A new energy vehicle high-voltage insulation detection method

Abstract

The application provides a new energy vehicle high-voltage insulation detection method, and the technical scheme is that discrete temperature and air pressure data under vehicle driving conditions are acquired, and a continuous three-dimensional temperature and air pressure distribution field is reconstructed in combination with a pre-stored space environment base; then, temperature and pressure calculation values of nodes without sensor coverage are extracted, and a transient air critical breakdown voltage is calculated by substituting the temperature and pressure calculation values into a gas discharge breakdown model; finally, a dynamic insulation safety margin of each node is determined according to the critical breakdown voltage and a high-voltage system working voltage, and an insulation early warning signal is output when the dynamic insulation safety margin is lower than a preset threshold. The application can solve the technical problems that the prior art is difficult to realize full-vehicle blind area-free insulation state dynamic environment compensation under extreme variation coupling conditions, insulation failure and missed report are prone to occur, and precise three-dimensional space positioning of insulation hidden dangers cannot be performed under limited computing power at a vehicle end.

Description

Technical Field

[0001] This invention relates to the field of insulation testing technology for new energy vehicles, and specifically to a method for testing the high-voltage insulation of new energy vehicles. Background Technology

[0002] With the widespread application of new energy heavy-duty trucks and off-road vehicles in high-altitude mountainous areas and long downhill slopes, high-power kinetic energy recovery systems are frequently activated. Under these conditions, the vehicles are exposed to both extremely low air pressure due to high altitude and severe heat generation from high-voltage wiring harnesses and motor windings caused by intense electric braking. This deep coupling of low air pressure and localized high-temperature expansion leads to a sharp decrease in the air density within the electrical gaps of high-voltage components. According to the laws of gas discharge physics, this causes a significant drop in the transient critical breakdown voltage of the continuous air medium within the electrical gaps, placing the vehicle's high-voltage insulation system in a highly dangerous state.

[0003] However, existing insulation testing technologies for new energy vehicles generally simplify the entire vehicle insulation system to a single lumped equivalent resistance parameter for monitoring, and the decision thresholds mostly use factory-fixed static safety thresholds. Due to the lack of dynamic compensation mechanisms for complex environmental conditions such as temperature and air pressure, existing detection technologies struggle to detect insulation gaps not covered by physical sensors, such as high-voltage wiring harness bends and the inside of high-voltage connector plugs, which are already on the verge of discharge breakdown, under harsh coupled conditions of high temperature and low air pressure. This can easily lead to missed detections of early partial discharge failures. Furthermore, the limited resources of onboard computing platforms cannot support massive continuous physical field partial differential three-dimensional numerical simulations, making it difficult to achieve high-precision environmental parameter extrapolation for areas not covered by physical sensors, and even more difficult to accurately locate insulation safety risk points in three-dimensional space. Moreover, they are highly susceptible to strong electromagnetic interference generated by large current steps and high-frequency vibration noise caused by bumps.

[0004] Therefore, how to achieve dynamic environmental compensation of vehicle high-voltage insulation status under complex and changing environments, high-precision safety assessment without monitoring blind spots, high-reliability multi-dimensional anti-shake early warning, and three-dimensional spatial positioning of insulation hazard points under vehicle-mounted limited computing power are problems that urgently need to be solved in this field. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a high-voltage insulation detection method for new energy vehicles. This method solves the technical problems of existing technologies, such as difficulty in achieving dynamic environmental compensation for the insulation status of the entire vehicle without blind spots under extreme variable coupling conditions like low air pressure and high heat, easy occurrence of missed insulation failures, and inability to accurately locate insulation hazards in three-dimensional space under limited computing power at the vehicle end.

[0006] The technical solution adopted in this invention is to acquire discrete temperature and air pressure data under vehicle driving conditions, and reconstruct a continuous three-dimensional temperature and air pressure distribution field by combining it with a pre-stored spatial environment substrate; then extract the temperature and pressure estimation values ​​of sensorless nodes, substitute them into the gas discharge breakdown model to calculate their transient air critical breakdown voltage; finally, determine the dynamic insulation safety margin of each node based on the critical breakdown voltage and the high-voltage system operating voltage, and output an insulation warning signal when it is lower than a preset threshold.

[0007] Specifically, it includes the following steps: Acquire discrete environmental state data under driving conditions; Based on the discrete environmental state data and the pre-stored spatial environment base, the spatial field of the high-pressure system is reconstructed to obtain a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field. Extract the estimated temperature and estimated pressure values ​​of sensor-free nodes in the three-dimensional temperature distribution field and the three-dimensional air pressure distribution field; Substitute the estimated temperature value and the estimated air pressure value into the gas discharge breakdown model to calculate the transient air critical breakdown voltage at the sensorless coverage node. The insulation safety margin is determined based on the critical breakdown voltage and the actual operating voltage, and an insulation warning signal is output when it falls below a preset threshold.

[0008] Furthermore, the discrete environmental state data under driving conditions is obtained through the following steps: The discrete environmental state data includes local real-time temperature data and global real-time air pressure data; A time synchronization link for the multimodal physical signals is established, and the acquired raw state signals are time-aligned. The time-aligned synchronization state signal is represented as follows:

[0009] in, This is the synchronization status signal after timing alignment. To synchronize with the target time, and For two adjacent actual sampling times, and At the actual sampling time and The collected values; The Kalman filter algorithm is used to smooth and denoise the synchronization state signal after time alignment to obtain the smoothed and denoised local real-time temperature data and global real-time air pressure data. The smoothed and denoised signal is correlated with the corresponding sensor spatial coordinates to form a discrete environmental state data matrix:

[0010] Where M is the discrete environmental state data matrix, N is the total number of deployed sensors, and the feature column vector is... , Let be the three-dimensional physical space coordinate vector of the i-th sensor. P represents the local real-time temperature data after smoothing and denoising, and P represents the global real-time air pressure data after smoothing and denoising.

[0011] Furthermore, based on the discrete environmental state data and the pre-stored space environment substrate, the high-pressure system is reconstructed to obtain a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field. Specific steps include: The space environment base of the high-voltage system is invoked, and the space environment base of the high-voltage system contains a preset number of orthogonal characteristic functions; In the three-dimensional physical topology model of the high-voltage system, spatial index induced points are preset, and the discrete environmental state data are sparsely approximated using the spatial index induced points to generate dimensionality-reduced state data. The reduced state data is input into the system state matrix, and the random coefficient values ​​of the current sampling period are calculated based on the sparse Gaussian process regression model. Based on the linear combination of the orthogonal characteristic function and the random coefficient value, a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field are reconstructed.

[0012] Furthermore, the construction steps of the space environment substrate of the high-voltage system include: Obtain a dataset containing the test history under alternating operating conditions, and calculate the coordinate vectors of any two spatial nodes. and The expected correlation between them is used to construct a spatial covariance matrix C, whose internal elements are... The calculation formula is:

[0013] in, Let T be the internal element of the spatial covariance matrix C, and let T be the total number of time sampling points in the test history dataset. To test the physical state measurements at corresponding spatial coordinates and times in the historical dataset, For spatial nodes The historical average measurement time at that location; Perform eigenvalue decomposition on the spatial covariance matrix C, solve the characteristic equation, and obtain the orthogonal characteristic functions and the corresponding eigenvalue sequence:

[0014] in, For the nth order eigenvalue, The eigenvector corresponding to the nth order eigenvalue represents the orthogonal eigenfunction; Contribution rates are sorted in descending order of eigenvalues:

[0015] in, The cumulative variance contribution rate of the first K eigenvalues, where M is the total number of nodes in the decomposition space. When the variance contribution is first greater than or equal to a preset variance contribution threshold, the feature vectors are truncated, retaining only the first K eigenvectors. to The space environment substrate that constitutes the high-voltage system is combined.

[0016] Furthermore, the step of pre-setting spatial index induced points in the three-dimensional physical topology model of the high-voltage system, and using the spatial index induced points to perform sparse approximation processing on the discrete environmental state data to generate dimensionality-reduced state data, specifically includes: Analysis of spatial nodes in the three-dimensional physical topology model Based on the electric field gradient and heat flux gradient, a comprehensive sensitivity assessment index is constructed to characterize the degree of insulation stress concentration and environmental abrupt changes:

[0017] Wherein, W(p) is the comprehensive sensitivity assessment index. Let be the electric field gradient vector. The heat flux gradient vector, and These are the electric field weighting coefficient and the heat flux weighting coefficient, respectively. Extract the set of spatial coordinates that are greater than the preset safety margin threshold from the comprehensive sensitivity evaluation index to divide the sensitive area; Within the sensitive area, a weighted iterative optimization algorithm is used to update the centroid offset and lock the optimal three-dimensional coordinates of the spatial index induced point.

[0018] in, Represents the optimal three-dimensional coordinates. To belong to the m-th candidate point sub-region The discrete spatial node coordinate vectors are used until the displacement of all candidate points is less than the preset convergence accuracy parameter. Based on the optimal three-dimensional coordinates, a normalized mapping operator matrix H is constructed. The formula for calculating the mapping weight element in the m-th row and j-th column is as follows:

[0019] in, The mapping weight element in the m-th row and j-th column of the normalized mapping operator matrix H is... Let J be the physical three-dimensional coordinates of the j-th sensor. A characteristic length scale parameter for controlling the range of attenuation of spatial physical correlation; The mapping operator matrix H is multiplied by the discrete state column vector y, which consists of the actual observation values ​​of each sensor, to generate a reduced-dimensional state data vector containing only L elements. :

[0020] in, This is a reduced-dimensional state data vector.

[0021] Furthermore, the reduced-dimensional state data is input into the system state matrix, and the random coefficient values ​​for the current sampling period are calculated based on the sparse Gaussian process regression model. Specific steps include: The transpose of the reduced-dimensional state data vector and the projection mapping matrix. Multiply to calculate the inner product response component vector of the dimensionality-reduced state data on the orthogonal feature functions:

[0022] in, The inner product response component vector, Let be the transpose of the projection mapping matrix, the projection mapping matrix It is composed of the discrete feature mapping values ​​of the orthogonal feature functions at each of the spatial index induced points; A ridge regression regularization penalty constant is introduced during the solution process. Solve for the vector of random coefficient values ​​for the current sampling period:

[0023] in, This is a vector of random coefficient values ​​for the current sampling period. Let I be the ridge regression regularization penalty constant, and let I be the identity matrix.

[0024] Furthermore, based on the linear combination of the orthogonal characteristic function and the random coefficient values, a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field are reconstructed. Specific steps include: A gridded coordinate system is established based on the three-dimensional physical boundary of the high-voltage system. The nth component in the random coefficient value vector is used as a weight, and is correlated with the orthogonal characteristic function of the corresponding order. The q-th grid coordinate node Linear superposition is performed at each point to reconstruct the initial spatial field values:

[0025] in, These are the initial spatial field values. To determine the total order of the truncated orthogonal characteristic functions, A vector of random coefficient values The nth component in The nth order orthogonal eigenfunction at the qth grid coordinate node The fundamental mapping operator at the location; Applying a spatial smoothing constraint based on the discrete Laplace mean operator to the initial spatial field values ​​generates smoothed, corrected, continuous three-dimensional distribution field values:

[0026] in, The numerical values ​​of the continuous three-dimensional distribution field after smoothing correction. The smoothing damping convergence coefficient is... To surround the central grid node The set of adjacent grid nodes, This represents the number of valid nodes contained in the set. These are the three-dimensional coordinates of adjacent grid nodes.

[0027] Furthermore, the estimated temperature and estimated air pressure values ​​are extracted from sensorless coverage nodes. Specific steps include: Extract the geometric envelope path of the weak point in the insulation of the high-voltage line harness and connector, and perform discretization sampling along the envelope path at a preset spatial step size to generate multiple virtual detection nodes; For any virtual detection node, the eight nearest neighboring spatial grid nodes are retrieved within the gridded coordinate system, and the corresponding estimated temperature value is calculated using the inverse distance weighted spatial interpolation algorithm. Compared with the estimated air pressure value :

[0028]

[0029] in, and These are the reconstructed temperature and pressure values ​​for the continuous field at the k-th adjacent grid node, respectively. The distance is the three-dimensional Euclidean distance between the virtual detection node and the kth adjacent grid node.

[0030] Furthermore, the estimated temperature and estimated air pressure values ​​are substituted into the gas discharge breakdown model to calculate the transient air critical breakdown voltage at the sensorless coverage node. Specific steps include: Based on the calculated temperature value Establish an ideal gas thermal expansion correction mechanism, combined with the calculated gas pressure value. Solving for the comprehensive equivalent air pressure under microscopic conditions :

[0031] in, To solve the comprehensive equivalent air pressure under microscopic environment, The absolute reference temperature is set under standard atmospheric physical conditions; Based on the geometry of the conductor at the sensorless coverage node, a pole gap correction factor is applied. For the original electrical clearance distance Make corrections to obtain the equivalent discharge electrode distance. :

[0032] Comprehensive equivalent air pressure and equivalent discharge electrode distance Input the Paschen gas discharge model and solve for the corresponding transient air critical breakdown voltage. :

[0033] in, Let be the transient critical breakdown voltage of air, A be the calibration constant of the air ionization collision cross section, and B be the physical threshold constant for gas ionization excitation. This represents the secondary electron emission coefficient of a specific insulating interface material.

[0034] Furthermore, the insulation safety margin is determined based on the critical breakdown voltage and the actual operating voltage, and an insulation warning signal is output when it falls below a preset threshold. Specific steps include: Based on the ratio of the transient air critical breakdown voltage to the high-voltage system operating voltage, calculate the insulation safety margin of the sensorless coverage node: Real-time acquisition of the high-voltage system operating voltage output to the DC bus by the battery management system under current operating conditions, and calculation of the dynamic insulation risk index:

[0035] in, This is a dynamic insulation risk index. This is the operating voltage of the high-voltage system; Combined with the calibrated safety tolerance benchmark threshold Calculate the insulation safety margin at the virtual detection node:

[0036] in, For insulation safety margin, The calibrated safety tolerance benchmark threshold; When the insulation safety margin of any of the sensorless coverage nodes is lower than a preset threshold, an insulation warning signal is output. The specific steps include: Establish a jitter prevention monitoring mechanism based on a time sliding window, and construct a Boolean decision function at the current time t. :

[0037] in, This is an indicator function that outputs 1 when the condition is true and 0 otherwise. For each discrete sampling moment within the current time sliding window, The length of the continuous sampling monitoring period. Threshold for determining security alarms; In response to When the output is 1, the judgment logic is triggered, and the insulation warning signal and the corresponding fault physical node coordinate information are output.

[0038] As can be seen from the above technical solution, the beneficial technical effects of the present invention are as follows: 1. By deeply coupling low air pressure with local high temperature effect, the transient air critical breakdown voltage of electrical gap is solved in real time, which breaks through the limitations of traditional fixed static threshold detection and significantly improves the detection rate of hidden dangers under complex coupled conditions such as severe long downhill slopes.

[0039] 2. By combining the spatial environment base and spatial induced point dimensionality reduction, the temperature and pressure parameters of nodes without physical sensor coverage can be accurately deduced with extremely low on-board computing power overhead, realizing accurate positioning and blind-spot-free monitoring of any weak insulation point in the entire vehicle.

[0040] 3. The introduction of a time-sliding window anti-jitter mechanism effectively suppresses random disturbances caused by sudden high-frequency vibrations, electromagnetic noise, and multi-channel signal delays, ensuring extremely high confidence and real-time performance of insulation safety warnings. Attached Figure Description

[0041] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0042] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the dynamic insulation early warning system according to an embodiment of the present invention. Detailed Implementation

[0043] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0044] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0045] This embodiment provides a high-voltage insulation testing method for new energy vehicles. The working principle of this embodiment is explained in detail below: The method flowchart of this embodiment is as follows: Figure 1 As shown, it includes the following steps: Acquire discrete environmental state data under driving conditions; Based on the discrete environmental state data and the pre-stored spatial environment base, the spatial field of the high-pressure system is reconstructed to obtain a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field. Extract the estimated temperature and estimated pressure values ​​of sensor-free nodes in the three-dimensional temperature distribution field and the three-dimensional air pressure distribution field; Substitute the estimated temperature value and the estimated air pressure value into the gas discharge breakdown model to calculate the transient air critical breakdown voltage at the sensorless coverage node. The insulation safety margin is determined based on the critical breakdown voltage and the actual operating voltage, and an insulation warning signal is output when it falls below a preset threshold.

[0046] In this embodiment, further obtaining discrete environmental state data under driving conditions specifically includes the following steps: The discrete environmental state data includes local real-time temperature data collected by high-pressure component sensors (sensors distributed at the battery, motor, and wiring harness ends) and global real-time air pressure data collected by air pressure sensors (air pressure sensors distributed on the chassis).

[0047] Because different sensors have different hardware sampling periods and communication delays, the system establishes a time synchronization link for multimodal physical signals, and sets the synchronization target time based on the global main loop clock of the vehicle controller. For asynchronously arriving discrete signals, a linear interpolation algorithm is used to calculate the reference mapping value at the target time, and the acquired raw state signals are then time-aligned. The synchronized state signal after time alignment... Represented as:

[0048] in, This is the synchronization status signal after timing alignment. Clearly define the corresponding real-time local temperature reference value or real-time global air pressure reference value after time synchronization. To synchronize with the target time, and For two adjacent actual sampling times, and At the actual sampling time and The collected values; In this embodiment, two adjacent actual sampling times and The actual sampling time difference between them is 100 milliseconds, and the synchronization target time is... It is the absolute midpoint between two adjacent sampling times to ensure strong spatiotemporal consistency of timing alignment.

[0049] To suppress the strong electromagnetic interference generated by high-voltage line harnesses during high-current step switching and the transient changes caused by high-frequency mechanical vibrations induced by vehicle traffic, the system uses a one-dimensional Kalman filter algorithm to smooth and denoise the time-aligned synchronization state signal. The one-dimensional state prior estimate at discrete time step k is defined as... The posterior state estimate is The system process noise covariance is Q, and the measurement noise covariance is R. The state prediction equation and prior estimation of the error covariance are presented in the filtering prediction stage. The calculation formula is:

[0050]

[0051] During the update phase, the system calculates the Kalman gain. And combined with the system observations at the current moment The state is corrected, and the observed value here is the timing synchronization state signal. The posterior estimate of the error covariance is then updated to... :

[0052]

[0053]

[0054] In this embodiment, the system process noise covariance Q is 0.01, and the measurement noise covariance R is 0.05.

[0055] After smoothing and denoising, the system correlates the denoised high-fidelity local real-time temperature data or global real-time air pressure data with the corresponding sensor spatial coordinates. Based on the three-dimensional physical topology model of the high-pressure system, the three-dimensional physical space coordinate vectors of each sensor are extracted. Here, 'i' represents the global index number of the sensors distributed across the motor, wiring harness, or battery. The denoised local real-time temperature data corresponding to the index number is then used. The denoised global real-time air pressure data P is numerically concatenated with the spatial coordinate vector, and a discrete environmental state data matrix M is constructed by traversing all sensor nodes of the vehicle.

[0056] Where M is the discrete environmental state data matrix, N is the total number of deployed sensors, and the feature column vector is... , Let be the three-dimensional physical space coordinate vector of the i-th sensor. P represents the local real-time temperature data after smoothing and denoising, and P represents the global real-time air pressure data after smoothing and denoising. In this embodiment, the total number of sensors N is 12.

[0057] In this embodiment, further, based on the discrete environmental state data and the pre-stored spatial environment substrate, the high-pressure system is reconstructed into a spatial field to obtain a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field. Specific steps include: The space environment base of the high-voltage system is invoked, and the space environment base of the high-voltage system contains a preset number of orthogonal characteristic functions; In the three-dimensional physical topology model of the high-voltage system, spatial index induced points are preset, and the discrete environmental state data are sparsely approximated using the spatial index induced points to generate dimensionality-reduced state data. The reduced state data is input into the system state matrix, and the random coefficient values ​​of the current sampling period are calculated based on the sparse Gaussian process regression model. Based on the linear combination of the orthogonal characteristic function and the random coefficient value, a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field are reconstructed.

[0058] The construction of the space environment substrate for the high-voltage system specifically includes the following steps: A dataset containing historical alternating operating condition tests was extracted. This dataset covers historical environmental state variables collected from new energy vehicles under alternating environmental stresses such as different altitudes and electric drive loads. The coordinate vectors of any two spatial nodes in the three-dimensional physical topology model of the high-voltage system are defined as follows: and Let Z(s,t) be the physical state measurement value corresponding to the spatial coordinates and time in the test history dataset. This represents the average historical measurement time at this spatial node. The system constructs a spatial covariance matrix C by calculating the expected correlation between spatial nodes, where the elements within the matrix are... The calculation formula is:

[0059] Where T is the total number of time sampling points in the test historical dataset. In this embodiment, the total number of time sampling points T is 10,000. It clearly characterizes the correlation and coupling strength of different physical locations in the high-pressure system during the heat flow and pressure transfer process.

[0060] Eigenvalue decomposition is performed on the spatial covariance matrix C to obtain the eigenvalue sequence and the original set of eigenfunctions corresponding to the environmental random field. Based on spatial random field theory and the KL (Karhunen-Loève) expansion method, the eigenvalue equation of the spatial covariance matrix is ​​solved to obtain the orthogonal basis:

[0061] in, The nth order eigenvalue obtained from the decomposition. Let be the eigenvector corresponding to the nth eigenvalue, representing a fundamental spatial distribution mode in the original eigenfunction set. Eigenvalue The magnitude of the value reflects the energy proportion of the corresponding fundamental spatial distribution mode in the global physical field fluctuations.

[0062] To achieve model dimensionality reduction while preserving core physical field features, the system sorts the eigenvalue sequence by eigenvalue magnitude and calculates the cumulative variance contribution rate of the first K eigenvalues. :

[0063] in, The cumulative variance contribution rate is denoted as M, where M is the total order of the spatial covariance matrix, i.e., the total number of spatial nodes participating in the decomposition. In this embodiment, M is order 256. The system sets a preset variance contribution threshold. When the variance contribution threshold is first reached or equal to the variance contribution threshold, the top K principal components at that time are truncated and retained. In this embodiment, the preset variance contribution threshold is 95%, and the number of the top K principal components retained is 5. The orthogonal eigenvectors corresponding to these K largest eigenvalues ​​are... to The combination constitutes the pre-stored high-voltage system space environment base.

[0064] The orthogonal characteristic function, i.e., the characteristic vector The mathematical essence of the covariance operator is the eigenfunction of the covariance operator. Its physical meaning lies in describing the inherent spatial state distribution modes of a high-pressure system determined by its physical boundaries such as its geometric shape, heat conduction path, and air flow channel. The first-order low-order modes reflect the macroscopic trend of global temperature or air pressure, while the high-order modes reflect the abrupt change of local gradient. The inner product of each mode in space is 0, which removes redundant information and can highly reconstruct complex continuous physical fields with a low computational dimension.

[0065] In this embodiment, spatial index induced points are preset in the three-dimensional physical topology model of the high-voltage system. These spatial index induced points are then used to perform sparse approximation processing on the discrete environmental state data to generate dimensionality-reduced state data. Specific steps include: Analyze the electric field and heat flux distribution data of each spatial node in the three-dimensional physical topology model of a high-voltage system. Define the spatial coordinate vector as... The system extracts the steady-state electric field intensity scalar field E(p) and steady-state heat flux scalar field T(p) at the corresponding coordinate locations. The system calculates the spatial gradient vectors of the electric field and heat flux at these coordinates and obtains their L2 norm to quantify the degree of abrupt change in local field strength. Based on this, the system constructs a comprehensive sensitivity evaluation index W(p) characterizing the degree of insulation stress concentration and environmental abrupt changes:

[0066] Wherein, W(p) is the comprehensive sensitivity assessment index. Let be the electric field gradient vector. The heat flux gradient vector, and These are the electric field weighting coefficient and the heat flow weighting coefficient, respectively, calibrated based on the physical dielectric properties and thermal conductivity properties of the insulating material.

[0067] In this embodiment, the electric field weighting coefficient The heat flux weighting coefficient is 0.6. The threshold is 0.4. The system traverses the entire vehicle topology space, extracts all spatial coordinate sets whose comprehensive sensitivity evaluation index W(p) values ​​are greater than the preset safety margin threshold, and divides them into gas-solid-electric-thermal sensitive regions of insulation stress concentration and environmental abrupt changes. In this embodiment, the preset safety margin threshold is 0.8.

[0068] Within the identified sensitive area, the system dynamically allocates candidate points using a weighted iterative optimization algorithm to lock in the optimal three-dimensional coordinates of the spatial index induced point. Set the total number of spatial index guidance points preset by the system to L, and initialize the coordinates of L candidate points. And randomly distributed within the sensitive area. In each iteration, the system distributes each discrete spatial node within the sensitive area. Subregions are defined by the candidate point with the closest Euclidean distance. In addition, it strictly utilizes comprehensive sensitivity assessment indicators. As a gravity weight constraint, the 3D coordinates of each candidate point are updated by centroid offset. (Candidate point 3D coordinates) The iterative update calculation formula is as follows:

[0069] When the displacement of all candidate points in two consecutive iterations is less than the preset convergence accuracy parameter, an iterative optimization termination command is triggered, and the final output is... This refers to the optimal three-dimensional coordinates of the spatial indexing induced points. In this embodiment, the total number of spatial indexing induced points L is 8, and the preset convergence accuracy parameter is 0.5 mm.

[0070] Based on the eight optimal 3D coordinates obtained, the system constructs a spatial mapping operator driven by a radial basis function kernel to project the discrete environmental state data to spatial indexed induced points in a dimension-reduced manner. The system calculates the optimal 3D coordinates of the induced points. Physical coordinates of each sensor The spatial Gaussian kernel distance between them is used to construct a kernel-normalized mapping operator matrix H, where the mapping weight element in the m-th row and j-th column of this matrix is... The calculation formula is:

[0071] in, The mapping weight element in the m-th row and j-th column of the normalized mapping operator matrix H is... Let J be the physical three-dimensional coordinates of the j-th sensor. A characteristic length scale parameter for controlling the range of attenuation of spatial physical correlation; In this embodiment, the feature length scale parameter The value is 150 mm. The system maps the spatial mapping operator matrix H to the actual observation values ​​from each sensor. The discrete state column vector y is linearly multiplied to compress and map the high-dimensional sensor state observation data into a reduced-dimensional state data vector containing only 8 elements. :

[0072] in, This is a reduced-dimensional state data vector.

[0073] This step centralizes and mathematically compresses the massive, unevenly distributed sensor data of the entire vehicle into a few high-risk physical stress core nodes, thus reducing the matrix operation workload of subsequent reconstruction algorithms while fully preserving the characteristics of local insulation deterioration.

[0074] In this embodiment, the reduced-dimensional state data is further input into the system state matrix, and the random coefficient values ​​of the current sampling period are calculated based on the sparse Gaussian process regression model. Specific steps include: First, a sparse Gaussian process regression model incorporating prior observational knowledge is constructed. However, conventional Gaussian process regression requires processing the entire dataset of massive spatial node data across the entire vehicle. The curse of dimensionality caused by performing the inversion operation on the covariance matrix. To address the issue of time complexity, the sparse Gaussian process regression model constructed in this invention refers to... The optimal set of three-dimensional coordinates of a spatial indexed induced point is used as a sparse approximate induced set to replace the full set of observation nodes in constructing a surrogate model of the spatial autocorrelation distribution of the target physical field. This model will rely on the full set of observation nodes. The inference process from individual sensor data can be mathematically transformed into relying only on a much smaller number of sensors. of The variational lower bound optimization solution at each induced point significantly reduces the computational complexity of inverting the system state matrix to [value missing]. This ensures that highly real-time online parameter inference can be achieved under limited onboard computing power.

[0075] Based on the sparse Gaussian process regression model architecture defined above, the system calculates the discrete feature mapping values ​​of each order of orthogonal feature function in the pre-stored spatial environment basis at each spatial index induced point, forming a projection mapping matrix. Dimensionally reduced state data vector Transformation by projection and the transpose of the projection mapping matrix Multiplying these components yields the inner product response vector v of the reduced-dimensional state data on the orthogonal eigenfunctions:

[0076] in, The inner product response component vector, Let be the transpose of the projection mapping matrix, the projection mapping matrix It is composed of the discrete feature mapping values ​​of the orthogonal feature functions at each of the spatial index induced points; To address the linear dependence problem that may be caused by the characteristic function at local discrete induced points, the system introduces a feature-dependent linear relationship during the solution of the system state matrix. The objective function equation for the regularization penalty term is used to decouple collinearity interference and inversely output the random coefficient value vector for the current sampling period. The solution formula is:

[0077] in, This is a vector of random coefficient values ​​for the current sampling period. Let be the ridge regression regularization penalty constant, and I be the identity matrix. In this embodiment, the ridge regression regularization penalty constant... It is 0.001.

[0078] In this embodiment, further, based on the linear combination of the orthogonal characteristic function and the random coefficient value, a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field are reconstructed. Specific steps include: After calculating the random coefficient value vector Subsequently, based on the three-dimensional physical boundary geometry of the high-voltage system's power battery casing and high-voltage wiring harness routing path, the system establishes a three-dimensional continuous gridded coordinate system with a preset spatial resolution. The system extracts the three-dimensional spatial coordinate vectors of all internal spatial grid nodes and uses the obtained random coefficient values... As a weighting variable, it is an orthogonal characteristic function of the corresponding order. Perform a linear superposition operation at the physical space level within a three-dimensional mesh coordinate system to generate an initial spatial field of three-dimensional temperature and pressure. The reconstruction formula is as follows:

[0079] in, For the q-th spatial grid coordinate node The initial spatial field physical reconstruction value is given by K, where K is the total order of the orthogonal characteristic functions in the truncated and preserved environmental basis calling module, which is order 5 in this embodiment. A vector of random coefficient values The nth weighted coefficient component in.

[0080] To address the high-frequency numerical oscillations at the grid boundaries caused by discrete feature expansion, the system applies a spatial smoothing constraint based on the discrete Laplace average operator to the initial spatial field generated by superposition. This eliminates step distortion between adjacent grid nodes, ultimately generating a smooth and continuous three-dimensional temperature and pressure distribution field. The smoothing correction formula is as follows:

[0081] in, The numerical values ​​of the continuous three-dimensional distribution field after smoothing correction. The smoothing damping convergence coefficient is... To surround the central grid node The set of adjacent grid nodes, This represents the number of valid nodes contained in the set. These represent the three-dimensional coordinates of adjacent grid nodes. In this embodiment, the smoothing damping convergence coefficient... It is 0.15.

[0082] In this embodiment, the calculated temperature value and the calculated air pressure value are further substituted into the gas discharge breakdown model to calculate the transient air critical breakdown voltage at the sensorless coverage node. Specific steps include: Extracting the estimated temperature and estimated air pressure values ​​from sensorless nodes involves the following steps: The geometric envelope path of the weak insulation points in the high-voltage wiring harness and connectors is extracted, and a target topology set without physical sensors is generated based on the three-dimensional digital model of the vehicle's high-voltage system. The system performs discretization sampling along the spatial extension direction of this target topology set at a pre-set equidistant spatial step size, generating three-dimensional spatial coordinate vectors of multiple virtual detection nodes. , where i represents the i-th virtual detection node without sensor coverage. In this embodiment, the preset equidistant spatial step size is 10 mm.

[0083] For each virtual detection node, the system uses an inverse distance weighted spatial interpolation algorithm to map its coordinates to the reconstructed continuous three-dimensional temperature and pressure distribution fields. The system retrieves the eight nearest neighboring spatial grid nodes within the gridded coordinate system and performs a weighted calculation to obtain the estimated temperature value corresponding to that virtual detection node. Compared with the estimated air pressure value The interpolation formulas are as follows:

[0084]

[0085] in, and These are the reconstructed temperature and pressure values ​​for the continuous field at the k-th adjacent spatial grid node, respectively. Let be the three-dimensional Euclidean distance between the i-th virtual detection node and the k-th adjacent spatial grid node.

[0086] In this embodiment, the calculated temperature value and the calculated air pressure value are further substituted into the gas discharge breakdown model to calculate the transient air critical breakdown voltage at the sensorless coverage node. Specific steps include: After acquiring the environmental parameters of each virtual detection node, the system calculates the thermal expansion correction factor for local air density based on the estimated temperature value, and combines this with the estimated air pressure value to obtain the comprehensive equivalent air pressure under the microscopic environment. Based on the physical compensation mechanism established by the ideal gas law, the actual air pressure is dynamically converted into an effective gas density pressure characterization quantity affected by temperature thermal expansion, resulting in the comprehensive equivalent air pressure. The solution formula is:

[0087] in, To solve the comprehensive equivalent air pressure under microscopic environment, The absolute reference temperature is set under standard atmospheric physical conditions. In this embodiment, the absolute reference temperature is... It is 293.15 Kelvin.

[0088] The system further evaluates the electric field distribution characteristics of the high-voltage conductor geometry at each sensorless coverage node, and calculates the electrode gap correction coefficient, which characterizes the microscopic discharge path distortion, based on the conductor tip curvature gradient. The system will determine the original electrical clearance distance at the physical node. With polar moment correction factor Multiply to obtain the equivalent discharge electrode distance in the boundary dynamic compensation quantity. The calculation formula is:

[0089] In this embodiment, the original electrical clearance distance The polar distance correction factor is 5.5 mm. It is 1.2.

[0090] Furthermore, the obtained comprehensive equivalent air pressure With equivalent discharge electrode distance By synchronously inputting the parameter-calibrated Paschen gas discharge model, the transient air critical breakdown voltage of the sensorless coverage node under the current state is solved. Its nonlinear calculation formula is:

[0091] in, Let be the transient critical breakdown voltage of air, A be the calibration constant of the ionization collision cross section of the air mixture, and B be the threshold constant of the physical properties of gas ionization excitation. The secondary electron emission coefficient is the specific insulating interface material. In this embodiment, the calibration constant A is 11.25, the physical threshold constant B is 273.75, and the secondary electron emission coefficient... It is 0.01.

[0092] In this embodiment, further, the insulation safety margin is determined based on the critical breakdown voltage and the actual operating voltage, and an insulation warning signal is output when it falls below a preset threshold. Specific steps include: Based on the ratio of the transient air critical breakdown voltage to the high-voltage system operating voltage, calculate the insulation safety margin of the sensorless coverage node: The system uses the vehicle control local area network bus (CAN bus) communication interface to read in real time the actual high-voltage system operating voltage output by the battery management system (BMS) to the DC bus under the current driving load requirements. In this embodiment, the high-voltage system operating voltage It is 800 volts.

[0093] To prevent air ionization discharge, the system calculates the ratio of the relative difference between the transient air critical breakdown voltage and the high-voltage system operating voltage to assess the proximity of insulation failure at the current physical node. For the i-th virtual detection node without sensor coverage, the system extracts the transient air critical breakdown voltage calculated previously. And calculate the dynamic insulation risk index of the node in the current spatial heterogeneous environment. :

[0094] The dynamic insulation risk index It is a dimensionless ratio. The smaller the value, the closer the actual working voltage that the node is currently subjected to is to the dynamic physical breakdown limit caused by severe temperature and air pressure, and the higher the probability of high-voltage arc flashover.

[0095] After obtaining the dynamic insulation risk index of each virtual detection node, the system performs a nonlinear comparison with a pre-set safety tolerance benchmark parameter to quantitatively characterize the remaining safety buffer range. Let the system-calibrated safety tolerance benchmark threshold be... This benchmark is primarily established based on initial design margins and lifetime verification data for high-voltage system insulation materials. The system calculates the insulation safety margin for this node. The calculation formula is:

[0096] in, For insulation safety margin, In this embodiment, the safety tolerance benchmark threshold is used as the calibrated safety tolerance benchmark threshold. It is 0.25.

[0097] In this embodiment, furthermore, when the insulation safety margin of any of the sensorless coverage nodes is lower than a preset threshold, an insulation warning signal is output. Specific steps include: To avoid false alarms caused by transient signal fluctuations from sensors or frame drops in communication data, the system introduces a continuous state monitoring anti-jitter mechanism based on a time-sliding window. The system's preset safety alarm judgment threshold is set to... The continuous sampling monitoring period window length is During real-time sampling of vehicle operation, the system utilizes a historical state cache queue to monitor the evolution trend of insulation safety margin at each node. The system constructs a Boolean decision function. At the current time t, the i-th node is within its corresponding continuous historical window, i.e., from time t... Insulation safety margin obtained up to time t All are strictly below the preset threshold. When the judgment logic is triggered, the alarm function state is reversed:

[0098] in, This is an indicator function that outputs a logical value of 1 when the logical condition within the parentheses is true, and a logical value of 0 otherwise. The length of the continuous sampling monitoring period. Threshold for determining safety alarms.

[0099] In this embodiment, the security alarm determination threshold The value is 0.05, representing the length of the continuous sampling monitoring period window. The sampling period is 10 cycles, with a time difference of 100 milliseconds between adjacent samplings, meaning the window monitoring duration is 1 second. The Boolean function for the alarm of any virtual detection node within the entire vehicle range... The output is 1, which is a logical true value. This means that when the insulation safety margin is lower than the danger limit of 0.05 for one consecutive second, the system immediately sends the highest level insulation warning signal to the vehicle controller and instrument panel, prompting the driver to reduce the kinetic energy recovery power or stop to check, and accurately reports the three-dimensional coordinate information of the corresponding fault physical node, so as to realize proactive early warning and accurate fault location.

[0100] Combination Figure 2 The following is an overall description of the dynamic cyclic early warning execution process based on the high-voltage insulation detection method in this embodiment: Step 1: The system acquires and synchronously denoises and cleans discrete environmental state data in real time through a time synchronization link of multimodal physical signals and a Kalman filter algorithm; Step 2: Call the pre-stored spatial environment base, use spatial index induced points in the three-dimensional physical topology model to perform sparse dimensionality reduction of high-dimensional data, and reconstruct the high-resolution continuous three-dimensional temperature distribution field and three-dimensional air pressure distribution field through the sparse Gaussian process regression model. Step 3: Discretize sampling in areas with weak insulation, such as wire harnesses and connectors, to generate virtual detection nodes without physical sensors, and use the inverse distance weighted spatial interpolation algorithm to deduce and calculate the transient air critical breakdown voltage at each virtual detection node; Step 4: Calculate the insulation safety margin at each virtual detection node by combining the actual high-voltage system operating voltage output by the vehicle battery management system; Step 5: Activate anti-shake monitoring and judgment based on time sliding window: In real time, determine whether the judgment condition is met that the insulation safety margin is less than 0.05 for 10 consecutive sampling cycles, i.e., within 1 second. In response to a "yes" judgment result, the judgment logic is triggered, the alarm function state is reversed, the system immediately triggers the highest level of insulation warning, sends an alarm to the vehicle controller and instrument panel, and outputs the specific three-dimensional coordinates of the high-risk fault point, prompting the driver to take safety measures or stop for inspection. If the judgment result is "No", that is, the insulation safety margin is not lower than 0.05 for 10 consecutive cycles, indicating that the current state is safe or that it is only an occasional transient signal fluctuation interference from the sensor, the system will not trigger an alarm, and will execute feedback control, return to the "data acquisition and synchronization cleaning" step, and start the cyclic monitoring of the next sampling cycle.

[0101] Through the aforementioned closed-loop dynamic cyclic monitoring process, this invention not only ensures accurate location of insulation hazards but also greatly improves the system's anti-disturbance capability and early warning confidence under extremely complex driving conditions.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for testing high-voltage insulation in new energy vehicles, characterized in that, Includes the following steps: Acquire discrete environmental state data under driving conditions; Based on the discrete environmental state data and the pre-stored spatial environment base, the spatial field of the high-pressure system is reconstructed to obtain a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field. Extracting the estimated temperature and estimated pressure values ​​for sensorless nodes in the three-dimensional temperature and pressure distribution fields involves the following steps: Extract the geometric envelope path of the weak point in the insulation of the high-voltage line harness and connector, and perform discretization sampling along the envelope path at a preset spatial step size to generate multiple virtual detection nodes; A gridded coordinate system is established based on the three-dimensional physical boundary of the high-voltage system; For any virtual detection node, the eight nearest neighboring spatial grid nodes are retrieved within the gridded coordinate system, and the corresponding estimated temperature value is calculated using the inverse distance weighted spatial interpolation algorithm. Compared with the estimated air pressure value : in, and These are the reconstructed temperature and pressure values ​​for the continuous field at the k-th adjacent grid node, respectively. The three-dimensional Euclidean distance between the i-th virtual detection node and the k-th adjacent grid node; Substituting the estimated temperature and estimated air pressure values ​​into the gas discharge breakdown model, the transient air critical breakdown voltage at the sensorless node is calculated. Specific steps include: Based on the calculated temperature value Establish an ideal gas thermal expansion correction mechanism, combined with the calculated gas pressure value. Solving for the comprehensive equivalent air pressure under microscopic conditions : in, To solve the comprehensive equivalent air pressure under microscopic environment, The absolute reference temperature is set under standard atmospheric physical conditions; Based on the geometry of the conductor at the sensorless coverage node, a pole gap correction factor is applied. For the original electrical clearance distance Make corrections to obtain the equivalent discharge electrode distance. : Comprehensive equivalent air pressure and equivalent discharge electrode distance Input the Paschen gas discharge model and solve for the corresponding transient air critical breakdown voltage. : in, Let be the transient critical breakdown voltage of air, A be the calibration constant of the air ionization collision cross section, and B be the physical threshold constant for gas ionization excitation. The secondary electron emission coefficient of a specific insulating interface material; The insulation safety margin is determined based on the critical breakdown voltage and the actual operating voltage, and an insulation warning signal is output when it falls below a preset threshold.

2. The high-voltage insulation testing method for new energy vehicles according to claim 1, characterized in that, The specific steps for obtaining discrete environmental state data under driving conditions include: The discrete environmental state data includes local real-time temperature data and global real-time air pressure data; A time synchronization link for the multimodal physical signals is established, and the acquired raw state signals are time-aligned. The time-aligned synchronization state signal is represented as follows: in, This is the synchronization status signal after timing alignment. To synchronize with the target time, and For two adjacent actual sampling times, and At the actual sampling time and The collected values; The Kalman filter algorithm is used to smooth and denoise the synchronization state signal after time alignment to obtain the smoothed and denoised local real-time temperature data and global real-time air pressure data. The smoothed and denoised signal is correlated with the corresponding sensor spatial coordinates to form a discrete environmental state data matrix: Where M is the discrete environmental state data matrix, N is the total number of deployed sensors, and the feature column vector is... , Let be the three-dimensional physical space coordinate vector of the i-th sensor. P represents the local real-time temperature data after smoothing and denoising, and P represents the global real-time air pressure data after smoothing and denoising.

3. The high-voltage insulation testing method for new energy vehicles according to claim 1, characterized in that, Based on the discrete environmental state data and the pre-stored space environment substrate, the high-pressure system is reconstructed to obtain a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field. Specific steps include: The space environment base of the high-voltage system is invoked, and the space environment base of the high-voltage system contains a preset number of orthogonal characteristic functions; In the three-dimensional physical topology model of the high-voltage system, spatial index induced points are preset, and the discrete environmental state data are sparsely approximated using the spatial index induced points to generate dimensionality-reduced state data. The reduced state data is input into the system state matrix, and the random coefficient values ​​of the current sampling period are calculated based on the sparse Gaussian process regression model. Based on the linear combination of the orthogonal characteristic function and the random coefficient value, a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field are reconstructed.

4. The high-voltage insulation testing method for new energy vehicles according to claim 3, characterized in that, The construction steps of the space environment substrate of the high-voltage system include: Obtain a dataset containing the test history under alternating operating conditions, and calculate the coordinate vectors of any two spatial nodes. and The expected correlation between them is used to construct a spatial covariance matrix C, whose internal elements are... The calculation formula is: in, Let T be the internal element of the spatial covariance matrix C, and let T be the total number of time sampling points in the test history dataset. To test the physical state measurements at corresponding spatial coordinates and times in the historical dataset, For spatial nodes The historical average measurement time at that location; Perform eigenvalue decomposition on the spatial covariance matrix C, solve the characteristic equation, and obtain the orthogonal characteristic functions and the corresponding eigenvalue sequence: in, For the nth order eigenvalue, The eigenvector corresponding to the nth order eigenvalue represents the orthogonal eigenfunction; Contribution rates are sorted in descending order of eigenvalues: in, The cumulative variance contribution rate of the first K eigenvalues, where M is the total number of nodes in the decomposition space. When the variance contribution is first greater than or equal to a preset variance contribution threshold, the feature vectors are truncated, retaining only the first K eigenvectors. to The space environment substrate that constitutes the high-voltage system is combined.

5. The high-voltage insulation testing method for new energy vehicles according to claim 3, characterized in that, The step of pre-setting spatial index induced points in the three-dimensional physical topology model of the high-voltage system, and using the spatial index induced points to perform sparse approximation processing on the discrete environmental state data to generate dimensionality-reduced state data includes: Analysis of spatial nodes in the three-dimensional physical topology model Based on the electric field gradient and heat flux gradient, a comprehensive sensitivity assessment index is constructed to characterize the degree of insulation stress concentration and environmental abrupt changes: Wherein, W(p) is the comprehensive sensitivity assessment index. Let be the electric field gradient vector. The heat flux gradient vector, and These are the electric field weighting coefficient and the heat flux weighting coefficient, respectively. Extract the set of spatial coordinates that are greater than the preset safety margin threshold from the comprehensive sensitivity evaluation index to divide the sensitive area; Within the sensitive area, a weighted iterative optimization algorithm is used to update the centroid offset and lock the optimal three-dimensional coordinates of the spatial index induced point. in, Represents the optimal three-dimensional coordinates. To belong to the m-th candidate point sub-region The discrete spatial node coordinate vectors are used until the displacement of all candidate points is less than the preset convergence accuracy parameter. Based on the optimal three-dimensional coordinates, a normalized mapping operator matrix H is constructed. The formula for calculating the mapping weight element in the m-th row and j-th column is as follows: in, The mapping weight element in the m-th row and j-th column of the normalized mapping operator matrix H is... Let J be the physical three-dimensional coordinates of the j-th sensor. A characteristic length scale parameter for controlling the range of attenuation of spatial physical correlation; The mapping operator matrix H is multiplied by the discrete state column vector y, which consists of the actual observation values ​​of each sensor, to generate a reduced-dimensional state data vector containing only L elements. : in, This is a reduced-dimensional state data vector.

6. The high-voltage insulation testing method for new energy vehicles according to claim 5, characterized in that, The reduced state data is input into the system state matrix, and the random coefficient values ​​for the current sampling period are calculated based on the sparse Gaussian process regression model. The specific steps include: The transpose of the reduced-dimensional state data vector and the projection mapping matrix. Multiply to calculate the inner product response component vector of the dimensionality-reduced state data on the orthogonal feature functions: in, The inner product response component vector, Let be the transpose of the projection mapping matrix, the projection mapping matrix It is composed of the discrete feature mapping values ​​of the orthogonal feature functions at each of the spatial index induced points; A ridge regression regularization penalty constant is introduced during the solution process. Solve for the vector of random coefficient values ​​for the current sampling period: in, This is a vector of random coefficient values ​​for the current sampling period. Let I be the ridge regression regularization penalty constant, and let I be the identity matrix.

7. The high-voltage insulation testing method for new energy vehicles according to claim 6, characterized in that, Based on the linear combination of the orthogonal characteristic function and the random coefficient value, a continuous three-dimensional temperature distribution field and a three-dimensional air pressure distribution field are reconstructed. The specific steps include: The nth component in the random coefficient value vector is used as the weight, and then compared with the orthogonal characteristic function of the corresponding order. The q-th grid coordinate node Linear superposition is performed at each point to reconstruct the initial spatial field values: in, These are the initial spatial field values. To determine the total order of the truncated orthogonal characteristic functions, A vector of random coefficient values The nth component in The nth order orthogonal eigenfunction at the qth grid coordinate node The fundamental mapping operator at the location; Applying a spatial smoothing constraint based on the discrete Laplace mean operator to the initial spatial field values ​​generates smoothed, corrected, continuous three-dimensional distribution field values: in, The numerical values ​​of the continuous three-dimensional distribution field after smoothing correction. The smoothing damping convergence coefficient is... To surround the central grid node The set of adjacent grid nodes, This represents the number of valid nodes contained in the set. These are the three-dimensional coordinates of adjacent grid nodes.

8. The high-voltage insulation testing method for new energy vehicles according to claim 1, characterized in that, The insulation safety margin is determined based on the critical breakdown voltage and the actual operating voltage, and an insulation warning signal is output when it falls below a preset threshold. The specific steps include: Based on the ratio of the transient air critical breakdown voltage to the high-voltage system operating voltage, calculate the insulation safety margin of the sensorless coverage node: Real-time acquisition of the high-voltage system operating voltage output to the DC bus by the battery management system under current operating conditions, and calculation of the dynamic insulation risk index: in, This is a dynamic insulation risk index. This is the operating voltage of the high-voltage system; Combined with the calibrated safety tolerance benchmark threshold Calculate the insulation safety margin at the virtual detection node: in, For insulation safety margin, The calibrated safety tolerance benchmark threshold; When the insulation safety margin of any of the sensorless coverage nodes is lower than a preset threshold, an insulation warning signal is output. The specific steps include: Establish a jitter prevention monitoring mechanism based on a time sliding window, and construct a Boolean decision function at the current time t. : in, This is an indicator function that outputs 1 when the condition is true and 0 otherwise. For each discrete sampling moment within the current time sliding window, The length of the continuous sampling monitoring period. Threshold for determining security alarms; In response to When the output is 1, the judgment logic is triggered, and the insulation warning signal and the corresponding fault physical node coordinate information are output.

Images