Convert the results obtained from time feature extraction and spatial feature extraction into evidence. Based on the Dempster-Shafer theory, synthesize multi-source evidence, calculate the fused belief function, and obtain the comprehensive belief measure, where the comprehensive belief measure is the confidence interval and probability distribution of each state quantity obtained based on evidence synthesis;

[0013] Step 5: Based on the comprehensive belief measure, the results of time feature extraction, and the results of spatial feature extraction, perform multi-source signal fusion through the Bar-Shalom-Campo fusion algorithm to obtain the power grid state estimation after multi-source fusion.

[0014] The present invention effectively solves the problems of consistency, noise processing, and spatio-temporal feature integration in multi-source data fusion through the spatio-temporal data characterization method that combines CI and Bar-Shalom-Campo fusion; lays a foundation for subsequent analysis by collecting and preprocessing data from sensors, PMUs, and SCADA systems; captures the dynamic changes of the power grid state by extracting time features (including time series, frequency domain, and time-frequency features); enhances the understanding of the power grid structure and environmental impacts by extracting spatial features based on power grid topology and geographical location data; uses the Dempster-Shafer theory to convert time and spatial features into evidence and calculate the comprehensive belief measure, which synthesizes the confidence intervals and probability distributions of each data source and effectively handles data uncertainty; finally, fuses these comprehensive belief measures through the Bar-Shalom-Campo algorithm to achieve accurate estimation of the power grid state. This method effectively improves the accuracy and reliability of power grid state estimation through the integration of spatio-temporal features and evidence synthesis.

[0015] Preferably, the time feature, frequency domain feature, and time-frequency feature are obtained based on the following steps:

[0016] Time feature extraction: Calculate the mean, variance, peak value, and standard deviation of each type of data based on the preprocessed data; and perform autocorrelation analysis based on the calculated mean:

[0017]

[0018] where: x i represents the i-th data point in the time series; N represents the total number of data points in the time series; μ represents the mean; x i+τ represents the (i + τ)-th data point in the time series; τ represents the lag time or delay;

[0019] Frequency domain feature extraction: Based on the preprocessed data, use the fast Fourier transform to convert the time series to the frequency domain, and then perform power spectral density calculation and main frequency calculation:

[0020] P(f) = |X(f)|2;

[0021] Where: P(f) represents the power spectral density, indicating the power distribution at different frequencies; X(f) represents the fast Fourier transform;

[0022] Among them, the main frequency corresponds to the frequency of the maximum power frequency density;

[0023] Time-frequency characteristic data: Based on the preprocessed data, short-time Fourier transform is used to obtain time-frequency characteristics:

[0024]

[0025] Where: w(n - t) is the window function; t is the unknown time; f is the frequency; x(n) represents the nth data point in the time series, indicating the value at time point n; e -j2πfn represents the complex exponential function, which is used to transform each data point in the time series into the frequency domain; represents the summation load, indicating the summation of the signal values at all time points n and the weighted values of the window function.

[0026] Preferably, step 3 includes the following steps:

[0027] Spatial distribution analysis: Analyze the distribution of power grid equipment in the geographical space:

[0028]

[0029]

[0030] Where: d ij represents the distance between the i-th and j-th devices; x i and y i represent the geographical coordinates of the i-th device, indicating longitude and latitude; x j and y j represent the geographical coordinates of the j-th device, indicating longitude and latitude; D i represents the device distribution density in the i-th area; N i represents the number of devices in the i-th area; A i represents the area of the i-th area;

[0031] Cluster analysis: Use the clustering algorithm to group the devices spatially and identify clusters of devices with similar geographical locations:

[0032]

[0033] Where: d ik represents the distance from the i-th device to the k-th cluster center; x ij represents the j-th feature of the i-th device; c kjDenote the cluster center value of the j-th feature of the k-th cluster;;

[0034] The way to update the cluster center is as follows:

[0035]

[0036] In the formula: c kj Denote the cluster center value of the j-th feature of the k-th cluster; N k Denote the number of devices in the k-th cluster; S k Denote the set of devices in the k-th cluster;

[0037] Spatial anomaly detection: Calculate the local density using the sliding window method, and obtain the anomaly detection result based on the local density:

[0038]

[0039] In the formula: ρ(x) denotes the local density at position x; W(x) denotes the set of devices within the sliding window at position x; dist(x, i) denotes the distance between position x and device i;

[0040] Adopt threshold detection for density anomalies. If ρ(x) is greater than the threshold, it is considered an anomaly.

[0041] Preferably, step 4 includes the following steps:

[0042] Conversion of time features and spatial features: Convert the results obtained from the extraction of time features and spatial features into the form of basic probability assignment in Dempster-Shafer theory:

[0043] Let the results of time feature and spatial feature extraction be denoted as E t and E s , For each feature E, calculate its corresponding BPA function m(E), which is defined as:

[0044] m(E) = P(E) = 1 - σ;

[0045] Where: P(E) denotes the confidence in each feature E; σ denotes the standard deviation of the feature;

[0046] Perform evidence combination based on Dempster-Shafer theory:

[0047]

[0048] In the formula: m combined (A) denotes the degree of belief in event A after combination; m1(B) and m2(C) denote BPA functions from different sources; B and C denote the basic events in the evidence sources;

[0049] Calculate the combined belief function and the comprehensive belief measure:

[0050]

[0051] In the formula: Bel(A) represents the degree of belief in event A; represents all sub-events included in A;

[0052] According to the belief function Bel(A) and the belief measure m combined (A), calculate the confidence interval CI; use the belief function and the combined BPA value to generate the probability distribution of each state quantity;

[0053] Among them, the comprehensive belief measure includes the confidence interval and the probability distribution.

[0054] Preferably, the confidence interval CI is calculated based on the following steps:

[0055]

[0056] In the formula: z represents the Z value corresponding to the execution level, which is 1.96; n is the sample size.

[0057] Preferably, step 5 includes the following steps:

[0058] Determine the weighting coefficients: Calculate the weight of the comprehensive belief degree based on the confidence degree of the belief degree function and the combined BPA value; Determine the weight of the time feature extraction result based on the standard deviation in the time feature data; Determine the weight of the space feature extraction result based on the fluctuation of the local density of the space feature;

[0059] State estimation fusion:

[0060]

[0061] In the formula: represents the fused power grid state estimation; respectively represent the comprehensive belief measure and the extraction results of the time feature and the space feature; w combined , w t , w s respectively represent the weights of the comprehensive belief measure, the time feature extraction result, and the space feature extraction result;

[0062] Fusion uncertainty estimation:

[0063]

[0064] In the formula: σ fused represents the uncertainty of the fused state estimation;

[0065] Iterative update: By continuously adjusting the weights of each data source, according to the deviation between the fused state estimate and the true power grid state, gradually optimize the weight allocation until the fusion result converges, and output the final power grid state estimate obtained by weighted fusion using the Bar-Shalom-Campo fusion algorithm, as well as the uncertainty of the fused power grid state estimate.

[0066] Preferably, the weight of the comprehensive belief measure is obtained based on the following method:

[0067]

[0068] In the formula: w combined represents the weight of the comprehensive belief measure; S i represents the trust measure of the i-th data source; n represents the total number of data sources; where the trust measure S of each data source i represents the degree of trust based on the confidence interval.

[0069] The beneficial effects of the present invention include:

[0070] Through the spatio-temporal data characterization method of CI and Bar-Shalom-Campo fusion, the present invention effectively solves the problems of consistency, noise processing, and spatio-temporal feature integration in multi-source data fusion; by collecting and preprocessing data from sensors, PMUs, and SCADA systems, it lays a foundation for subsequent analysis; by extracting time features (including time series, frequency domain, and time-frequency features), it captures the dynamic changes of the power grid state; based on power grid topology and geographical location data, it extracts spatial features to enhance the understanding of the power grid structure and environmental impacts; it uses the Dempster-Shafer theory to convert time and spatial features into evidence, calculates the comprehensive belief measure, which combines the confidence intervals and probability distributions of each data source, and effectively processes the uncertainty of the data; finally, it fuses these comprehensive belief measures through the Bar-Shalom-Campo algorithm to achieve an accurate estimate of the power grid state. This method effectively improves the accuracy and reliability of power grid state estimation through the integration of spatio-temporal features and evidence synthesis. Description of the Drawings

[0071] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0072] Figure 1 It is the overall step block diagram provided by the embodiment of the present invention. Detailed Embodiments

[0073] In order to make the technical problems, technical solutions and beneficial effects to be solved by this application clearer and more understandable, the following further details this application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0074] Refer to Figure 1 As shown, a further description is made of the optimal embodiment of the present invention;

[0075] A spatio-temporal data characterization method integrating CI and Bar-Shalom-Campo includes the following steps:

[0076] Step 1: Collect data of each sensor, PMU data and SCADA system data in the power grid, and preprocess the collected data (process data noise, outliers, perform data cleaning and standardization to ensure the accuracy of subsequent analysis);

[0077] Sensor data: Includes real-time measurement data such as voltage, current, power, and frequency.

[0078] PMU data: Provides high-precision phasor measurement and can accurately obtain the phase angle and amplitude of each point in the power grid.

[0079] SCADA system data: Includes equipment status, switch operation records, historical data, etc., and is used to monitor and control the operation of the power grid.

[0080] Step 2: Extract time features based on the preprocessed data to obtain time features, frequency domain features, and time-frequency features;

[0081] Step 3: Extract spatial features based on the power grid topology information, preprocessed sensor data, and geographical location data to obtain topological features and geographical features;

[0082] Step 4: Convert the results obtained from time feature extraction and spatial feature extraction into evidence, and based on the Dempster-Shafer theory, synthesize multi-source evidence, calculate the fused belief function, and obtain the comprehensive belief measure, where the comprehensive belief measure is the confidence interval and probability distribution of each state quantity obtained based on evidence synthesis;

[0083] Step 5: Based on the comprehensive belief measure, the results of time feature extraction, and the results of spatial feature extraction, perform multi-source signal fusion through the Bar-Shalom-Campo fusion algorithm to obtain the power grid state estimation after multi-source fusion.

[0084] The present invention uses a spatio-temporal data characterization method that combines CI and Bar-Shalom-Campo to effectively solve the problems of consistency, noise processing, and spatio-temporal feature synthesis in multi-source data fusion; it collects and preprocesses data from sensors, PMUs, and SCADA systems to lay a foundation for subsequent analysis; it extracts temporal features (including time series, frequency domain, and time-frequency features) to capture the dynamic changes in the power grid state; it extracts spatial features based on power grid topology and geographical location data to enhance the understanding of the power grid structure and environmental impacts; it uses the Dempster-Shafer theory to convert temporal and spatial features into evidence and calculates a comprehensive belief metric that combines the confidence intervals and probability distributions of each data source, effectively handling data uncertainty; finally, it fuses these comprehensive belief metrics through the Bar-Shalom-Campo algorithm to achieve an accurate estimation of the power grid state. This method effectively improves the accuracy and reliability of power grid state estimation through the synthesis of spatio-temporal features and evidence synthesis.

[0085] As a possible implementation of this embodiment, the temporal features, frequency domain features, and time-frequency features are obtained based on the following steps:

[0086] Temporal feature extraction: Calculate the mean, variance, peak value, and standard deviation for each type of data based on the preprocessed data; and perform autocorrelation analysis based on the calculated mean:

[0087]

[0088] In the formula: x i represents the i-th data point in the time series; N represents the total number of data points in the time series; μ represents the mean; x i+τ represents the i + τ-th data point in the time series; τ represents the lag time or delay;

[0089] Frequency domain feature extraction: Based on the preprocessed data, use the fast Fourier transform to transform the time series into the frequency domain, and then perform power spectral density calculation and dominant frequency calculation:

[0090] P(f) = |X(f)|^2;

[0091] In the formula: P(f) represents the power spectral density, indicating the power distribution at different frequencies; X(f) represents the fast Fourier transform;

[0092] where the dominant frequency corresponds to the frequency with the maximum power frequency density;

[0093] Time-frequency feature data: Based on the preprocessed data, use the short-time Fourier transform to obtain the time-frequency features:

[0094]

[0095] where: w(n - t) is the window function; t is the unknown time; f is the frequency; x(n) represents the nth data point in the time series, representing the value at time point n; e -j2πfn represents the complex exponential function, which is used to transform each data point in the time series to the frequency domain; represents the summation load, indicating the summation of the signal values and the weighted values of the window function for all time points n.

[0096] In this embodiment, in power grid state estimation, the extraction of time characteristics is crucial, which can describe in detail the dynamic behavior and change trend of power grid parameters. By calculating basic statistics such as mean, variance, peak value, and standard deviation, the core characteristics of power grid data can be effectively captured. In addition, autocorrelation analysis can reveal the periodic changes and long-term trends of the power grid state by calculating the correlation between different time points in the time series. These time characteristics play an important role in identifying abnormal situations, load fluctuations, and potential faults in the power grid. For example, by analyzing the periodic changes of voltage, possible voltage stability problems can be predicted, and corresponding preventive measures can be taken.

[0097] As a possible implementation manner of this embodiment, step 3 includes the following steps:

[0098] Spatial distribution analysis: Analyze the distribution of power grid equipment in the geographical space:

[0099]

[0100]

[0101] where: d ij represents the distance between the ith and jth devices; x i and y i represent the geographical coordinates of the ith device, representing longitude and latitude; x j and y j represent the geographical coordinates of the jth device, representing longitude and latitude; D i represents the device distribution density in the ith area; N i represents the number of devices in the ith area; A i represents the area of the ith area;

[0102] Clustering analysis: Use a clustering algorithm to group the devices spatially and identify device clusters with similar geographical locations:

[0103]

[0104] where: d ik represents the distance from the ith device to the kth clustering center; x ijRepresents the j-th feature of the i-th device; c kj Represents the cluster center value of the j-th feature of the k-th cluster;;

[0105] Among them, the cluster center update method is as follows:

[0106]

[0107] In the formula: c kj Represents the cluster center value of the j-th feature of the k-th cluster; N k Represents the number of devices in the k-th cluster; S k Represents the set of devices in the k-th cluster;

[0108] Spatial anomaly detection: Use the sliding window method to calculate the local density, and obtain the anomaly detection result based on the local density:

[0109]

[0110] In the formula: ρ(x) represents the local density at position x; W(x) represents the set of devices within the sliding window at position x; dist(x, i) represents the distance between position x and device i;

[0111] Adopt threshold detection for density anomalies. If ρ(x) is greater than the threshold, it is considered an anomaly.

[0112] In this embodiment, spatial feature extraction can provide detailed information on the grid structure and spatial distribution by integrating grid topology information and geographical location data. These spatial features are crucial for accurately evaluating the state of the power grid. For example, topological features help understand important nodes and connection relationships in the power grid, while geographical features can reflect the geographical distribution of the power grid. By combining these features, the load distribution of the power grid, the scope of fault impact, and the sensitivity to environmental factors can be better analyzed. This comprehensive spatial information helps improve the accuracy of power grid monitoring and fault location, thereby optimizing the operation and maintenance strategies of the power grid.

[0113] As a possible implementation manner of this embodiment, step 4 includes the following steps:

[0114] Conversion of time features and spatial features: Convert the results obtained from time feature extraction and spatial feature extraction into the form of basic probability assignment in Dempster-Shafer theory:

[0115] Let the results of time feature extraction and spatial feature extraction be represented as E t and E s , for each feature E, calculate its corresponding BPA function m(E), which is defined as:

[0116] m(E) = P(E) = 1 - σ;

[0117] Among them: P(E) represents the confidence in each feature E; σ represents the standard deviation of the feature;

[0118] Perform evidence synthesis based on the Dempster-Shafer theory:

[0119]

[0120] In the formula: m combined (A) represents the belief in event A after synthesis; m1(B) and m2(C) represent BPA functions from different sources; B and C represent elementary events in the evidence sources;

[0121] Calculate the fused belief function and the combined belief measure:

[0122]

[0123] In the formula: Bel(A) represents the degree of belief in event A; represents all sub-events included in A;

[0124] According to the belief function Bel(A) and the belief function m combined (A), calculate the confidence interval CI; use the belief function and the synthesized BPA value to generate the probability distribution of each state variable;

[0125] Among them, the combined belief measure includes the confidence interval and the probability distribution. As a possible implementation of this embodiment, the confidence interval CI is calculated based on the following steps:

[0126]

[0127] In the formula: z represents the Z value corresponding to the execution level, which is 1.96; n is the sample size.

[0128] In this embodiment, converting time and space features into evidence and performing synthesis based on the Dempster-Shafer theory can effectively fuse information from different data sources. This method can not only handle the noise and uncertainty of data, but also calculate the fused belief function and provide a combined belief measure. This combined belief measure can provide the confidence interval and probability distribution of each power grid state, thereby improving the accuracy and reliability of state estimation. For example, the combined belief measure can help evaluate the possibility and severity of power grid faults, thus providing a more scientific basis for decision-makers to formulate countermeasures.

[0129] As a possible implementation of this embodiment, step 5 includes the following steps:

[0130] Determine the weighting coefficients: Calculate the weight of the comprehensive belief degree based on the belief degree function and the confidence of the synthesized BPA value; Determine the weight of the time feature extraction result based on the standard deviation in the time feature data; Determine the weight of the spatial feature extraction result based on the fluctuation of the local density of the spatial feature;

[0131] State estimation fusion:

[0132]

[0133] In the formula: represents the fused power grid state estimation; respectively represent the comprehensive belief measure and the extraction results of time features and spatial features; w combined 、w t 、w s respectively represent the weights of the comprehensive belief measure, the time feature extraction result, and the spatial feature extraction result;

[0134] Fusion uncertainty estimation:

[0135]

[0136] In the formula: σ fused represents the uncertainty of the fused state estimation;

[0137] Iterative update: By continuously adjusting the weights of each data source, according to the deviation between the fused state estimation and the true power grid state, gradually optimize the weight allocation until the fusion result converges, and output the final power grid state estimation obtained by weighted fusion using the Bar-Shalom-Campo fusion algorithm, as well as the uncertainty of the fused power grid state estimation.

[0138] In this embodiment, it is detailed how to calculate the weight of the comprehensive belief degree based on the belief degree function and the confidence of the synthesized BPA value, the standard deviation of the time feature data to determine the weight of the time feature extraction result, and the local density fluctuation of the spatial feature to determine the weight of the spatial feature extraction result. This method of determining weights based on statistical features and trust ensures a reasonable distribution of the contribution degree of each data source in the fusion process. By dynamically adjusting the weights, it can effectively cope with the changes in the data quality and importance of each data source, making the fusion result more accurate and robust. For example, when the data of a certain sensor is abnormal, its weight will automatically decrease, thereby reducing its impact on the final state estimation.

[0139] In multi-source data fusion, uncertainty estimation is an important indicator to evaluate the reliability of the fusion result. Through σ fusedThe calculation is based on the reciprocal sum of the weights of each data source to comprehensively evaluate the overall uncertainty. This method effectively integrates the uncertainty information of each data source, making the uncertainty estimation after fusion more accurate and quantifying the credibility of the fusion result. In power grid state monitoring, this is crucial for identifying potential risks and making dispatching decisions.

[0140] As a possible implementation manner of this embodiment, the weight of the comprehensive belief measure is obtained based on the following method:

[0141]

[0142] In the formula: w combined represents the weight of the comprehensive belief measure; S i represents the trust measure of the i-th data source; n represents the total number of data sources; where the trust measure S of each data source i represents the degree of trust based on the confidence interval.

[0143] In this embodiment, the method of representing the degree of trust based on the confidence interval enables the contribution degree of each data source to be objectively evaluated and reasonably allocated. Through this reasonable allocation of trust measures, the fusion algorithm can better utilize the information of high-credibility data sources and improve the accuracy and reliability of the overall state estimation.

[0144] In this embodiment, by processing the comprehensive belief measure through the Bar-Shalom-Campo fusion algorithm, effective fusion of multi-source signals can be achieved. This method combines temporal and spatial features, considers the relationships and complementarities between different data sources, and thus provides a more accurate power grid state estimation. Applied to actual power grid monitoring, this algorithm can optimize the power grid state estimation result based on the fusion of multiple data sources and information. This not only improves the prediction accuracy of the power grid state but also enhances the early warning ability for potential faults. For example, when the power grid load changes rapidly, this algorithm can quickly integrate the data of sensors, PMUs, and SCADA systems to provide timely decision support for power grid operators and reduce the operation risks of the power grid.

[0145] The above are only the preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent replacements, and improvements made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims

1. A spatio-temporal data representation method integrating CI and Bar-Shalom-Campo, characterized in that It includes the following steps: Step 1: Collect the data of each sensor, PMU data, and SCADA system data in the power grid, and preprocess the collected data; Step 2: Extract time features based on the preprocessed data to obtain time features, frequency domain features, and time-frequency features; Step 3: Extract spatial features based on the power grid topology information, preprocessed sensor data, and geographical location data to obtain topology features and geographical features; Step 4: Convert the results obtained from time feature extraction and spatial feature extraction into evidence, and based on the Dempster-Shafer theory, synthesize multi-source evidence, calculate the fused belief function, and obtain the comprehensive belief measure, where the comprehensive belief measure is the confidence interval and probability distribution of each state quantity obtained based on evidence synthesis; Step 5: Based on the comprehensive belief measure, the results of time feature extraction, and the results of spatial feature extraction, perform multi-source signal fusion through the Bar-Shalom-Campo fusion algorithm to obtain the power grid state estimation after multi-source fusion.

2. The spatio-temporal data representation method of fusing CI and Bar-Shalom-Campo according to claim 1, wherein The time features, frequency domain features, and time-frequency features are obtained based on the following steps: Time feature extraction: Calculate the mean, variance, peak value, and standard deviation of each type of data based on the preprocessed data; and perform autocorrelation analysis based on the calculated mean: where: x i represents the i-th data point in the time series; N represents the total number of data points in the time series; μ represents the mean; x i+τ represents the (i + τ)-th data point in the time series; τ represents the lag time or delay; Frequency domain feature extraction: Based on the preprocessed data, use the fast Fourier transform to convert the time series to the frequency domain, and then calculate the power spectral density and the main frequency: P(f) = |X(f)| 2 ; In the formula: P(f) represents the power spectral density, indicating the power distribution at different frequencies; X(f) represents the fast Fourier transform; Among them, the main frequency corresponds to the frequency of the maximum power frequency density; Time-frequency feature data: Based on the preprocessed data, use the short-time Fourier transform to obtain time-frequency features: Where: w(n - t) is the window function; t is the unknown time; f is the frequency; x(n) represents the nth data point in the time series, indicating the value at time point n; e -j2πfn represents the complex exponential function, which is used to transform each data point in the time series into the frequency domain; represents the summation load, indicating the summation of the signal values and the weighted values of the window function for all time points n.

3. The spatio-temporal data characterization method of CI and Bar-Shalom-Campo fusion according to claim 1, wherein The said Step 3 includes the following steps: Spatial distribution analysis: Analyze the distribution of power grid equipment in the geographical space: where: d ij represents the distance between the i-th and j-th devices; x i and y i represent the geographical coordinates of the i-th device, representing longitude and latitude; x j and y j represent the geographical coordinates of the j-th device, representing longitude and latitude; D i represents the device distribution density in the i-th area; N i represents the number of devices in the i-th area; A i represents the area of the i-th area; Clustering analysis: Use the clustering algorithm to group the equipment spatially and identify equipment clusters with similar geographical locations: where: d ik represents the distance from the i-th device to the k-th cluster center; x ij represents the j-th feature of the i-th device; c kj represents the cluster center value of the j-th feature of the k-th cluster;; Among them, the clustering center update method is as follows: where: c kj represents the clustering center value of the j-th feature of the k-th cluster; N k represents the number of devices in the k-th cluster; S k represents the set of devices in the k-th cluster; Spatial anomaly detection: Use the sliding window method to calculate the local density, and obtain the anomaly detection result based on the local density: In the formula: ρ(x) represents the local density at position x; W(x) represents the set of equipment within the sliding window at position x; dist(x,i) represents the distance between position x and equipment i; Adopt threshold detection for density anomalies. If ρ(x) is greater than the threshold, it is considered an anomaly.

4. The spatio-temporal data representation method integrating CI and Bar-Shalom-Campo according to claim 1, wherein The said Step 4 includes the following steps: Conversion of time features and spatial features: Convert the results obtained from time feature extraction and spatial feature extraction into the basic probability assignment form in the Dempster-Shafer theory: Let the results of time feature and space feature extraction be denoted as E t and E s , for each feature E, calculate its corresponding BPA function m(E), which is defined as: m(E) = P(E) = 1 - σ; Among them: P(E) represents the confidence in each feature E; σ represents the standard deviation of the feature; Perform evidence synthesis based on the Dempster-Shafer theory: Where: m combined (A) represents the degree of belief in event A after synthesis; m1(B) and m2(C) represent BPA functions from different sources; B and C represent elementary events in the evidence sources; Calculate the fused belief function and the comprehensive belief measure: where: Bel(A) represents the degree of belief in event A; represents all sub-events included in A; According to the belief function Bel(A) and the basic probability assignment function m combined (A) to calculate the confidence interval CI; using the belief function and the synthesized BPA values to generate the probability distributions of each state variable; Among them, the comprehensive belief measure includes the confidence interval and the probability distribution.

5. The spatio-temporal data characterization method by fusing CI and Bar-Shalom-Campo according to claim 4, wherein The confidence interval CI is calculated based on the following steps: Where: z represents the Z value corresponding to the execution level, which is 1.96; n is the sample size.

6. The spatio-temporal data representation method integrating CI and Bar-Shalom-Campo according to claim 1, characterized in that, Step 5 includes the following steps: Determine the weighting coefficients: Calculate the weight of the comprehensive belief degree based on the belief degree function and the confidence of the synthesized BPA value; Determine the weight of the time feature extraction result based on the standard deviation in the time feature data; Determine the weight of the spatial feature extraction result based on the fluctuation of the local density of the spatial features; State estimation fusion: In the formula: represents the fused power grid state estimation; respectively represent the comprehensive belief measure and the extraction results of time features and spatial features; w combined 、w t 、w s respectively represent the weights of the comprehensive belief measure, the extraction results of time features and spatial features; Fusion uncertainty estimation: where: σ fused represents the uncertainty of the fused state estimate; Iterative update: By continuously adjusting the weights of each data source and according to the deviation between the fused state estimation and the true power grid state, gradually optimize the weight allocation until the fusion result converges, and output the final power grid state estimation obtained by weighted fusion using the Bar-Shalom-Campo fusion algorithm, as well as the uncertainty of the fused power grid state estimation.

7. The spatio-temporal data characterization method of CI and Bar-Shalom-Campo fusion according to claim 6, characterized in that The weight of the comprehensive belief measure is obtained based on the following method: Where: w combined represents the weight of the comprehensive belief measure; S i represents the trust measure of the i-th data source; n represents the total number of data sources; the trust metric S of each data source i represents the degree of trust based on the confidence interval.

Images