Substation dangerous area determination method and system based on electric field gradient sensing

Abstract

The application discloses a substation dangerous area judgment method and system based on electric field gradient sensing, and relates to the field of safety monitoring, comprising: arranging an electric field sensor array in a key equipment area of a substation; carrying out noise reduction processing on data of each sensor node to obtain an electric field vector set; obtaining an electric field intensity change rate of each monitoring point through a spatial vector differential algorithm to generate an electric field gradient tensor; dividing the electric field gradient value to construct a dynamic danger level mapping rule; carrying out spatial interpolation through a Delaunay triangulation algorithm to generate an electric field gradient equipotential surface; constructing an electric field gradient space distribution model to generate a dynamic danger boundary envelope in real time; and correcting the danger area boundary through a pre-trained gradient compensation coefficient matrix and triggering a hierarchical early warning signal. The application has the advantages that through the electric field sensor array and the wavelet denoising technology, the electric field gradient change of the substation is monitored in real time, the dangerous area is dynamically evaluated, and the accuracy and real-time performance of the danger area judgment are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring, and in particular to a method and system for determining hazardous areas in substations based on electric field gradient sensing. Background Technology

[0002] As one of the core facilities of the power system, substations operate in a complex environment with high-voltage electrical equipment. Equipment failures, operational errors, or natural disasters can cause sudden changes in electric field strength, potentially leading to electric shocks and other safety accidents. With the development of intelligent monitoring technology, hazard zone determination methods based on electric field gradient perception not only improve safety but also effectively enhance the operational efficiency and maintenance management level of substations, demonstrating significant application prospects.

[0003] Current methods for identifying hazardous areas in substations largely rely on traditional static monitoring and single-sensor data analysis, often lacking high-precision capture of changes in electric field gradients. Many methods use data monitoring based on simple voltage, current, or temperature sensors, which struggles to reflect subtle spatial changes in the electric field in real time, resulting in lag or inaccuracy in identifying potential hazardous areas. Furthermore, existing methods typically ignore the dynamic relationship between equipment load current and electric field gradients, making it unable to handle hazardous area boundary adjustments under complex current fluctuations. In addition, some methods fail to fully utilize 3D modeling and spatial interpolation algorithms, failing to accurately depict the equipotential surface of the electric field gradient, leading to coarse hazardous area boundary delineation that fails to reflect the dynamic changes in the complex electric field distribution within the substation. Summary of the Invention

[0004] To improve existing methods and systems, this paper provides a method and system for determining hazardous areas in substations based on electric field gradient sensing. With the development of intelligent monitoring technology, this method not only improves safety but also effectively enhances the operating efficiency and maintenance management level of substations, showing significant application prospects.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for determining hazardous areas in substations based on electric field gradient sensing includes:

[0007] Electric field sensor arrays are deployed in the key equipment areas of the substation to form a distributed sensing network covering the equipment area, passage area and outer boundary;

[0008] The electric field intensity vectors and corresponding spatial coordinates and timestamps of each sensor node are acquired in real time. The impulse noise of the original data is suppressed by the wavelet threshold denoising algorithm to obtain the denoised electric field vector set.

[0009] Based on the denoised electric field vector set, the rate of change of electric field intensity at each monitoring point is obtained through the spatial vector differential algorithm, and the electric field gradient tensor is generated.

[0010] By setting a gradient threshold, the electric field gradient values ​​are divided, and a dynamic hazard level mapping rule is constructed.

[0011] The Delaunay triangulation algorithm is used to spatially interpolate discrete monitoring points to generate three-dimensional electric field gradient equipotential surfaces.

[0012] Based on the gradient threshold and the obtained three-dimensional electric field gradient equipotential surface, an electric field gradient spatial distribution model is constructed to generate a dynamic danger boundary envelope in real time.

[0013] Based on the real-time load current of key equipment in the substation, the boundary of the danger zone is corrected by a pre-trained gradient compensation coefficient matrix, the coordinate set of the danger zone after compensation is obtained, and a graded early warning signal is triggered.

[0014] Preferably, the step of arranging an electric field sensor array in the key equipment area of ​​the substation to form a distributed sensing network covering the equipment area, the passage area, and the outer boundary specifically includes:

[0015] The electric field sensor array consists of at least four sets of triaxial electric field sensor nodes forming a cross-shaped topology. The node spacing is dynamically adjusted according to the device voltage, and the deployment height covers the space from the device surface to 1.5 meters above the ground.

[0016] Each sensor node is equipped with a three-dimensional electric field probe and a high-precision positioning module to collect spatial coordinate data in real time.

[0017] The electric field sensor array covers the substation equipment area, passageway area, and outer boundary.

[0018] Preferably, the real-time acquisition of the electric field intensity vector and corresponding spatial coordinates and timestamps of each sensor node, and the suppression of impulse noise in the original data using a wavelet threshold denoising algorithm to obtain the denoised electric field vector set specifically includes:

[0019] The electric field intensity vector, spatial coordinates, and timestamp data are acquired in real time through each sensor node. The electric field data includes the components of the electric field in three directions.

[0020] The electric field intensity vector, spatial coordinates and timestamp data collected by each sensor node are organized into time series data in chronological order, with each data point corresponding to a timestamp;

[0021] Wavelet decomposition is performed on each electric field component using wavelet basis functions, decomposing the electric field signal into detail coefficients and approximation coefficients in different frequency bands;

[0022] Detail coefficients in the wavelet decomposition process are preserved and suppressed by using soft thresholding and hard thresholding.

[0023] The processed electric field components are reconstructed into a denoised electric field signal by inverse wavelet transform, and the denoised electric field vector is obtained.

[0024] Preferably, the step of obtaining the rate of change of electric field intensity at each monitoring point based on the denoised electric field vector set and generating the electric field gradient tensor through a space vector differential algorithm specifically includes:

[0025] For each pair of adjacent monitoring points, calculate the difference between the electric field components of the electric field vector sets after noise reduction.

[0026] Based on the spatial coordinate difference, the partial derivatives of the electric field are calculated using the finite difference method, where the partial derivatives represent the electric field gradient.

[0027] Based on the above differential calculation steps, the rate of change of electric field intensity in the surrounding neighborhood of each monitoring point is calculated to obtain each component of the electric field gradient.

[0028] All components of the electric field gradient are combined into a gradient tensor, and a gradient matrix reflecting the change of the electric field in each direction in space is obtained for each monitoring point.

[0029] Preferably, the step of dividing the electric field gradient values ​​by setting a gradient threshold and constructing a dynamic hazard level mapping rule specifically includes:

[0030] Based on historical operating data of substations, electric field gradient thresholds are analyzed and obtained, with each threshold corresponding to an electric field gradient.

[0031] Based on the electric field gradient threshold, the electric field gradient value is divided into different levels, including safe zone, warning zone, danger zone and high-risk zone;

[0032] A dynamic hazard level mapping rule is constructed to automatically determine the hazard level based on the real-time collected electric field gradient data and updated equipment data, and to update it in real time when the electric field gradient value changes.

[0033] Preferably, the step of generating a three-dimensional electric field gradient equipotential surface by spatial interpolating discrete monitoring points using the Delaunay triangulation algorithm specifically includes:

[0034] A discrete point set is generated based on the spatial coordinates of each sensor node. The discrete points are connected and divided into non-overlapping triangles using the Delaunay triangulation algorithm. The vertices of each triangle unit are the original monitoring points, and each triangle unit contains only three points.

[0035] Based on the Delaunay triangulation results, interpolation judgment is performed according to the three vertices in the triangular element and their corresponding electric field gradient values. The electric field gradient value of the interpolation point is calculated by linear interpolation.

[0036] After obtaining the electric field gradient values ​​at different locations in the sensor coverage area through spatial interpolation, the points with the same gradient value are connected to form equipotential surfaces, thus constructing a three-dimensional electric field gradient equipotential surface.

[0037] Preferably, the step of constructing an electric field gradient spatial distribution model based on a gradient threshold and the obtained three-dimensional electric field gradient equipotential surface, and generating a dynamic danger boundary envelope in real time, specifically includes:

[0038] Based on the gradient threshold and the obtained three-dimensional electric field gradient equipotential surface, a spatial distribution model of the electric field gradient is constructed.

[0039] Iterate through the vertex gradient values ​​of each triangular unit and perform spatial cutting of the equipotential surface based on the electric field gradient threshold;

[0040] Vertices with gradient values ​​greater than the danger zone threshold are marked as dangerous boundary points, and high-risk connected components in the space are obtained based on the dangerous boundary points.

[0041] The depth-first search algorithm is used to identify spatially adjacent triangular units to form closed high-risk regions. For each high-risk connected region, the set of its outermost vertices is extracted, and the minimum boundary polygon is constructed using the convex hull algorithm.

[0042] Virtual interpolation points are inserted inside the boundary polygon and reconstructed into a non-uniform rational B-spline surface;

[0043] Add a timestamp attribute to each non-uniform rational B-spline surface to obtain the dynamic danger boundary envelope.

[0044] Preferably, the step of correcting the danger zone boundary based on the real-time load current of key equipment in the substation using a pre-trained gradient compensation coefficient matrix, obtaining the compensated danger zone coordinate set, and triggering a graded early warning signal specifically includes:

[0045] Real-time acquisition of load current data for key equipment in substations;

[0046] Based on historical data of equipment load current and safety thresholds, the boundaries of the danger zone are initially set, and the gradient compensation coefficient matrix is ​​obtained through machine learning model training.

[0047] By combining the real-time load current value with the gradient compensation coefficient matrix, the boundary of the danger zone is dynamically corrected, and a new set of danger zone coordinates is calculated and obtained.

[0048] Based on changes in the boundaries of the danger zone, tiered early warning signals are established.

[0049] Preferably, the graded early warning signal specifically includes: when a warning zone is detected, triggering an audible and visual alarm and generating a yellow warning boundary; when the duration of the high-risk zone is >3 seconds, activating the device power reduction protocol and sending a code alarm to the monitoring center; when personnel enter the high-risk zone, triggering the area power outage protection and sending a code emergency command.

[0050] Furthermore, a substation hazard zone determination system based on electric field gradient sensing is proposed, including:

[0051] Electric field sensor module: The module deploys an array of electric field sensors in the key equipment area to form a distributed sensing network, covering the equipment area, the channel area and the surrounding boundary;

[0052] Data acquisition and processing module: The module is used to acquire the electric field intensity vector, spatial coordinates and timestamp data of the sensor node in real time, and obtain the denoised electric field vector set through wavelet denoising algorithm;

[0053] Electric field gradient calculation module: The module calculates the rate of change of electric field intensity at the monitoring point based on the denoised electric field vector set and uses the space vector differential algorithm to generate the electric field gradient tensor.

[0054] Hazard Level Module: The module divides the electric field gradient value into different levels according to the set gradient threshold, including safe zone, warning zone, danger zone and high-risk zone, and updates the dynamic hazard level in real time;

[0055] Equipotential surface generation module: The module uses the Delaunay triangulation algorithm to perform spatial interpolation on discrete monitoring points to generate a three-dimensional electric field gradient equipotential surface, which reflects the distribution characteristics of the electric field in space.

[0056] Dangerous boundary construction module: The module constructs an electric field gradient spatial distribution model based on gradient threshold and three-dimensional electric field gradient equipotential surface, generates a dynamic dangerous boundary envelope, identifies high-risk connected regions, and constructs the minimum boundary polygon;

[0057] Load current and boundary correction module: The module obtains the compensated set of dangerous zone coordinates by dynamically correcting the dangerous zone boundary based on the real-time load current data of key equipment in the substation, and triggers graded early warning signals.

[0058] Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0059] Compared with the prior art, the advantages of the present invention are:

[0060] By combining an electric field sensor array, real-time data acquisition, spatial vector differentiation algorithms, and wavelet denoising techniques, this method can accurately and in real-time monitor electric field changes in key areas of substations and dynamically assess potential hazardous zones. The advantages of this approach lie in its high-precision electric field gradient calculation and three-dimensional spatial modeling, enabling real-time identification of the electric field distribution and changes in equipment areas, passageways, and peripheral boundaries within the substation, thus forming accurate hazardous zone determinations. Spatial interpolation using the Delaunay triangulation algorithm further optimizes the generation of equipotential surfaces for the electric field gradient, effectively improving the accuracy of hazardous boundaries. Furthermore, the method incorporates real-time load current and machine learning compensation coefficients to correct boundaries, enhancing the dynamic response capability of hazardous zones and ensuring equipment safety under load changes. Attached Figure Description

[0061] Figure 1 This is a schematic diagram of the method proposed in this invention;

[0062] Figure 2 This is a schematic diagram of the electric field sensor array arrangement proposed in this invention;

[0063] Figure 3 This is a schematic diagram of obtaining the noise-reduced electric field vector set proposed in this invention;

[0064] Figure 4 This is a schematic diagram of the generated electric field gradient tensor proposed in this invention;

[0065] Figure 5 This is a schematic diagram illustrating the dynamic hazard level mapping rule proposed in this invention;

[0066] Figure 6 This is a schematic diagram of the generation of three-dimensional electric field gradient equipotential surfaces proposed in this invention;

[0067] Figure 7 This is a schematic diagram of the dynamic danger boundary envelope proposed in this invention;

[0068] Figure 8 This is a schematic diagram of obtaining the coordinate set of the danger zone after compensation, as proposed in this invention.

[0069] Figure 9 This is a schematic diagram illustrating the comparison of the early warning response time of the present invention.

[0070] Figure 10 This is a schematic diagram illustrating the effect of the dangerous area update delay comparison in this invention.

[0071] Figure 11 This is a schematic diagram showing the comparison of the accuracy of equipotential surface generation in this invention.

[0072] Figure 12 This is a schematic diagram comparing the wavelet denoising and Kalman filtering effects of the present invention.

[0073] Figure 13 This is an enlarged view of the signal abrupt change region in this invention. Detailed Implementation

[0074] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0075] A substation hazard zone determination system based on electric field gradient sensing includes:

[0076] Electric field sensor module: The module deploys an array of electric field sensors in the key equipment area to form a distributed sensing network, covering the equipment area, the channel area and the surrounding boundary;

[0077] Data acquisition and processing module: The module is used to acquire the electric field intensity vector, spatial coordinates and timestamp data of the sensor node in real time, and obtain the denoised electric field vector set through wavelet denoising algorithm;

[0078] Electric field gradient calculation module: The module calculates the rate of change of electric field intensity at the monitoring point based on the denoised electric field vector set and uses the space vector differential algorithm to generate the electric field gradient tensor.

[0079] Hazard Level Module: The module divides the electric field gradient value into different levels according to the set gradient threshold, including safe zone, warning zone, danger zone and high-risk zone, and updates the dynamic hazard level in real time;

[0080] Equipotential surface generation module: The module uses the Delaunay triangulation algorithm to perform spatial interpolation on discrete monitoring points to generate a three-dimensional electric field gradient equipotential surface, which reflects the distribution characteristics of the electric field in space.

[0081] Dangerous boundary construction module: The module constructs an electric field gradient spatial distribution model based on gradient threshold and three-dimensional electric field gradient equipotential surface, generates a dynamic dangerous boundary envelope, identifies high-risk connected regions, and constructs the minimum boundary polygon;

[0082] Load current and boundary correction module: The module obtains the compensated set of dangerous zone coordinates by dynamically correcting the dangerous zone boundary based on the real-time load current data of key equipment in the substation, and triggers graded early warning signals.

[0083] Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0084] See Figure 1 As shown, the substation hazard zone determination method based on electric field gradient sensing includes:

[0085] Step 1: Deploy electric field sensor arrays in the key equipment areas of the substation to form a distributed sensing network covering the equipment area, passage area, and outer boundary;

[0086] Step 2: Real-time acquisition of the electric field intensity vector, corresponding spatial coordinates, and timestamp of each sensor node; suppression of impulse noise in the original data using a wavelet threshold denoising algorithm to obtain the denoised electric field vector set.

[0087] Step 3: Based on the denoised electric field vector set, obtain the rate of change of electric field intensity at each monitoring point through the spatial vector differential algorithm, and generate the electric field gradient tensor;

[0088] Step 4: Divide the electric field gradient values ​​by setting a gradient threshold and construct a dynamic hazard level mapping rule;

[0089] Step 5: Spatial interpolation of discrete monitoring points is performed using the Delaunay triangulation algorithm to generate three-dimensional electric field gradient equipotential surfaces;

[0090] Step 6: Based on the gradient threshold and the obtained three-dimensional electric field gradient equipotential surface, construct the electric field gradient spatial distribution model and generate the dynamic danger boundary envelope in real time;

[0091] Step 7: Based on the real-time load current of key equipment in the substation, the boundary of the danger zone is corrected by the pre-trained gradient compensation coefficient matrix, the coordinate set of the danger zone after compensation is obtained, and a graded early warning signal is triggered.

[0092] See Figure 2 As shown, deploying an array of electric field sensors in the key equipment area of ​​the substation to form a distributed sensing network covering the equipment area, passageway area, and outer boundary specifically includes:

[0093] The electric field sensor array consists of at least four sets of triaxial electric field sensor nodes forming a cross-shaped topology. The node spacing is dynamically adjusted according to the device voltage, and the deployment height covers the space from the device surface to 1.5 meters above the ground.

[0094] Each sensor node is equipped with a three-dimensional electric field probe and a high-precision positioning module to collect spatial coordinate data in real time.

[0095] The electric field sensor array covers the substation equipment area, passageway area, and outer boundary.

[0096] See Figure 3 As shown, the electric field intensity vectors, corresponding spatial coordinates, and timestamps of each sensor node are acquired in real time. Wavelet thresholding denoising algorithm is used to suppress impulse noise in the original data, resulting in a denoised electric field vector set, specifically including:

[0097] The electric field intensity vector, spatial coordinates, and timestamp data are acquired in real time through each sensor node. The electric field data includes the components of the electric field in three directions.

[0098] The electric field intensity vector, spatial coordinates and timestamp data collected by each sensor node are organized into time series data in chronological order, with each data point corresponding to a timestamp;

[0099] Wavelet decomposition is performed on each electric field component using wavelet basis functions, decomposing the electric field signal into detail coefficients and approximation coefficients in different frequency bands;

[0100] Detail coefficients in the wavelet decomposition process are preserved and suppressed by using soft thresholding and hard thresholding.

[0101] The processed electric field components are reconstructed into a denoised electric field signal by inverse wavelet transform, and the denoised electric field vector is obtained.

[0102] Specifically, the electric field intensity vector is acquired in real time through sensor nodes, including spatial coordinates and timestamp data. Each sensor records the components of the electric field intensity in three spatial directions at different time points. For each time point, the data format is as follows:

[0103] E(t)=(E x (t),E y (t),E z (t))

[0104] Where E(t) is the electric field intensity vector, t is the timestamp, and E x (t),E y (t),E z (t) represent the components of the electric field in the three spatial directions x, y, and z, respectively;

[0105] The electric field intensity vector, spatial coordinates, and timestamp data collected by each sensor node are organized in chronological order to form a time series dataset;

[0106] The electric field signal is decomposed using wavelet transform. Let f(t) represent the electric field signal. Wavelet decomposition using the wavelet basis function ψ(t) yields the approximation coefficients and detail coefficients of the signal. The formula is as follows:

[0107]

[0108] Among them, a n The coefficients are obtained from wavelet decomposition, and ψ(t) is the mother wavelet function. The signal is decomposed into multiple frequency bands, which typically include high-frequency detail coefficients and low-frequency approximation coefficients.

[0109] The soft thresholding method compares the coefficient with the threshold. If the absolute value of the coefficient is greater than the threshold, the coefficient is subtracted from the threshold; otherwise, it is set to zero. The hard thresholding method sets the absolute values ​​of the detailed coefficients that are less than the threshold to zero, while the absolute values ​​that are greater than the threshold remain unchanged.

[0110] The processed detail coefficients are subjected to inverse wavelet transform to reconstruct the denoised electric field signal from each electric field component. Each electric field component obtained by inverse wavelet transform will form the denoised electric field vector.

[0111] See Figure 4 As shown, based on the denoised electric field vector set, the rate of change of electric field intensity at each monitoring point is obtained through a space vector differential algorithm, and the electric field gradient tensor is generated, specifically including:

[0112] For each pair of adjacent monitoring points, calculate the difference between the electric field components of the electric field vector sets after noise reduction.

[0113] Based on the spatial coordinate difference, the partial derivatives of the electric field are calculated using the finite difference method, where the partial derivatives represent the electric field gradient.

[0114] Based on the above differential calculation steps, the rate of change of electric field intensity in the surrounding neighborhood of each monitoring point is calculated to obtain each component of the electric field gradient.

[0115] All components of the electric field gradient are combined into a gradient tensor, and a gradient matrix reflecting the change of the electric field in each direction in space is obtained for each monitoring point.

[0116] Specifically, for each pair of adjacent monitoring points, the difference between their electric field vectors is calculated, as is the difference in spatial coordinates between adjacent monitoring points. Using the finite difference method, the partial derivatives of the electric field, i.e., the electric field gradient, are calculated. The electric field gradient reflects the rate of change of the electric field intensity in space. In each spatial direction, the partial derivatives of the electric field are approximated using the following formula:

[0117]

[0118] Where, Δx=x j -x i Δy=y j -y i Δz=z j -z i The spatial coordinate difference between two adjacent monitoring points i and j in three directions;

[0119] The components of the electric field gradient are expressed as: E is the electric field gradient vector;

[0120] Based on the electric field gradient, the rate of change of electric field intensity in space is calculated. For each monitoring point, the rate of change of electric field intensity in its surrounding neighborhood is calculated, which is all components of the electric field gradient. By calculating multiple adjacent monitoring points, the change of electric field gradient in the entire region can be obtained, and the electric field gradient tensor, which reflects the change of electric field in various directions in space, can be obtained.

[0121] See Figure 5 As shown, by setting a gradient threshold to divide the electric field gradient values, the dynamic hazard level mapping rule is constructed, specifically including:

[0122] Based on historical operating data of substations, electric field gradient thresholds are analyzed and obtained, with each threshold corresponding to an electric field gradient.

[0123] Based on the electric field gradient threshold, the electric field gradient value is divided into different levels, including safe zone, warning zone, danger zone and high-risk zone;

[0124] A dynamic hazard level mapping rule is constructed to automatically determine the hazard level based on the real-time collected electric field gradient data and updated equipment data, and to update it in real time when the electric field gradient value changes.

[0125] Specifically, by analyzing the historical operating data of the substation, empirical thresholds for electric field gradients are obtained. The historical data records electric field gradient values ​​at different times and locations. By analyzing the data, patterns of electric field changes and potential danger signals can be identified.

[0126] Based on the set electric field gradient threshold, the electric field gradient values ​​are divided into different regions or levels. Common levels include:

[0127] Safe zone: When the electric field gradient value is below the safe threshold, it indicates that the electric field change is small and the power equipment is in normal operation. It is generally considered to be a risk-free area.

[0128] Warning zone: The electric field gradient value is within the medium threshold range, indicating that the electric field is beginning to change significantly, which may indicate potential problems or an impending failure. The equipment requires special attention.

[0129] Danger Zone: When the electric field gradient value exceeds the high-risk threshold, it indicates an abnormal change in the electric field, which may cause equipment failure or pose a serious electrical safety risk, requiring immediate action.

[0130] High-risk area: When the electric field gradient value reaches an extreme value, it usually indicates that the equipment has entered a dangerous state or is experiencing a serious malfunction, requiring emergency intervention;

[0131] By constructing dynamic hazard level mapping rules, the system automatically determines the current hazard level based on real-time collected electric field gradient data and updates it in real time. The system periodically collects electric field gradient data from various monitoring points and generates a complete gradient dataset based on spatial location, timestamps, and other information. Besides electric field gradient data, the operating status of the equipment also needs to be used as a basis for updating the hazard level. For example, data such as equipment load, current, voltage, and temperature can also affect the hazard level determination. By setting appropriate logical rules, the electric field gradient determination level can be automatically adjusted when the equipment status changes.

[0132] See Figure 6 As shown, the Delaunay triangulation algorithm is used to spatially interpolate discrete monitoring points to generate a three-dimensional electric field gradient equipotential surface. Specifically, this includes:

[0133] A discrete point set is generated based on the spatial coordinates of each sensor node. The discrete points are connected and divided into non-overlapping triangles using the Delaunay triangulation algorithm. The vertices of each triangle unit are the original monitoring points, and each triangle unit contains only three points.

[0134] Based on the Delaunay triangulation results, interpolation judgment is performed according to the three vertices in the triangular element and their corresponding electric field gradient values. The electric field gradient value of the interpolation point is calculated by linear interpolation.

[0135] After obtaining the electric field gradient values ​​at different locations in the sensor coverage area through spatial interpolation, the points with the same gradient value are connected to form equipotential surfaces, thus constructing a three-dimensional electric field gradient equipotential surface.

[0136] Specifically, based on the spatial coordinates of each sensor node, a discrete set of points is generated. The Delaunay triangulation algorithm divides these discrete 3D point sets into a set of non-overlapping triangular units. In 3D space, each triangular unit is a tetrahedron, and the four vertices of each tetrahedron correspond to the original monitoring points. The core idea of ​​this algorithm is that in each triangular unit, all the point sets are connected so that no other point is located inside the circumcircle of a certain triangle. This ensures that the generated triangulation is optimal and there are no redundant triangles.

[0137] After Delaunay triangulation, the three vertices of each tetrahedral element correspond to the location of the monitoring point and their electric field gradient values ​​in various spatial directions. After obtaining the electric field gradient values ​​between all monitoring points, spatial interpolation can be performed on the entire sensor coverage area to calculate the electric field gradient values ​​at each location.

[0138] After obtaining the electric field gradient value at each location in the entire region, a specific gradient value is selected, and then all points in space that are equal to that value are found. These points are connected to form an equipotential surface. If multiple equipotential surfaces exist, different thresholds can be selected for different electric field gradient values, and the above process can be repeated to construct multiple equipotential surfaces.

[0139] Through the above steps, the generated three-dimensional electric field gradient equipotential surface will be a three-dimensional region composed of multiple interconnected equipotential surfaces.

[0140] See Figure 7 As shown, based on the gradient threshold and the obtained three-dimensional electric field gradient equipotential surface, an electric field gradient spatial distribution model is constructed, and a dynamic hazard boundary envelope is generated in real time. Specifically, this includes:

[0141] Based on the gradient threshold and the obtained three-dimensional electric field gradient equipotential surface, a spatial distribution model of the electric field gradient is constructed.

[0142] Iterate through the vertex gradient values ​​of each triangular unit and perform spatial cutting of the equipotential surface based on the electric field gradient threshold;

[0143] Vertices with gradient values ​​greater than the danger zone threshold are marked as dangerous boundary points, and high-risk connected components in the space are obtained based on the dangerous boundary points.

[0144] The depth-first search algorithm is used to identify spatially adjacent triangular units to form closed high-risk regions. For each high-risk connected region, the set of its outermost vertices is extracted, and the minimum boundary polygon is constructed using the convex hull algorithm.

[0145] Virtual interpolation points are inserted inside the boundary polygon and reconstructed into a non-uniform rational B-spline surface;

[0146] Add a timestamp attribute to each non-uniform rational B-spline surface to obtain the dynamic danger boundary envelope.

[0147] Specifically, the process iterates through each triangular element in the three-dimensional space and calculates the electric field gradient at each vertex. For each vertex of the triangular element, the magnitude of the electric field gradient is calculated. Based on the electric field gradient threshold, the equipotential surface is spatially cut, and vertices with electric field gradients greater than the threshold are marked as dangerous boundary points.

[0148] The depth-first search algorithm extracts adjacent dangerous regions from the marked dangerous boundary points to form one or more high-risk connected regions. The depth-first search algorithm starts from a dangerous boundary point and visits all its adjacent points until there are no more adjacent points, thus forming a connected region.

[0149] For each high-risk connected component, a closed danger region is formed, which is extracted using the convex hull algorithm. The convex hull is a minimal convex polygon or polyhedron that contains all the points in the danger region. For the vertex set of the high-risk connected component, the outermost boundary points are extracted using the convex hull algorithm to obtain the outermost vertex set of the danger region.

[0150] Virtual interpolation points are inserted within the minimum boundary polygon to accurately describe the boundary of the region. Then, based on these virtual points and boundary points, a non-uniform rational B-spline surface is used for reconstruction. The boundary of each generated danger zone can change over time, so a timestamp attribute is added to each non-uniform rational B-spline surface to indicate the generation time of the surface.

[0151] See Figure 8 As shown, based on the real-time load current of key equipment in the substation, the boundary of the danger zone is corrected through a pre-trained gradient compensation coefficient matrix to obtain the coordinate set of the compensated danger zone and trigger graded early warning signals, specifically including:

[0152] Real-time acquisition of load current data for key equipment in substations;

[0153] Based on historical data of equipment load current and safety thresholds, the boundaries of the danger zone are initially set, and the gradient compensation coefficient matrix is ​​obtained through machine learning model training.

[0154] By combining the real-time load current value with the gradient compensation coefficient matrix, the boundary of the danger zone is dynamically corrected, and a new set of danger zone coordinates is calculated and obtained.

[0155] Based on changes in the boundaries of the danger zone, tiered early warning signals are established.

[0156] Specifically, based on historical load current data and equipment safety thresholds, the boundaries of the substation's hazardous area are initially set. Using historical load current data and the corresponding hazardous area boundaries, a machine learning model is trained to obtain a gradient compensation coefficient matrix. The real-time collected load current is combined with the gradient compensation coefficient matrix to dynamically correct the boundaries of the hazardous area.

[0157] Based on changes in the boundary of the danger zone, different levels of early warning signals are set, including: when a warning zone is detected, an audible and visual alarm is triggered and a yellow warning boundary is generated; when the duration of the high-risk zone is >3 seconds, the equipment power reduction protocol is activated and a code alarm is sent to the monitoring center; when personnel enter the high-risk zone, the area power outage protection is triggered and an emergency code command is sent.

[0158] To verify the feasibility and effectiveness of this invention application, the following experiments were conducted:

[0159] Experimental conditions:

[0160] Sensor array: It adopts a cross topology and consists of 4 sets of triaxial electric field sensor nodes.

[0161] Data acquisition: Real-time acquisition of electric field intensity vector, spatial coordinates, and timestamps. Sampling frequency 1kHz, using wavelet thresholding denoising algorithm (soft / hard thresholding combination) to suppress impulse noise;

[0162] Standard environment: Temperature 25℃±2℃, Humidity 65%±5% (refer to standard DL / T 860-2018); Extreme environment: Humidity >90% (simulating hot and humid climate), Temperature 30℃ (supplementary test);

[0163] Fault characteristics: Pollution flashover scenario: abrupt change in electric field gradient >40% (salt density 0.2 mg / cm³) 2 The amplitude of the partial discharge pulse was increased by 50% (the measured average was 42%).

[0164] Load current: Real-time acquisition, range 0-3000A. Sample size: 200 tests per scenario (e.g., pollution flashover path test) to ensure statistical significance (false negative rate is calculated based on this).

[0165] The differences in early warning response time of the present invention's solution (standard humidity), the present invention's solution (humidity > 90%), comparison group 1 (threshold judgment plus central server decision), comparison group 2 (local rule engine plus cloud verification), and comparison group 3 (edge ​​preprocessing plus threshold judgment) were compared under scenarios such as personnel intrusion, equipment overload, insulation failure, and insulator flashover.

[0166] See Figure 9 The five comparative experiments of the present invention (standard humidity), the present invention (humidity > 90%), and comparison groups 1, 2, and 3 showed average early warning response times of 103.3, 111.3, 367.5, 325.0, and 267.5 seconds respectively under four scenarios: personnel intrusion, equipment overload, insulation fault, and insulator flashover. The compensation matrix learns the humidity-gradient mapping relationship through historical data and automatically adjusts the safety threshold. When humidity > 90%, the delay increases by only 7.8%, improving the stability of load fluctuation response. The present invention provides a millisecond-level response system, meeting the golden handling window of the DL / T 860-2018 standard, providing a millisecond-level protection solution for complex faults in substations. It is particularly suitable for scenarios such as flashover prevention in UHV converter stations, intelligent safety upgrades in dense urban substations, and high-reliability monitoring of new energy power grids.

[0167] Experimental conditions: 200 sensor nodes (triaxial electric field sensor, 100Hz sampling). The performance of this invention's scheme, Comparison Group 1 (central server polling), and Comparison Group 2 (cloud-edge collaboration) in terms of hazardous area update latency under four scenarios: normal load, single fault, multi-device concurrent fault, and multi-device fault + low temperature.

[0168] See Figure 10 As shown, the update latency of the dangerous area in the present invention is 48ms, 50ms, 55ms and 63ms respectively under four scenarios: normal load, single fault, multi-device concurrent fault, and multi-device fault + low temperature. This is significantly better than that of comparison group 1 and comparison group 2, and the advantage is even more obvious when a fault occurs.

[0169] This invention employs an edge streaming computing architecture and parallel spatial interpolation algorithm to achieve simultaneous execution of triangulation and gradient calculation, along with a dynamic boundary compensation mechanism. This results in a 63ms update latency under extreme loads, and a stable response in multi-device failure scenarios with only a 5ms increase in latency compared to single-device failures, demonstrating significant stability advantages and meeting the IEC 61850 TR2 standard. These advantages set a new technological benchmark in the field of smart grid security protection, providing substations with millisecond-level real-time security protection capabilities and effectively addressing the pain point of insufficient accuracy in substation hazard zone identification.

[0170] Experimental conditions and physical model: A silicone rubber composite sleeve with an air interface was added to the cable terminal area, conforming to the IEC60840 standard. A transient electric field coupled EMTP model (simulating dynamic load changes) was used. The performance differences in equipotential surface generation accuracy of the proposed solution were compared: Delaunay triangulation, control group 1 (bilinear interpolation), control group 2 (inverse distance weighting), and control group 3 (Kriging interpolation).

[0171] G ref G represents the real electric field gradient. calc This represents the electric field gradient generated by the scheme.

[0172] See Figure 11 As shown, the present invention demonstrates its accuracy advantage by maintaining the lowest RMSE value in all regions. In the cable terminal area, the error of the present invention (0.048) is 59% lower than that of the optimal control group (Kriging interpolation, 0.118). This advantage stems from the Delaunay triangulation algorithm, which adaptively fits complex boundaries (such as silicone rubber sleeves for cable terminals) with non-overlapping triangular elements, avoiding topological distortion of traditional meshes. Under 90% humidity conditions, the RMSE of the present invention increases by only 0.005 kV / m (from 0.048 to 0.053), demonstrating significant stability advantages. This characteristic is attributed to wavelet denoising technology, which suppresses impulse noise through soft / hard thresholding and reconstructs the denoised electric field vector.

[0173] In complex regions with non-uniform media (such as near insulators), the error of this invention (0.027) is only 40% of that of Kriging interpolation (0.067), highlighting its adaptability to abrupt changes in the medium. Its technical basis lies in vertex adaptive encryption, gradient threshold space cutting, dynamic boundary reconstruction technology to reconstruct the surface to fit the actual geometry of the device, and adding timestamp attributes to the NURBS surface to achieve historical boundary tracing. This invention achieves the following breakthroughs in equipotential surface generation by employing the Delaunay triangulation algorithm and convex hull algorithm: RMSE < 0.05kV / m in complex regions (compliant with IEC 60840 standard); and dynamic response: NURBS surface timestamps support early warning of boundary change trends.

[0174] Experimental conditions: Signal model: power frequency fundamental wave, frequency 50Hz, amplitude 10kV / m, expression 10·sin(2π·50·t); Step jump, occurrence time t=0.2s, amplitude 8kV / m, expression 8·u(t-0.2) (u(t) is the step function). Noise model: Gaussian white noise, signal-to-noise ratio (SNR) 20dB, simulating background electromagnetic interference in a substation. Impulse noise: amplitude ±30% (relative to the fundamental wave amplitude), occurrence rate 5% (random triggering), characteristic: simulating transient interference during switch operation. Comparison of the performance of wavelet filtering (the scheme of this invention) and Kalman filtering.

[0175] Wavelet filtering uses the db4 wavelet basis for multi-level decomposition, hierarchical thresholding, and a soft thresholding function to suppress noise. In the filtering simulation, wavelet denoising (blue dashed line) demonstrates: abrupt change features are preserved (0.2s step change response delay is only 0.5ms), and the power frequency waveform is completely preserved (50Hz sine wave without distortion). Kalman filtering adjusts parameters based on the electric field gradient rate of change to achieve adaptive load current and dynamic noise optimization. Figure 12 , 13 The Kalman filter output (green dashed line) shows that at the abrupt change point t=0.2s, the Kalman filter bias is 1.65kV / m, while the bias in this invention application is 0.1kV / m. The wavelet filter satisfies the description of "effectively suppressing impulse noise while preserving abrupt change characteristics."

[0176] The proposed solution outputs values ​​closest to the true values ​​at the abrupt change moment (t = 0.200s), thus reducing the false negative rate. The wavelet denoising processing delay is <1ms, meeting the 50ms refresh cycle requirement. This proposed solution demonstrates significant advantages in transient response and noise robustness, providing reliable technical support for substation dynamic monitoring.

[0177] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0178] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0179] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for determining hazardous areas in substations based on electric field gradient sensing, characterized in that, include: Electric field sensor arrays are deployed in the key equipment areas of the substation to form a distributed sensing network covering the equipment area, passage area and outer boundary; The electric field intensity vectors and corresponding spatial coordinates and timestamps of each sensor node are acquired in real time. The impulse noise of the original data is suppressed by the wavelet threshold denoising algorithm to obtain the denoised electric field vector set. Based on the denoised electric field vector set, the rate of change of electric field intensity at each monitoring point is obtained through a space vector differential algorithm, and an electric field gradient tensor is generated, specifically including: For each pair of adjacent monitoring points, calculate the difference between the electric field components of the electric field vector sets after noise reduction. Based on the spatial coordinate difference, the partial derivatives of the electric field are calculated using the finite difference method, where the partial derivatives represent the electric field gradient. Based on the above differential calculation steps, the rate of change of electric field intensity in the surrounding neighborhood of each monitoring point is calculated to obtain each component of the electric field gradient. All components of the electric field gradient are combined into a gradient tensor, and a gradient matrix reflecting the change of the electric field in each direction in space is obtained for each monitoring point. By setting a gradient threshold, the electric field gradient values ​​are divided, and a dynamic hazard level mapping rule is constructed. The Delaunay triangulation algorithm is used to spatially interpolate discrete monitoring points to generate three-dimensional electric field gradient equipotential surfaces. Based on the gradient threshold and the obtained three-dimensional electric field gradient equipotential surface, an electric field gradient spatial distribution model is constructed to generate a dynamic danger boundary envelope in real time. Based on the real-time load current of key equipment in the substation, the boundary of the danger zone is corrected by a pre-trained gradient compensation coefficient matrix, the coordinate set of the danger zone after compensation is obtained, and a graded early warning signal is triggered.

2. The method for determining the hazardous area of ​​a substation based on electric field gradient sensing according to claim 1, characterized in that, The deployment of electric field sensor arrays in the key equipment area of ​​the substation to form a distributed sensing network covering the equipment area, passage area, and outer boundary specifically includes: The electric field sensor array consists of at least four sets of triaxial electric field sensor nodes forming a cross-shaped topology. The node spacing is dynamically adjusted according to the device voltage, and the deployment height covers the space from the device surface to 1.5 meters above the ground. Each sensor node is equipped with a three-dimensional electric field probe and a high-precision positioning module to collect spatial coordinate data in real time. The electric field sensor array covers the substation equipment area, passageway area, and outer boundary.

3. The method for determining the hazardous area of ​​a substation based on electric field gradient sensing according to claim 1, characterized in that, The process of acquiring the electric field intensity vector, corresponding spatial coordinates, and timestamp of each sensor node in real time, and then using a wavelet threshold denoising algorithm to suppress impulse noise in the original data to obtain the denoised electric field vector set specifically includes: The electric field intensity vector, spatial coordinates, and timestamp data are acquired in real time through each sensor node. The electric field data includes the components of the electric field in three directions. The electric field intensity vector, spatial coordinates and timestamp data collected by each sensor node are organized into time series data in chronological order, with each data point corresponding to a timestamp; Wavelet decomposition is performed on each electric field component using wavelet basis functions, decomposing the electric field signal into detail coefficients and approximation coefficients in different frequency bands; Detail coefficients in the wavelet decomposition process are preserved and suppressed by using soft thresholding and hard thresholding. The processed electric field components are reconstructed into a denoised electric field signal by inverse wavelet transform, and the denoised electric field vector is obtained.

4. The method for determining the hazardous area of ​​a substation based on electric field gradient sensing according to claim 1, characterized in that, The step of dividing the electric field gradient values ​​by setting a gradient threshold and constructing a dynamic hazard level mapping rule specifically includes: Based on historical operating data of substations, electric field gradient thresholds are analyzed and obtained, with each threshold corresponding to an electric field gradient. Based on the electric field gradient threshold, the electric field gradient value is divided into different levels, including safe zone, warning zone, danger zone and high-risk zone; A dynamic hazard level mapping rule is constructed to automatically determine the hazard level based on the real-time collected electric field gradient data and updated equipment data, and to update it in real time when the electric field gradient value changes.

5. The method for determining the hazardous area of ​​a substation based on electric field gradient sensing according to claim 1, characterized in that, The step of generating a three-dimensional electric field gradient equipotential surface by spatial interpolating discrete monitoring points using the Delaunay triangulation algorithm specifically includes: A discrete point set is generated based on the spatial coordinates of each sensor node. The discrete points are connected and divided into non-overlapping triangles using the Delaunay triangulation algorithm. The vertices of each triangle unit are the original monitoring points, and each triangle unit contains only three points. Based on the Delaunay triangulation results, interpolation judgment is performed according to the three vertices in the triangular element and their corresponding electric field gradient values. The electric field gradient value of the interpolation point is calculated by linear interpolation. After obtaining the electric field gradient values ​​at different locations in the sensor coverage area through spatial interpolation, the points with the same gradient value are connected to form equipotential surfaces, thus constructing a three-dimensional electric field gradient equipotential surface.

6. The method for determining the hazardous area of ​​a substation based on electric field gradient sensing according to claim 1, characterized in that, The process of constructing a spatial distribution model of the electric field gradient based on the gradient threshold and the obtained three-dimensional electric field gradient equipotential surface, and generating a dynamic danger boundary envelope in real time, specifically includes: Based on the gradient threshold and the obtained three-dimensional electric field gradient equipotential surface, a spatial distribution model of the electric field gradient is constructed. Iterate through the vertex gradient values ​​of each triangular unit and perform spatial cutting of the equipotential surface based on the electric field gradient threshold; Vertices with gradient values ​​greater than the danger zone threshold are marked as dangerous boundary points, and high-risk connected components in the space are obtained based on the dangerous boundary points. The depth-first search algorithm is used to identify spatially adjacent triangular units to form closed high-risk regions. For each high-risk connected region, the set of its outermost vertices is extracted, and the minimum boundary polygon is constructed using the convex hull algorithm. Virtual interpolation points are inserted inside the boundary polygon and reconstructed into a non-uniform rational B-spline surface; Add a timestamp attribute to each non-uniform rational B-spline surface to obtain the dynamic danger boundary envelope.

7. The method for determining the hazardous area of ​​a substation based on electric field gradient sensing according to claim 1, characterized in that, The process of using real-time load current of key substation equipment to correct the boundary of the danger zone through a pre-trained gradient compensation coefficient matrix, obtaining the coordinate set of the compensated danger zone, and triggering graded early warning signals specifically includes: Real-time acquisition of load current data for key equipment in substations; Based on historical data of equipment load current and safety thresholds, the boundaries of the danger zone are initially set, and the gradient compensation coefficient matrix is ​​obtained through machine learning model training. By combining the real-time load current value with the gradient compensation coefficient matrix, the boundary of the danger zone is dynamically corrected, and a new set of danger zone coordinates is calculated. Based on changes in the boundaries of the danger zone, tiered early warning signals are established.

8. The method for determining the hazardous area of ​​a substation based on electric field gradient sensing according to claim 1, characterized in that, The tiered early warning signals specifically include: when a warning zone is detected, an audible and visual alarm is triggered and a yellow warning boundary is generated; when the duration of the high-risk zone is > 3 seconds, a power reduction protocol is activated and a code alarm is sent to the monitoring center; when personnel enter the high-risk zone, regional power outage protection is triggered and an emergency code command is sent.

9. A substation hazard zone determination system based on electric field gradient sensing, used to implement the substation hazard zone determination method based on electric field gradient sensing as described in any one of claims 1-8, characterized in that, include: Electric field sensor module: The module deploys an array of electric field sensors in the key equipment area to form a distributed sensing network, covering the equipment area, the channel area and the surrounding boundary; Data acquisition and processing module: The module is used to acquire the electric field intensity vector, spatial coordinates and timestamp data of the sensor node in real time, and obtain the denoised electric field vector set through wavelet denoising algorithm; Electric field gradient calculation module: The module calculates the rate of change of electric field intensity at the monitoring point based on the denoised electric field vector set and uses the space vector differential algorithm to generate the electric field gradient tensor. Hazard Level Module: The module divides the electric field gradient value into different levels according to the set gradient threshold, including safe zone, warning zone, danger zone and high-risk zone, and updates the dynamic hazard level in real time; Equipotential surface generation module: The module uses the Delaunay triangulation algorithm to perform spatial interpolation on discrete monitoring points to generate a three-dimensional electric field gradient equipotential surface, which reflects the distribution characteristics of the electric field in space. Dangerous boundary construction module: The module constructs an electric field gradient spatial distribution model based on gradient threshold and three-dimensional electric field gradient equipotential surface, generates a dynamic dangerous boundary envelope, identifies high-risk connected regions, and constructs the minimum boundary polygon; Load current and boundary correction module: The module obtains the compensated set of dangerous zone coordinates by dynamically correcting the dangerous zone boundary based on the real-time load current data of key equipment in the substation, and triggers graded early warning signals. Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

Images