Artificial contamination test method under high voltage environment

Abstract

The present application relates to the field of high-voltage engineering and insulation technology, and particularly relates to a method for artificial contamination test under high-voltage environment, which comprises the following steps: firstly, constructing a target deposition distribution function and a target response set on a parameterized grid on the surface of an insulator, and merging an environmental boundary condition time sequence set to form a target deposition and response data set; then, implementing layered deposition of soluble contaminants and insoluble particles under the electrode system time-varying electric field and particle charging control in the on-power state, collecting surface state field to generate an in-situ deposition state data set; then, obtaining electrical and surface responses and estimating a median flashover voltage according to segmented loading and stepwise flashover test to form an electrical and physical response data set; finally, based on the kinetic model, closed-loop optimization is performed on the electrode system time-varying trajectory, the environmental program and the deposition spectrum within the leakage current constraint. The present application realizes reproducible configuration, stable threshold estimation and fast test convergence.

Description

Technical Field

[0001] This invention relates to the fields of high voltage engineering and insulation technology, and in particular to artificial pollution test methods under high voltage environments. Background Technology

[0002] Coastal, wind-blown, and industrial areas of the power grid face widespread pollution risks from the combined effects of soluble salts and insoluble particles. Pollution alters surface conductivity and creepage paths during wetting and drying cycles, inducing increased leakage current and flashover tripping, impacting power supply reliability and equipment lifespan. Laboratory-based artificial pollution tests are fundamental for operation and maintenance grading, insulator type selection, and pollution prevention strategy verification; however, they have long suffered from problems such as unreproducible configurations, inconsistent statistical standards, and difficulty in quantifying safety boundaries, hindering the extrapolation of test results to engineering applications. Summary of the Invention

[0003] To address the numerous problems existing in the prior art, this invention provides an artificial contamination test method under high pressure. This invention uses a parameterized grid as a reference, first defines the target deposition and response, then writes the configuration in layers with a time-varying electric field and charged particles under power-on conditions, subsequently obtains the median flashover voltage through piecewise loading and statistical estimation, and finally optimizes the electrode trajectory, environmental program, and deposition spectrum within the leakage current constraint in a closed loop. The results are that the target configuration is reproducible, the threshold estimation is stable, and the test converges quickly.

[0004] A method for artificial contamination testing under high pressure includes the following steps:

[0005] A parameterized mesh for the insulator surface is established, the target deposition distribution function and target response set are generated, and the environmental boundary condition time series set is merged to obtain the target deposition and response dataset;

[0006] In the energized state, a time-varying electric field distribution is formed through the electrode system, soluble pollutant aerosols and insoluble particulate aerosols are supplied to the surface of the insulator and particulate charge control is implemented. Based on the time series set of environmental boundary conditions and the supply switching, a layered pollutant deposition structure is constructed, surface state field data is collected, and an in-situ deposition state dataset is generated.

[0007] Loading is performed based on the time series set of environmental boundary conditions and voltage condition changes. The time series of electrical response and surface state field are collected. The median flashover voltage is obtained by flashover test and statistical methods. The relevant increments are calculated to generate an electro-physical response dataset.

[0008] Within the leakage current constraint, based on the electrophysiological response dataset and the target deposition and response dataset, a dynamic model of surface state and migration behavior is established with the deviation indices of deposition configuration and conductive connectivity, and electrical response and median flashover voltage as optimization objectives. The time series set of time-varying trajectory of electrode system and environmental boundary conditions is updated, the pollutant spectrum parameter set is adjusted, and the convergence parameter set is output.

[0009] Preferably, the environmental boundary condition time series set consists of relative humidity time series, temperature time series, wind speed time series and pollutant deposition flux time series, and is sampled and aligned with a unified time reference as input to the target deposition and response dataset.

[0010] Preferably, the electrode system is a multi-electrode shaping array, with the phase and amplitude of the electrode channels set independently, and the electrode channels arranged with a fixed geometric spacing, forming a time-varying electric field distribution that moves along the axial and radial directions of the insulator through sequential driving.

[0011] Preferably, particle charge control employs corona charging, with the charge polarity being either unipolar positive or unipolar negative. The particle charge is measured using a Faraday cylinder in conjunction with an electrostatic meter, and the measured value is used to correct the aerosol supply parameters.

[0012] Preferably, the construction sequence of the layered pollutant deposition structure is as follows: first, insoluble particulate aerosols are supplied to form a particulate skeleton; then, in the wetting stage, the relative humidity is increased and soluble pollutant aerosols are supplied to form a coating layer; and finally, in the drying stage, the relative humidity is reduced to stabilize the layered structure.

[0013] Preferably, the surface state field data includes a water film thickness field, a surface equivalent conductivity field, and a soluble pollutant surface concentration field. The water film thickness field is obtained by non-contact optical measurement and conversion, the surface equivalent conductivity field is obtained by segmented measurement using a multi-point electrode array and inversion, and the soluble pollutant surface concentration field is obtained by either micro-elution conductivity measurement or spectral measurement.

[0014] Preferably, the loading is performed in a segmented manner. The loading sequence includes at least three types of segments: relative humidity step, temperature step, wind speed change, pollutant deposition flux pulse, and voltage transient. The start and end times of each segment are recorded with time stamps and synchronized with the timestamp of the acquisition channel.

[0015] Preferably, the median flashover voltage is obtained through a step-up flashover test, with the voltage steps for rising and falling being consistent, and statistical regression is used to estimate the results of multiple tests.

[0016] Preferably, the deviation index between deposition configuration and conductive connectivity is measured after obtaining the conductivity map by thresholding the surface equivalent conductivity field. The difference in deposition configuration is measured by the optimal transmission distance, and the difference in conductive connectivity is measured by the persistent coherence bottleneck distance.

[0017] Preferably, the dynamic model of surface state and migration behavior is established through sparse structure identification, and the candidate basis function set and state derivative are sparsely regressed to obtain the analytical form. The time-varying trajectory of the electrode system is updated after obtaining the gradient by the adjoint method. The set of pollutant spectrum parameters is adjusted by introducing feasible region constraints of upper limit of leakage current and upper limit of leakage current change rate and using Bayesian optimization.

[0018] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0019] By establishing a parameterized grid of the insulator surface and the target deposition distribution function, and merging the time series set of environmental boundary conditions, integrated spatiotemporal alignment and configuration reproducibility across devices and batches were achieved.

[0020] By driving the time-varying electric field of the electrode system and the particle charge control in parallel, the stratified deposition of soluble contaminants and insoluble particles is carried out in the powered state, realizing the directional writing of conductive connectivity paths and the programmable construction of spatial non-uniformity.

[0021] By using segmented loading and stepped flashover tests in conjunction with statistical regression methods, repeatable estimation of median flashover voltage and causal attribution of environmental segments were achieved.

[0022] By constructing a dynamic model of surface state and migration behavior within the leakage current constraint and performing closed-loop optimization, the coordinated updating and rapid convergence of the time-varying trajectory of the electrode system, the time series set of environmental boundary conditions, and the set of pollutant spectrum parameters were achieved.

[0023] By creating standardized fields for in-situ sedimentation state datasets and electrical and physical response datasets, an engineering data foundation that is fully traceable, replayable, and usable for model identification was achieved. Attached Figure Description

[0024] Figure 1 This is a schematic flowchart of the method of the present invention;

[0025] Figure 2 This is a schematic diagram illustrating the dataset relationships and closed-loop optimization in this invention. Detailed Implementation

[0026] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation.

[0027] like Figures 1-2 As shown, a method for artificial contamination testing under high pressure includes the following steps:

[0028] A parameterized mesh for the insulator surface is established, the target deposition distribution function and target response set are generated, and the environmental boundary condition time series set is merged to obtain the target deposition and response dataset;

[0029] This invention addresses the input-side unification and alignment issues in artificial pollution tests under high-voltage environments. It establishes a parameterized mesh on the insulator surface, generates a target deposition distribution function and a target response set, and merges the environmental boundary condition time series set to obtain the target deposition and response dataset. The complex three-dimensional umbrella skirt surface is "flattened" into a two-dimensional coordinate system, allowing any point to be mapped using coordinates along the creepage path. Coordinates along the direction of the busbar This unique representation ensures that field quantities such as water film thickness, salt density, and conductivity can be aligned and compared on a unified index. The method for constructing a parameterized mesh on the insulator surface includes: reading the insulator's CAD or laser scanning model, and then... (The sentence is incomplete and requires more context to translate accurately.) Sampling is performed along the radial generatrix in angular increments. Sampling, generation A regular mesh is used to record additional labels such as local curvature and skirt numbers, which can be used for visual mask and electric field sensitivity calculations. Its core lies in using the surface coordinates of the insulator surface as a unified benchmark, and using spatial non-uniformity templates and response indicators as constraints, to provide a data structure and index system that can be directly called upon for subsequent in-situ deposition, loading, and closed-loop optimization.

[0030] In principle, the insulator surface is an irregular curved surface of revolution. If the deposition target is directly described on three-dimensional coordinates, subsequent deposition and measurement are difficult to align. This invention employs a dual-parameterization method along the creepage path direction and along the umbrella skirt generatrix direction to unfold the three-dimensional surface into a regular grid. Each grid cell is associated with a location index, local curvature label, and visualization mask, ensuring that any target field can be aligned with the in-situ measured water film thickness field and surface equivalent conductivity field under the same index. The spatial non-uniformity template is used to express the intensity zoning and connectivity expectation of the target deposition. It can be assigned values ​​according to structural regions such as the inner edge, outer edge, and ribs of the umbrella skirt, and allows for strip-like or patchy distributions along the creepage path.

[0031] The target deposition distribution function is defined as the surface density of contaminants per unit area and its component ratio, including the surface density of soluble contaminants, the surface density of insoluble particles, and their volume fraction labels. It also records the stratification relationship of the target and the surface energy constraint intention, facilitating subsequent shaping using a framework-first, then coating approach. Given the desired soluble salt surface density and insoluble particle surface density for each grid cell, it can be considered as a contaminant quality blueprint to be "written"; simultaneously, the relative weights of sections such as the outer edge zone and the back of the umbrella skirt are defined. Methods for constructing the target deposition distribution function include: statistically analyzing the distribution of salt spray and sand / dust contamination on actual lines or mapping these statistics to anti-fouling standards. On the mesh, a desired areal density matrix is ​​formed; layering fields can be added to indicate the process sequence, such as "skeleton first, then coating". The target response set describes the desired electrical and surface state responses, including the median flashover voltage target, feature labels for the leakage current time series, and the spatial distribution target of the surface equivalent conductivity. The environmental boundary condition time series set describes the controllable external drivers that affect the deposition distribution and surface state during the experiment, including relative humidity time series, temperature time series, wind speed time series, and pollutant deposition flux time series, all aligned with a unified time base. The merged target deposition and response dataset is unified under a single version number, fixing field names, indices, units, and measurement methods to ensure seamless reference in subsequent steps.

[0032] The purpose of merging is to solidify the "desired sedimentary pattern" and the "external sequence that drives its formation" into a single dataset, facilitating subsequent controller access by time segment. Environmental boundary conditions (humidity, temperature, wind speed, and pollutant flux sequences) can be adjusted in real time within the physical high-pressure chamber; in numerical simulation, they are applied as boundary inputs to the CFD-electric field coupling model. Therefore, this step serves both the physical high-pressure chamber experiment and the simulation: offline preparation → obtaining a unified dataset; online execution → physical experiment replays the environmental sequence according to the dataset, and simulation loads the boundary according to the same sequence.

[0033] The implementation method first obtains the geometric model and operating level of the tested insulator. Grid row indices are set according to creepage path length, and grid column indices are set according to the direction of the shed busbars. The grid spacing is selected based on the target resolution. A spatial non-uniformity template is generated for each grid cell by labeling high-weighted bands in the vicinity of the outer edge of the shed, medium-weighted bands in the transition zone of the ribs, and low-weighted bands near the fittings. Then, a target deposition distribution function is given, specifically by assigning higher soluble pollutant target surface density and insoluble particle target surface density to the high-weighted bands, and recording the order and interface requirements in the stratification field. Next, a target response set is constructed, providing the target feature labels for the median flashover voltage target value range, leakage current time series, and the target spatial distribution map of surface equivalent conductivity. Subsequently, an environmental boundary condition time series set is prepared, including relative humidity time series, temperature time series, wind speed time series, and pollutant deposition flux time series. The timestamps are aligned with the grid indices to obtain the target deposition and response dataset. The dataset is stored with fixed field names, including the target deposition distribution function, spatial non-uniformity template, target response set and environmental boundary condition time series set, and includes units and allowed value ranges to ensure reuse across devices and batches.

[0034] This dataset solidifies geometric, target, and operational condition information under the same grid and time reference, significantly reducing alignment errors and human interpretation differences in subsequent steps. The spatial non-uniformity template ensures that the expected location and intensity of non-uniform deposition are determined before experimentation, avoiding the blind spots of the in-situ deposition stage. The target response set describes electrical and surface condition indicators side-by-side, providing clear criteria for subsequent loading and optimization. The unified sampling and alignment of the environmental boundary condition time series set allows sequential operational conditions to be directly mapped to the grid index, facilitating fragmented loading and synchronous acquisition.

[0035] For example, for a composite insulator with a primary line voltage level of 220V and a creepage distance of 5600 mm, the number of grid rows along the creepage path can be set to 560, and the number of grid columns along the shed busbar direction to 120. The spatial non-uniformity template sets a high-value zone in the outer three rows, a medium-value zone in the four rows at the ribs, and a low-value zone in the ten rows near the fittings. The target deposition distribution function sets higher target surface densities for soluble contaminants and insoluble particles in the high-value zone, and requires the formation of an insoluble particle skeleton before the formation of a soluble contaminant coating layer in the stratification field. The target response set sets a median flashover voltage target value range and provides segment energy percentage labels for the leakage current time series and a target spatial distribution map of the surface equivalent conductivity. The environmental boundary condition time series set provides relative humidity time series, temperature time series, wind speed time series, and contaminant deposition flux time series with a 1-second sampling interval. The target deposition and response datasets obtained through the above processing can be directly used as unified inputs for in-situ deposition control, loading sequence design, and closed-loop optimization, thereby improving the reproducibility and statistical comparability of the experiment.

[0036] Preferably, the environmental boundary condition time series set consists of relative humidity time series, temperature time series, wind speed time series and pollutant deposition flux time series, and is sampled and aligned with a unified time reference as input to the target deposition and response dataset.

[0037] This invention defines the environmental boundary condition time series set as a group of data including relative humidity time series, temperature time series, wind speed time series, and pollutant deposition flux time series. These are then sampled and aligned using a unified time reference and used as input to the target deposition and response dataset. The reason for using grouped time series instead of single-point values ​​is that the hygroscopic absorption, dissolution, recrystallization, and membrane connectivity of pollutants all exhibit strong time and path dependencies. Furthermore, the leakage current and flashover threshold during high-pressure loading are highly sensitive to wet-dry cycles, rise / fall rates, and flux pulses. Unifying the four types of external drivers onto the same time axis provides a unique environmental index for each response record during subsequent deposition control, loading segment design, and parameter identification, avoiding alignment errors and interpretation ambiguities.

[0038] In terms of implementation, a common clock is used as the time reference for the entire link. Each sensing and execution node timestamps and caches the data locally under the same time reference, and then the data aggregation node aligns them. Relative humidity and temperature sensors are deployed at representative heights near the insulator surface. Wind speed sensors are deployed near the middle of the insulator in the direction of the incoming flow. Pollutant deposition flux is obtained through mass accumulation sensing or weighing of a reference sampling plate; flux is defined as the mass increment per unit area per unit time reaching the reference surface. To reduce boundary layer effects and installation disturbances, thin rods and remote leads are used for sensor installation. Zero-point and range checks are performed before and after measurement. When the sampling interval and control step size are consistent, direct alignment is performed; when inconsistent, resampling and interpolation are performed based on the common time axis. Quality labels are attached to interpolated segments for easy subsequent analysis and screening. For abrupt changes, a median strategy using forward and backward windows is used for consistency checks. Missing measurements are filled using linear or spline methods with the shortest acceptable window, and the filling mark and original missing measurement length are clearly marked.

[0039] At the data structure level, the environmental boundary condition time series set is fixed with unified field names, unified units, and unified coordinate numbers, and the time key is unique and monotonically increasing. After establishing a time-to-space mapping relationship with the spatial index of the parameterized grid on the insulator surface, the environmental drive at any time can generate the corresponding deposition and wetting boundaries under the grid coordinates. When loading the target deposition and response dataset into this set, the time key remains unchanged, and relative humidity, temperature, wind speed, and pollutant deposition flux are used as exogenous variables in the subsequent control and identification process, so that the time history of deposition control, the segment boundaries of the loaded sequence, and the regression window of parameter identification can strictly correspond on the same time axis.

[0040] Example 1 constructs an environmental program including relative humidity, temperature, wind speed, and pollutant deposition flux. Relative humidity initially increases to a high-humidity plateau at fixed steps per minute, remains constant in the middle stage, and then decreases to a dry level at fixed steps in the final stage. Temperature is maintained at a constant value, wind speed switches between two levels using pulses, and pollutant deposition flux is pulsed once before and after the wet impact. All channels are sampled and aligned using the same second-level time base, with missing data segments lasting no more than ten seconds and filled using a linear method, while filling markers are recorded. This set serves as input to the target deposition and response dataset. Deposition is controlled at the high-humidity plateau triggering the coating stage of soluble pollutants. Voltage transient segments are added after the flux pulses in the loading sequence. Parameter identification uses windows corresponding to the plateau segments to extract the rates of change of water film and conductivity. Finally, model fitting and optimization are completed on the same time key.

[0041] This invention does not elaborate on mature measurement methods, but emphasizes the need to ensure the consistency of time keys and the stability of field names in applications. It also stresses that traceability records in multi-channel scenarios should be stored along with the data, including sampling intervals, calibration times, missing measurement flags, and interpolation methods. Through these principles and implementation methods, the environmental boundary condition time series set not only serves as input but also as an alignment framework for subsequent control, loading, and identification, supporting the reproduction experiments and closed-loop convergence of the final deposition morphology and electrical response.

[0042] In the energized state, a time-varying electric field distribution is formed through the electrode system, soluble pollutant aerosols and insoluble particulate aerosols are supplied to the surface of the insulator and particulate charge control is implemented. Based on the time series set of environmental boundary conditions and the supply switching, a layered pollutant deposition structure is constructed, surface state field data is collected, and an in-situ deposition state dataset is generated.

[0043] In the powered state, the present invention forms a time-varying electric field distribution through an electrode system, which, in conjunction with the continuous or segmented supply of soluble pollutant aerosols and insoluble particulate aerosols, and under the combined effect of particle charge control and environmental boundary condition time series set, constructs a layered pollutant deposition structure with a skeleton and coating relationship on the surface of the insulator. At the same time, the spatially resolved surface state field data is used as a record to generate an in-situ deposition state dataset.

[0044] The electrode system consists of multi-channel electrodes, with independently set channel phases and amplitudes. Arranged around the tested insulator, they are sequentially driven to form a high-field distribution on the surface that moves along the creepage path and the direction of the skirt busbars. The time-varying electric field serves two purposes: first, shaping the migration and trapping trajectories of charged particles near the wall surface using the normal electric field component; second, adjusting the transient connectivity path of the thin water film using the tangential electric field component, ensuring the deposition location aligns with the desired subsequent conductive connectivity. Particle charging control employs a unipolar charging method, with the charge-to-mass ratio as the core indicator, maintained stable through online metering and closed-loop correction. Soluble pollutant aerosols, acting as precursors to soluble salts, determine the surface equivalent conductivity and redissolution behavior during wet-dry cycles. Insoluble particulate aerosols, acting as a rough framework, provide nucleation sites and persistent morphology.

[0045] The environmental boundary condition time series dataset includes relative humidity, temperature, wind speed, and pollutant deposition flux time series. Relative humidity and temperature determine the formation and evaporation of the thin water film, wind speed determines near-wall shear and the residence probability of incident particles, and pollutant deposition flux provides the mass input per unit area per unit time. Through coupling with supply switching, a stratification rhythm is achieved by first depositing insoluble particles to form a framework, and then supplying soluble pollutants to complete the coating during the high relative humidity stage. To avoid ambiguity, this invention clarifies the following terms: water film thickness field is the spatial distribution of the thin water film thickness of each grid cell on the surface; surface equivalent conductivity field is the spatial distribution of the equivalent conductivity per unit length of each grid cell on the surface; soluble pollutant surface concentration field is the spatial distribution of the surface density of soluble salts in each grid cell on the surface; insoluble particle framework density field is the spatial distribution of the surface density of insoluble particles in each grid cell on the surface. The above four fields, together with experimental metadata, constitute the in-situ deposition state dataset.

[0046] To highlight the core mechanism, this invention provides only three minimized evolutionary relationships in this step to guide the alignment and verification of control and observation variables: , , ,in, Indicates the thickness of the water film. This indicates the surface density of soluble pollutants. This indicates the surface density of insoluble particles. Represents a time series of relative humidity. Indicates the normal electric field strength of the surface. Indicates the near-wall wind speed. This represents the time series of deposition fluxes of soluble pollutants. This represents the time series of insoluble particle deposition flux. These are non-negative coefficients related to materials and geometry. Let be a dimensionless function of the thin film supply term under the combined influence of electric field and wind speed. This is a dimensionless function representing the coupling term of dissolution and migration under the influence of the electric field. The specific values ​​of the coefficients and the dimensionless function will be determined in the subsequent identification steps. This invention does not elaborate on the derivation here, but only serves to clarify the correspondence between control variables and state variables, ensuring the consistency between data and process.

[0047] In terms of acquisition strategy, the water film thickness field is measured non-contactly using the oblique incidence reflection method or structured light method; the surface equivalent conductivity field is scanned synchronously using a segmented electrode array; the soluble pollutant surface concentration field is obtained through micro-elution and conductivity measurement or spectral inversion; and the insoluble particle skeleton density field is verified by surface scattering characteristics and benchmark sampling. Temporally, the acquisition frequency and environmental boundary condition time series are kept consistent, and segment time stamps are used to record supply switching and electrode system waveform switching to ensure consistency in subsequent statistics and playback.

[0048] In Example 2, the environmental program was set to a humidified platform followed by drying. First, an insoluble particle deposition flux was used to form the framework under a high relative humidity platform in an powered-on state. Then, the flux was switched to soluble contaminants to complete the coating. The electrode system's channel phase was progressively scanned, causing the high-field region trajectory to move along the outer edge of the umbrella skirt. At this time, the water film thickness field increased with increasing relative humidity, and the surface equivalent conductivity field formed a continuous band at the outer edge. The soluble contaminant surface concentration field and the insoluble particle framework density field overlapped on the same band. The in-situ deposition state dataset was verified and used for the loading design to ensure that the subsequent electrophysiological response connectivity path was consistent with the deposition path.

[0049] Preferably, the electrode system is a multi-electrode shaping array, with the phase and amplitude of the electrode channels set independently, and the electrode channels arranged with a fixed geometric spacing, forming a time-varying electric field distribution that moves along the axial and radial directions of the insulator through sequential driving.

[0050] This invention employs a multi-electrode shaping array as the electrode system, with channel phase and channel amplitude set independently. The channels are arranged around the insulator at fixed geometric intervals, and a continuously moving, time-varying electric field distribution is formed in the axial and radial directions of the insulator through sequential driving. The normal component of this electric field distribution is used to control the capture of charged particles in the near-wall region, while the tangential component is used to guide the transient migration and connectivity of the thin water film. This enables the controllable shaping of the layered pollutant deposition structure under energized conditions, and generates an in-situ deposition state dataset using spatially resolved surface state field data.

[0051] The multi-electrode shaping array can be viewed as performing spatiotemporal shaping of the potential in the test space. The influence function of each electrode channel is pre-obtained in the offline solver, and then superimposed in real-time according to the channel phase and channel amplitude during the experiment. To clarify the relationship between the control and state variables, this invention provides a minimized field-fluid coupling expression in this core component: , , , , , ,in, The electric field intensity vector, For spatial location, For time, This refers to the number of electrode channels. For the first The time-varying drive potential of the channel, For the first Channel influence function, The surface normal unit vector, It is the unit vector of the surface tangential direction. and These are the normal and tangential field components, respectively. For insoluble particle deposition flux, For soluble pollutant deposition flux, and For the collection efficiency coefficient, The particle charge-to-mass ratio, and This represents the time series of volume fractions of insoluble particles and soluble pollutants in the air. For surface equivalent conductivity, For the dry surface baseline conductivity term, The coupling coefficient is... For water film thickness, This represents the surface density of soluble contaminants. The above expression guides the sequential drive strategy and supply switching strategy of the electrode system, and verifies the physical consistency of the state field during the data acquisition phase.

[0052] In this implementation, the multi-electrode shaping array is arranged in a ring or spiral pattern, with the channel spacing and the ratio of the insulator's outer diameter remaining fixed to ensure the repeatability of the field shape. The channel phase uses traveling wave coding to generate high-field region scanning along the axial direction, and the channel amplitude uses envelope modulation to focus the high-field region radially. The time step of the sequence-driven operation is consistent with the environmental boundary condition time series set, and the driving state of the electrode system is written to the in-situ deposition state dataset using the same time reference. Particle charging control uses unipolar charging and is verified online using a Faraday cylinder and electrostatic metering. The verification results are used to correct the supply mass flow rate and charge-to-mass ratio, ensuring the coordination of flux and electric field. During the acquisition phase, the surface state field is sampled at high frequency at the instant of electrode system switching. The water film thickness field is obtained using oblique incidence reflection or structured light inversion, the surface equivalent conductivity field is obtained using a segmented electrode array scan, the soluble pollutant surface concentration field is obtained using micro-elution conductivity or spectroscopy, and the insoluble particle skeleton density field is obtained through surface scattering inversion and verified using reference sampling. All fields, along with the electrode system and supply status, are written with the same time key, thus forming a replayable in-situ deposition status dataset.

[0053] Example 3 uses 24 channels and employs traveling wave coding with uniform phase advancement to continuously move the high-field region along the creepage path. The supply strategy is to first supply insoluble particles and then soluble contaminants. The environmental boundary condition time series dataset uses a high relative humidity platform with drying. This method forms a continuous skeleton band at the outer edge and a uniform coating layer along the same path. The surface equivalent conductivity field is band-connected. The water film thickness field and the soluble contaminant surface concentration field in the in-situ deposition state dataset highly overlap on the outer edge band. During subsequent loading, the main channel of the leakage current is consistent with the deposition path, verifying the controllability of the electrode system over the connected path.

[0054] Preferably, particle charge control employs corona charging, with the charge polarity being either unipolar positive or unipolar negative. The particle charge is measured using a Faraday cylinder in conjunction with an electrostatic meter, and the measured value is used to correct the aerosol supply parameters.

[0055] In this invention, particle charge control is used as an adjustable pre-processing step to determine deposition distribution and connectivity in artificial contamination experiments under high-pressure environments. A corona charging device applies a unipolar positive or negative charge to soluble pollutant aerosols and insoluble particulate aerosols, achieving a stable charge-to-mass ratio. The particles then enter a deposition zone containing a time-varying electric field, undergoing directional migration and near-wall trapping along the electric field direction. Online calibration of the charge-to-mass ratio is performed using a Faraday cylinder and an electrostatic meter. The measured values ​​are used to correct the aerosol supply mass flow rate and corona voltage in a closed-loop manner, thereby achieving the target deposition flux under given normal electric field and air volume fraction conditions.

[0056] To establish a calculable mapping relationship between control quantities and observable quantities, this invention provides two types of minimum necessary expressions in this core step. The first is the determination and correction of the charge-to-mass ratio: ,in This indicates the charge-to-mass ratio, expressed in coulombs per kilogram. This indicates the current measured by the Faraday cylinder, and the unit is ampere; This indicates the mass flow rate of aerosols entering the Faraday cylinder, expressed in kilograms per second. The Faraday cage collection efficiency is represented by a dimensionless coefficient, the value of which is obtained from geometric and flow regime calibrations. This formula allows for real-time calculation of... The feedback input to the joint controller for supply and charging ensures that the set target charge-to-mass ratio remains stable even in the presence of disturbances. Secondly, there is the estimation of deposition flux driven by near-wall electromigration. ,in This indicates the deposition flux pointing towards the surface, expressed in kilograms per square meter per second. The collection efficiency is expressed as a dimensionless coefficient. This represents the normal electric field strength on the surface, measured in volts per meter. This represents the volume fraction of the corresponding component in the air, in dimensionless units. From this, the target charge-to-mass ratio for the controller inverse solution can be obtained: ,in This indicates the target deposition flux, expressed in kilograms per square meter per second. This indicates the target charge-to-mass ratio.

[0057] In this implementation, the corona charging device is positioned before the supply pipeline enters the climate chamber. The electrodes employ a needle-plate or needle-ring geometry to prevent self-excited discharge from crossing the insulator shell. The polarity of the charge electrode is selected before the test begins and remains constant throughout the test to ensure a predictable electromigration direction after superposition with the time-varying electric field components. The supply pipeline uses a conductive flexible tube with at least two grounding drain points and a length not exceeding 2 meters to reduce leakage charge and backflow pulsation. The Faraday cylinder is positioned in the sampling branch, and the sampling flow rate is stably maintained by a mass flow controller. Each calibration cycle for the sampling branch does not exceed 10 minutes. During calibration, the electrostatic meter performs dual zero-point and range checks and records the data in the traceability log. The charge-to-mass ratio controller uses a dual-channel adjustment mechanism: channel one adjusts the corona voltage, and channel two adjusts the spray mass flow rate. Both channels sample and output using the same time reference, and the control law employs a proportional-integral correction. ,in Vector components representing control commands, including corona voltage and mass flow rate. and Representing proportional and integral gain The sampling period is indicated. To avoid outliers caused by multiple charges of particles, this invention sets up an electrostatic rectification section after the outlet of the charging chamber. Charge fluctuations are reduced through weak field averaging and linear aerodynamic rectification, and the sliding median of the Faraday cylinder current is used as the feedback quantity.

[0058] This step works in conjunction with the electric field shaping and environmental boundary condition time series dataset of this invention. When the normal electric field strength increases, the controller automatically reduces the charge-to-mass ratio or spray mass flow rate to maintain the target deposition flux. When the volume fraction in the air increases, the controller reduces the corona voltage to avoid space charge shielding and nucleation anomalies caused by overcharging. Since the Faraday cylinder and the electrostatic meter are based on the same time reference, their measured values, along with the supply instructions, are written into the control channel field of the in-situ deposition state dataset for subsequent loading segments and statistical identification.

[0059] Example 4: Unipolar positive charging is used to form an outer edge connected framework, and the target deposition flux is set to be... kilograms per square meter per second, surface normal electric field strength is The air volume fraction is volts per meter. The collection efficiency is The target charge-to-mass ratio is obtained from the inverse equation. Millicoulombs per kilogram, the controller updates the corona voltage and mass flow rate at 1-second intervals, and the Faraday cylinder current remains stable at... Amperes, corresponding to mass flow rates Kilograms per second, in-situ deposition state datasets show that the surface equivalent electrical conductivity field of the outer edge zone continuously increases in line with the target non-uniformity template.

[0060] Preferably, the construction sequence of the layered pollutant deposition structure is as follows: first, insoluble particulate aerosols are supplied to form a particulate skeleton; then, in the wetting stage, the relative humidity is increased and soluble pollutant aerosols are supplied to form a coating layer; and finally, in the drying stage, the relative humidity is reduced to stabilize the layered structure.

[0061] This invention employs a sequential deposition process to insoluble particles and soluble contaminants onto the surface of an insulator under an electrically charged state. First, an insoluble particle framework is constructed, providing roughness and nucleation sites. Then, during a wetting phase, relative humidity is increased and soluble contaminants are supplied, causing them to dissolve and redistribute within a thin water film, and then, guided by a tangential electric field, cover the framework surface. Finally, during a drying phase, relative humidity is reduced to complete recrystallization and pore consolidation, thereby obtaining a layered contaminant deposition structure with controllable connectivity and a stable morphology. This sequence coordinates the three processes—electric field-induced near-wall particle trapping, water film-mediated salt migration, and drying / solidification—within the same time frame, ensuring that subsequent loading and identification can reuse the same spatial and temporal index.

[0062] The core of layered construction can be characterized by three minimal expressions. First, the effective occupancy of the framework increases with the surface density of insoluble particles, which can be approximated as: ,in This indicates that the effective occupancy is a dimensionless quantity. This indicates the surface density of insoluble particles. This represents the convergence coefficient related to particle morphology and surface energy. This occupancy, along with the spatial non-uniformity template, determines the initial region of framework connectivity. Secondly, the temporal history of the coating thickness during the wetting stage is jointly dominated by deposition and dissolution redistribution, and can be written as: ,in Indicates the local equivalent thickness of soluble pollutants. The depositional conversion coefficient, determined by geometry and porosity. This represents the time series of deposition fluxes of soluble pollutants. This represents the volume fraction loss coefficient due to drying. This represents the relative humidity time series. Thirdly, the surface equivalent conductivity is related to the water film, salt load, and connectivity threshold, and can be written as: ,in Indicates the surface equivalent conductivity. Indicates the baseline conductivity of the dry surface. This represents the coupling coefficient between the water film and the salt load. Indicates the thickness of the water film. Indicates the surface density of soluble contaminants on the surface. This represents a step indicator function; it takes the value 1 when the quantity inside the parentheses is positive and 0 otherwise. This represents the critical occupancy threshold for connectivity. Together, these three equations link framework connectivity, coating growth, and electrical conductivity response within the same spatial grid and time key, facilitating cross-verification between measured and control quantities during deposition.

[0063] The key implementation points are threefold. First, in the framework stage, only insoluble particles are supplied, and the normal electric field component is maintained as the dominant component. The tangential component is used to move high-field regions along the creepage path, allowing the framework to connect in a strip or patchy pattern according to the spatial non-uniformity template. Second, in the wetting stage, the relative humidity is increased to a range that can form a stable thin water film without causing runoff. At the same time, the supply of soluble pollutants is switched, and the proportion of the tangential component is increased, so that salt forms a continuous coating on the near-surface of the framework rather than filling cavities. Third, in the drying stage, the relative humidity is reduced and the tangential component is decreased to inhibit water film migration. This allows the formed coating layer to recrystallize in the pores and mechanically interlock with the framework, thereby improving shape retention without changing the framework topology.

[0064] In data acquisition, all three stages recorded the water film thickness field, surface equivalent conductivity field, soluble pollutant surface concentration field, and insoluble particle skeleton density field using the same time reference. The phase and amplitude trajectories of the electrode system, as well as the supply switching timestamps, were also recorded simultaneously. The focus of the skeleton stage was the low-frequency baseline of the insoluble particle skeleton density field and the surface equivalent conductivity field; the focus of the wetting stage was the coupling increment of the water film thickness field and the soluble pollutant surface concentration field; and the focus of the drying stage was the spatial stability of the surface equivalent conductivity field after the coating layer crystallized.

[0065] Example 5 uses a sea salt scenario. During the framework stage, insoluble particles are supplied to the high-field zone at the outer edge, while the water film thickness field remains low, forming a banded connectivity. During the wetting stage, relative humidity is maintained at a plateau, and the supply of soluble contaminants is switched. The tangential component guides the coating to close along the outer edge, and the surface equivalent conductivity field increases on the same band. During the drying stage, after reducing relative humidity, the surface equivalent conductivity field remains stable in a banded pattern. In-situ deposition state datasets show that the coating layer overlaps well with the framework, satisfying the target depositional distribution function and spatial non-uniformity template.

[0066] Preferably, the surface state field data includes a water film thickness field, a surface equivalent conductivity field, and a soluble pollutant surface concentration field. The water film thickness field is obtained by non-contact optical measurement and conversion, the surface equivalent conductivity field is obtained by segmented measurement using a multi-point electrode array and inversion, and the soluble pollutant surface concentration field is obtained by either micro-elution conductivity measurement or spectral measurement.

[0067] This invention performs spatiotemporal analysis of the insulator surface state under energized and segmented loading conditions. The surface state field data consists of the water film thickness field, the surface equivalent conductivity field, and the soluble contaminant surface concentration field. After being aligned with a parameterized grid and a unified time reference, the data is written into the in-situ deposition state dataset and the electrical and physical response dataset. All three types of field quantities are obtained using non-contact or weakly perturbed methods to avoid disturbing the high-voltage boundary layer, the thin water film, and the conductive connectivity path.

[0068] The water film thickness field is obtained based on non-contact optical measurement and conversion. In the application of this invention, oblique incidence reflection interferometry or structured light phase method is selected. By calibrating the reflection phase difference or fringe phase at the same grid position, the local film thickness is calculated. The core relationship can be written as follows: ,in Indicates the thickness of the water film. Indicates the center wavelength of the imaging light. Indicates the refractive index of the water film. Indicates the angle of incidence. This represents the phase difference between the reference and the measured reflected signal. The above formula is only used to map the phase measurement to the minimum necessary relationship for thickness; the actual phase acquisition and unwrapping are completed by the imaging system. To adapt to the curved surface of the insulator and the shading of the skirt, this invention pre-generates a visible mask on a parameterized grid and performs multi-view stitching; for the time-varying wetting and drying processes, sampling is performed with a unified time key and the fringe phase is unfolded to ensure the continuity of the thickness field at the segment boundaries. In effect, the water film thickness field provides the spatial distribution of the measured wetting intensity, providing a quantitative basis for interpreting the changes in leakage current segment energy and median flashover voltage.

[0069] The surface equivalent conductivity field is obtained through segmented measurement and inversion using a multi-point electrode array. The electrode array is arranged segmentally along the creepage path, with each segment exciting only a small number of electrodes at any given time to reduce electromagnetic coupling and bulk heating. Measurement modeling employs a linearized transfer relation, combining the segmented measured port current and port voltage into an observation vector, and establishing a matrix mapping between this vector and the equivalent conductivity discrete on a parameterized grid. Inversion uses regularized least squares. ,in This represents the estimated equivalent conductance grid vector. This represents the spliced ​​current and voltage observation vectors. This represents the sensitivity matrix determined by geometry and electrode arrangement. Represents the discrete smoothing operator. This represents the regularization weight. To avoid interference from power-on measurements on the field shape, this invention interleaves the electrode array drive and the main voltage program in time, and uses segment time markers to align the start and end of each scan with the environmental boundary condition time series set. In effect, the inverted surface equivalent conductivity field exhibits significant changes at the formation and breakage points of the conduction path, consistent with the connectivity events after thresholding the conduction map, and can be directly used for the calculation of connectivity difference indices and closed-loop optimization objectives.

[0070] The surface concentration field of soluble contaminants is obtained through either micro-elution conductivity measurement or spectroscopic measurement. Micro-elution employs a short-term contact between a fixed-area microcavity or a hydrophilic membrane on a grid cell to obtain the eluent conductivity, which is then converted into surface density using a calibration curve. The conversion relationship is denoted as follows: ,in This indicates the surface density of soluble pollutants. The eluent conductivity is represented; spectroscopic methods convert the areal density at known salt absorption peaks using ratios or differential absorbance, resolving quantification issues when multiple components overlap. Both methods archive a one-to-one mapping of "sampling window—in-situ location—parameterized grid cell"; the time interval and sampling volume for repeated sampling at the same location are fixed in the metadata, ensuring comparability across batches. In terms of effectiveness, the concentration field reveals differences in salt loading under the same water film thickness field, providing a crucial supplementary measure for explaining the anisotropy of the equivalent conductivity field and differences in electrical response.

[0071] The synchronization and alignment of the three types of field quantities is a key operation of this invention. All measurements are sampled with a unified time reference and synchronized with the segment time markers; all spatial data are recorded on a parameterized grid with the same coordinates, accompanied by a visual mask and quality label. Missing measurement points are handled using interpolation or resampling strategies and written into the quality label. Thus, the water film thickness field, the surface equivalent conductivity field, and the soluble pollutant surface concentration field can be constructed with increments within the same segment window for coupling regression of electrical and surface states, calculation of connectivity difference indices, and identification of dynamic models.

[0072] Example 6: Three types of field quantities were obtained in the wetted platform segment. The water film thickness field showed an increase in thickness in the outer edge region, and the soluble pollutant surface concentration field increased in the same region. The inverted surface equivalent conductivity field formed a continuous high-value band, which significantly corresponded to the leakage current segment energy of the subsequent voltage step segment.

[0073] Loading is performed based on the time series set of environmental boundary conditions and voltage condition changes. The time series of electrical response and surface state field are collected. The median flashover voltage is obtained by flashover test and statistical methods. The relevant increments are calculated to generate an electro-physical response dataset.

[0074] This invention employs a dual-drive approach, using a time series of environmental boundary conditions and voltage variations as a unified time base, to load insulators that have undergone layered deposition. This yields electrical response time series and surface state field time series. The median flashover voltage is then calculated using flashover testing and statistical methods. Correlation increments within the synchronization window are calculated to generate an electrophysical response dataset. This dataset is used in subsequent closed-loop optimization to measure the deviations between connectivity thresholds, migration rates, and the target deposition distribution.

[0075] In principle, the environmental boundary condition time series defines the time-varying trajectories of relative humidity, temperature, wind speed, and pollutant deposition flux, while voltage condition variations define the amplitude program and transient segments of the applied voltage. Aligning both on the same time axis allows for a one-to-one correspondence between the external excitation and response of each segment. The electrical response time series includes at least the leakage current time series and its synchronous flashover event markers; the surface state field time series includes at least the water film thickness field time series and the surface equivalent conductivity field time series. To avoid conceptual ambiguity, this paper refers to a segment as a loading window marked with start and end times on the time axis, and an increment as the difference between the end and start times of the same grid or channel within a segment.

[0076] In estimating the median flashover voltage, this invention employs a stepped flashover test and statistical regression. Several test points are established using voltage steps, and the occurrence of flashover is recorded at each test point. The voltage is denoted as... The flashover probability is A probabilistic model is established using logistic regression: The median flashover voltage is defined as the voltage that makes the probability equal to one-half. ,in and For regression coefficients. The median flashover voltage reflects the tolerance threshold under a given deposition configuration and environmental program, and is written as a core scalar in the electrophysiological response dataset. The strength and connectivity of the electrical response are described by quantitative indicators such as fragment energy and peak value. The leakage current fragment energy is defined as: ,in This is a time series of leakage current. and These represent the start and end times of the segment. The increment of the surface state field is calculated in units of grids, for example, the increment of the surface equivalent conductivity field: Water film thickness field increment: ,in Spatial grid coordinates, and The spatiotemporal distributions of surface equivalent conductivity and water film thickness are shown, respectively. All increments are written with the same time key and spatial index to ensure repeatability of cross-segment comparison and regression identification.

[0077] The key implementation points are as follows: First, the loading program follows the order of "environment first, voltage second," first setting the surface wetting state with a relative humidity step or plateau segment, and then applying voltage steps or voltage transients within the window. Second, the acquisition system timestamps data using a time base consistent with the environmental boundary condition time series set. The leakage current time series uses a high sampling rate channel, and the surface state field is acquired using a snapshot method synchronized with the voltage step. If necessary, sampling is encrypted in key segments. Third, the start and end times of all segments, environmental measurement values, voltage amplitude and polarity, flashover status, and sensor quality indicators are all written into the dataset metadata to form a replayable test record.

[0078] Example 7: Voltage Steps on a Humid Platform. Relative humidity was increased to a stable level before loading and maintained, while temperature and wind speed remained constant. The voltage was increased in fixed steps until the first flashover occurred, and this process was repeated multiple times in the same steps to obtain samples of flashover and non-flashover events. The median flashover voltage was obtained using logistic regression. The fragment energy index showed a significant correlation between the increase in the surface equivalent conductivity field in the outer strip region and the fragment energy of the leakage current, verifying that the connectivity path written during the deposition stage was reproduced by the electrical response during the loading stage.

[0079] Preferably, the loading is performed in a segmented manner. The loading sequence includes at least three types of segments: relative humidity step, temperature step, wind speed change, pollutant deposition flux pulse, and voltage transient. The start and end times of each segment are recorded with time stamps and synchronized with the timestamp of the acquisition channel.

[0080] This invention employs a segmented loading method, arranging the time series of environmental boundary conditions and voltage condition changes into a set of atomized segments, and recording them synchronously with the timestamps of the acquisition channels under the same time base. The purpose of segmented loading is to decompose the effects of wetting and drying, convection and diffusion, flux pulses and electromigration, and transient electrical stress on connected paths into identifiable causal units, thereby simultaneously supporting the estimation and parameter identification of median flashover voltage at both the statistical and dynamic levels.

[0081] In the context of this invention, a segment refers to a control window with a defined start and end time on a unified time axis. Each segment modifies only a few exogenous quantities to maintain identifiability. Segment types include relative humidity steps, temperature steps, wind speed changes, pollutant deposition flux pulses, and voltage transients. Each segment is time-stamped and written into metadata. The time stamp and the timestamp of the acquisition channel use the same common clock. Time drift is corrected at the acquisition end using time synchronization and then corrected again at the aggregation end using an alignment algorithm. This ensures that the electrical response time series and the surface state field time series have a repeatable registration relationship at the segment boundaries.

[0082] To facilitate control and analysis, a fragment pointer function is used to formally describe the loader:

[0083]

[0084] in For the first Indicator functions for each segment, and These represent the start and end times of the segment. Environmental channels are given in the form of segment overlays, such as a relative humidity time series: Wind speed time series: Time series of pollutant deposition flux: Voltage program: ,in This is a time series of relative humidity. This is a time series of wind speeds. This is a time series of pollutant deposition flux. This is a time series of voltage operating conditions, with subscripts. The amount is the baseline value, with The amount is the increment applied within the fragment. This is a set of relative humidity step segment indices. This is a set of indices for wind speed change segments. This is a set of indexes for contaminant deposition flux pulse segments. This is the set of voltage transient segment indices. The above expression is not for derivation, but rather to clarify the mapping between the control quantity and the segment boundaries, ensuring consistency between the window used in subsequent response calculations and statistical estimations.

[0085] The leakage current time series and surface state field time series are acquired on the response side using the same time reference. The electrical response intensity within a segment is described by segment energy and peak value. The leakage current segment energy is defined as: ,in This is the time series of leakage current. The surface state field is calculated in increments on a grid basis, and the surface equivalent conductivity field increment is: Water film thickness field increment: ,in Spatial grid coordinates, The spatiotemporal distribution of the surface equivalent conductivity. This represents the spatiotemporal distribution of the water film thickness. For each segment, a structure is constructed using... and and The resulting response vector, along with the control increments within the segment and and and They are jointly written into the data structure of electrical and physical responses to form a fragment-level record of the electrical and physical responses.

[0086] The median flashover voltage was estimated using a stepped flashover test combined with statistical methods. Several segments in the voltage program were set as voltage step segments, and the occurrence of flashover in each step segment was recorded. Logistic regression was used to model the flashover probability and solve for the median flashover voltage, which characterizes the tolerance threshold under the current deposition configuration and environmental program. Since all segments were recorded with a uniform time stamp, the estimated median flashover voltage can be correlated with specific environmental combinations, facilitating the separation of the contributions of humidity and wind speed to the threshold.

[0087] The key implementation points include three aspects. First, the orchestration strategy follows the principle of minimizing variable changes, modifying only a few exogenous quantities in each segment while maintaining the baseline for the rest to avoid confusion. Second, the synchronization strategy uses a common clock at both the control and acquisition ends, with time stamps falling on the sampling grid. If cross-grid errors exist, they are aligned by interpolation between adjacent points during data aggregation, and the alignment method is preserved in the metadata. Third, the quality control strategy writes sensor status and missing measurement flags into each segment, and adds quality labels to resampling and interpolation segments to ensure consistency in subsequent statistical screening.

[0088] Example 8 selected three segments: relative humidity step, wind speed change, and voltage transient. The procedure was as follows: first, the relative humidity increased to a plateau; then, a wind speed pulse was applied to the plateau; and finally, a short-term voltage transient was applied within the window. The start and end times of each segment were recorded with second-level time stamps. The leakage current time series and the surface state field time series were synchronized within the three segment windows. The results showed that the water film thickness field in the wind speed pulse segment exhibited a negative increment, the leakage current segment energy in the voltage transient segment was significantly increased, and the median flashover voltage increased compared to the baseline loading without a wind speed pulse.

[0089] Within the leakage current constraint, based on the electrophysiological response dataset and the target deposition and response dataset, a dynamic model of surface state and migration behavior is established with the deviation indices of deposition configuration and conductive connectivity, and electrical response and median flashover voltage as optimization objectives. The time series set of time-varying trajectory of electrode system and environmental boundary conditions is updated, the pollutant spectrum parameter set is adjusted, and the convergence parameter set is output.

[0090] This invention constructs a closed-loop optimization problem within the leakage current constraint using an electrophysiological response dataset and a target deposition and response dataset. It jointly corrects the time-varying trajectory of the electrode system, the time series set of environmental boundary conditions, and the set of contaminant spectral parameters until a convergent parameter set is obtained. The core idea is to use a dynamic model of surface state and migration behavior as a forward predictor, with deposition configuration distance, conductivity connectivity deviation, electrical response deviation, and median flashover voltage deviation as objective terms, to solve a minimization problem while ensuring a safe leakage current boundary.

[0091] The kinetic model uses the water film thickness field, the surface density field of soluble pollutants, and the surface density field of insoluble particles as state variables. It employs the time series set of environmental boundary conditions, the time-varying trajectory of the electrode system, and the pollutant deposition flux as exogenous drivers, and the surface equivalent conductivity field as a derived variable. To ensure sign consistency, the following minimal expressions for the state and derived variables are defined for coupling control variables and observables: , , , ,in, For water film thickness, The surface density of soluble pollutants. The surface density of insoluble particles. This is a time series of relative humidity. The surface normal electric field intensity, Near-wall wind speed, and The time series of deposition fluxes for soluble pollutants and insoluble particles are shown separately. For surface equivalent conductivity, For the baseline conductivity of the dry surface, For the effective occupancy of the skeleton, The connectivity threshold, For indicator functions, and It is a dimensionless coupling function. The coefficients are non-negative. The values ​​of the coefficients and the function are identified from the electrophysical response dataset within the calibration window. This set of equations is not used to derive theoretical upper limits, but rather to generate predictors that are aligned term-by-term with the measured quantities in the optimization loop.

[0092] The optimization objective is a weighted sum of multiple indicators, written as: ,in, This represents the set of control variables that need to be updated, including parameters of the time-varying trajectory of the electrode system, incremental parameters of the environmental boundary condition time series set, and the set of pollutant spectrum parameters. The depositional configuration distance is used to measure the spatial difference between the predicted surface density field of soluble contaminants and the target depositional distribution function. This is a conductivity connectivity deviation index used to measure the difference between the thresholded surface equivalent conductivity field connectivity structure and the target connectivity expectation. To account for the time series differences in leakage current, the fragment energy and peak value can be quantized and aligned with the target response set. The predicted median flashover voltage, The target median flashover voltage, The weights are non-negative. The constraints are written as follows: , ,in This is a time series of leakage current. This is the upper limit of leakage current. This represents the upper limit of the leakage current change rate. The constraint is set within the boundaries of test safety and insulator thermal load, serving as the feasible region for the optimization process.

[0093] The solution process is an iterative discrete-time domain solution. First, fragment-level data from the electrophysiological response dataset and the target deposition and response dataset are read, and the state field and weights are initialized on a unified time reference and grid. Second, forward prediction is performed using the kinetic model and the current control variable set to generate the water film thickness field, the surface density field of soluble contaminants, the surface density field of insoluble particles, and the surface equivalent conductivity field. The median flashover voltage is then predicted using a voltage program. Third, the deposition configuration distance, conductivity connectivity deviation, leakage current time series difference, and median flashover voltage deviation are calculated to form a cost function and its gradient information. The gradient can be obtained through the adjoint method or through perturbation estimation. Fourth, the control variable set is updated according to the feasible region projection method. The update of the time-varying trajectory of the electrode system uses the parameterized trajectory of phase and amplitude as variables; the update of the environmental boundary condition time series set uses the fragment amplitude and fragment duration as variables; and the update of the contaminant spectrum parameter set uses particle size distribution, charge-to-mass ratio, soluble contaminant volume fraction, and insoluble particle volume fraction as variables. Fifth, perform segment-by-segment checks on the constraints. If any segment violates the upper limit of leakage current or the upper limit of change rate, then scale and update the step size within that segment and write back the control quantity set. Sixth, determine convergence. If the cost function decreases insufficiently or the change in control quantity is less than the threshold, then output the convergence parameter set.

[0094] The key to implementation lies in the coupled updating and data alignment of the three types of controls. The time-varying trajectory of the electrode system directly affects the ratio of the normal electric field to the tangential electric field on the surface, thus influencing water film replenishment and particle capture; the time series set of environmental boundary conditions determines the rhythm of wetting, drying, and convection; and the set of pollutant spectrum parameters controls the injection intensity and charging capacity of soluble pollutants and insoluble particles. Coordinated adjustment of these three factors under a unified time key avoids local optima caused by relying solely on a single path for regulation.

[0095] Example 9 aims to improve the median flashover voltage. The initial condition is outer-edge band connectivity, resulting in a strong electrical response. The optimization process reduces the proportion of the tangential component of the time-varying trajectory of the electrode system, shortens the duration of the relative humidity plateau, and lowers the volume fraction of soluble contaminants. This reduces the conductivity connectivity deviation index, decreases the leakage current segment energy, and increases the median flashover voltage, ultimately satisfying the constraints and outputting a set of convergent parameters.

[0096] Preferably, the median flashover voltage is obtained through a step-up flashover test, with the voltage steps for rising and falling being consistent, and statistical regression is used to estimate the results of multiple tests.

[0097] This invention centers on a stepped flashover test, aligning the environmental boundary condition time series with voltage condition changes around a unified time reference. Flashover event data is obtained through repeated ascending and descending steps, and a statistical regression method is used to estimate the median flashover voltage. In this application, this method is used to quantify the tolerance threshold under stratified contaminant deposition structures and provides traceable scalar targets and confidence information for closed-loop optimization.

[0098] In principle, the voltage program is discretized into several voltage steps with consistent amplitude and dwell time within the sequence. Each step only changes the voltage amplitude; other exogenous quantities are given by the environmental boundary condition time series set and locked within a preset segment, ensuring causal identification. For each step, the binary result of whether flashover occurred is recorded, along with the synchronous electrical response time series and surface state field time series. The voltage is defined as... In voltage The flashover probability is Logistic regression is used to establish the probability-voltage relationship: ,in These are the regression coefficients. The median flashover voltage is defined as the voltage that makes the probability equal to 0.5: To avoid systematic bias caused by path dependence, the ascending and descending steps use the same voltage step and dwell time, and are included in the same regression during estimation. If necessary, a "path" indicator is introduced for covariate correction. If flashover does not occur at the lowest step or is inevitable at the highest step, censored likelihood is used to process boundary samples during regression to ensure unbiased estimation.

[0099] In this implementation, the voltage step is generated by a programmable power supply, with a step size of [missing information]. Time spent on the steps Set and maintain this setting before the test. The flashover event criterion uses a dual-threshold determination: the absolute value of the leakage current exceeds a certain threshold. And the duration is not less than Or voltage within the time window A relative drop exceeding the threshold occurred within the range. The event indicator quantity is denoted as To ensure time alignment, the start and end times of the voltage steps are time-stamped and written into the metadata. The leakage current time series and the surface state field time series are recorded on the same time key, and the segment energy and increment are calculated at the step boundaries respectively, serving as diagnostic quantities after regression rather than regression independent variables. After each round of ascending steps, a fixed-duration recovery phase is executed, followed by descending steps; the starting and ending steps of the two paths are consistent, forming a symmetrical sampling design.

[0100] The connection with the rest of this invention is reflected in two points. First, the stage residence time is consistent with the segment length in the time series set of environmental boundary conditions, so that the voltage stage forms a one-to-one correspondence with the relative humidity step, wind speed change, and pollutant deposition flux pulse. Second, the flashover criterion is synchronized with the segment, so that the estimation of the median flashover voltage can be clearly attributed to a specific environmental combination, thereby supporting the subsequent decomposition analysis of deposition configuration and connectivity deviation.

[0101] The data processing flow consists of four steps. Step 1: Fragment the leakage current time series and surface state field time series according to steps, and calculate the fragment energy and field increment of each step as quality control indicators. Step 2: Organize the step voltages and flashover indications for both the rising and falling paths to form an independent and identically distributed sample set. Step 3: Use maximum likelihood estimation to estimate the logistic regression coefficients. If necessary, add path indicators and test their significance. Step 4: Calculate the median flashover voltage. Its variance approximation, along with fragment quality labels and censoring information, are written into the electrophysical response dataset.

[0102] Example 10, used for reproducing the experiment. At rated voltage. As a benchmark, the initial step is set as The highest step is set as Step Duration of stay Seconds. The ascending step is executed first, followed by the descending step. The environmental boundary conditions time series dataset maintains a relative humidity plateau, constant temperature, and constant wind speed. A total of 30 step samples were obtained from the two paths, of which 14 samples showed flashover. Logistic regression yielded... , ,thus The fragment energy distribution shows that the leakage current fragment energy is significantly larger in samples with a high proportion of outer connectivity, and the regression residual is also relatively larger, suggesting that the influence of connectivity on the threshold needs to be measured separately in closed-loop optimization.

[0103] Preferably, the deviation index between deposition configuration and conductive connectivity is measured after obtaining the conductivity map by thresholding the surface equivalent conductivity field. The difference in deposition configuration is measured by the optimal transmission distance, and the difference in conductive connectivity is measured by the persistent coherence bottleneck distance.

[0104] This invention unifies the differences in deposition configuration and electrical connectivity into calculable deviation indices during the joint evaluation of electrical properties and surface conditions. The core approach involves first thresholding the surface equivalent conductivity field to generate a conductivity map, and then separately measuring the deposition configuration and electrical connectivity. Through this sequence, continuous field information is mapped to comparable objects in both topological and geometric domains, facilitating its use as an objective variable in closed-loop optimization.

[0105] First, a field-to-plot mapping is performed. The surface equivalent conductivity field at the same time key is selected as input, with corresponding fields from the electrical and physical response dataset. The surface equivalent conductivity, represented in grid coordinates, is... , This represents the spatial index on the parametric mesh of the insulator surface. It sets the threshold. The set of conductive regions is then obtained: On the parameterized mesh, adjacency relationships are established using four-neighbor or eight-neighbor domains to construct a conduction graph. ,in The set of grid nodes that meet the threshold condition. This is the set of edges between adjacent connected nodes. Threshold The determination of the threshold value adopts the principle of maximizing the correlation with the leakage current segment energy: the correlation coefficients between the connectivity component characteristics and the leakage current segment energy at different thresholds are calculated on a set of candidate thresholds, and the threshold with the highest correlation is selected as the working threshold. This approach is a practical detail in the field, ensuring consistency between the continuity diagram and the electrical response. Depositional configuration differences are used to measure the spatial mismatch between the surface density of soluble contaminants on the surface and the target deposition distribution. The normalized surface density field of soluble contaminants is denoted as... The normalized result of the target deposition distribution function on the same grid is denoted as... Both are considered as probabilistic measures defined on the surface parameter domain. Using surface geodesic distance as the metric, the difference in sedimentary configuration is calculated using the first-order optimal transport distance: ,in The parameter domain representing the surface of the insulator. This represents the geodetic distance obtained by weighting along the creeping circuit path and the direction of the umbrella skirt busbar in this parameter domain. Indicates and For all coupling sets at the edge, Represents the insulator surface parameter domain Another location on the surface. The calculation is performed under a mask of the conductivity map, i.e., integration is only performed within the conductivity region to highlight depositional mismatches associated with electrical connectivity. This distance is spatially sensitive to “displacement,” “outward shift,” and “void,” and directly reflects the geometric difference between the depositional amount and the target.

[0106] Conductivity connectivity differences are used to measure the deviation of the topology of the conduction map from the desired connectivity. A topology evolution is constructed using a super-level set filter of the surface equivalent conductivity field. Thresholds are progressively swept from high to low, and the connectivity components and holes in the conduction region at each threshold are calculated to obtain the corresponding persistence map. The persistence map of the measured field is recorded as follows: The persistent graph corresponding to the target connected template is The difference between the two is measured using bottleneck distance: ,in For matching between two persistent graphs, It has an infinite norm. The bottleneck distance is sensitive to topological events such as "channel merging," "bridging," and "separation," can stably characterize the generation and disappearance of conductive paths, is insensitive to local noise, and is suitable for cross-batch comparisons.

[0107] The two types of indicators are used in closed-loop optimization as follows: and As independent objective terms, they are added to the cost function, corresponding to the deposition geometry and conductive topology respectively. Both are calculated under the same time key and the same grid to avoid data alignment errors. To form a unified scale with the electrical response, dimensionless processing is adopted, dividing each term by a reference value derived from the unloaded baseline or historical median level.

[0108] Example 11 is used for verification of a banded connectivity target. A continuous band is defined at the outer edge of the target connectivity template, and the target deposition distribution function is assigned a high weight to this band. After deposition is performed on a wetted platform, the surface equivalent conductivity field is obtained. Thresholding is used to generate a connectivity map and two indices are calculated. The results show that the differences in deposition configuration are small, but the differences in connectivity are large, indicating that the deposition amount is in place, but the conduction path is locally broken. Subsequent closed-loop optimization, without increasing the total deposition amount, only adjusts the tangential field ratio of the electrode system and the wetted segment duration, causing continuous bridging in the connectivity map, reducing the bottleneck distance, and simultaneously increasing the median flashover voltage to the target range.

[0109] Preferably, the dynamic model of surface state and migration behavior is established through sparse structure identification, and the candidate basis function set and state derivative are sparsely regressed to obtain the analytical form. The time-varying trajectory of the electrode system is updated after obtaining the gradient by the adjoint method. The set of pollutant spectrum parameters is adjusted by introducing feasible region constraints of upper limit of leakage current and upper limit of leakage current change rate and using Bayesian optimization.

[0110] This invention addresses the closed-loop process of artificial contamination testing under high-pressure environments. Focusing on the "observable-identifiable-controllable" chain, it first establishes a dynamic model of surface state and migration behavior using sparse structure identification. Secondly, it uses the adjoint method to calculate and update the gradients of the parameters of the time-varying trajectory of the electrode system. Finally, within the feasible region of the upper limit of leakage current and the upper limit of leakage current change rate, Bayesian optimization is employed to adjust the set of contaminant spectrum parameters, simultaneously reducing differences in deposition configuration, conductivity connectivity, electrical response, and median flashover voltage deviation.

[0111] The kinetic model uses the water film thickness field, the surface density field of soluble pollutants, and the surface density field of insoluble particles as state variables. It employs the time series set of environmental boundary conditions, the time-varying trajectory of the electrode system, and the deposition flux as exogenous drivers, and the surface equivalent conductivity field as a derived variable. Let the state vector be denoted as... The exogenous vector is In sparse structure identification, a candidate basis function set containing constant terms, linear terms, cross terms, and physical heuristic terms is constructed. The state derivative is obtained by using denoised derivative estimation, and then sparse regression is performed to obtain the analytical form. ,in This is a sparse coefficient vector. The state variables have the following meanings: The thickness of the water film; The surface density of soluble pollutants; This represents the surface density of insoluble particles. The meaning of exogenous mass is as follows: This is a time series of relative humidity. It is a temperature time series; This is a time series of near-wall wind speeds; This is a time series of the surface normal electric field intensity; and The deposition flux time series are for soluble contaminants and insoluble particles, respectively. To ensure physical consistency, nonnegativity constraints are applied to some coefficients, and a signed prior is applied to diffusion and convection-related terms. Robustness is tested using cross-validation and hold-out fragments. Surface equivalent conductivity is given with minimal necessary coupling. ,in For surface equivalent conductivity, For the baseline conductivity of the dry surface, is the coupling coefficient. The above equation is used to map state variables to measurable electrical quantities, facilitating alignment with electrical and physical response datasets.

[0112] The time-varying trajectory of the electrode system is represented parametrically. The driving potential of each electrode channel is written as spline coefficients or orthogonal basis expansion, and the summation parameter is denoted as... Using a multi-objective cost function as the objective quantity. ,in As an indicator of differences in sedimentary configuration, It is an indicator of differences in electrical conductivity and connectivity. This is an indicator of the time series difference in leakage current. The predicted median flashover voltage, The target median flashover voltage, The weights are non-negative. They are obtained using the adjoint method. Let the dynamic equation be... Lagrange quantity is The adjoint equation is written as , ,in As an adjoint variable, the state trajectory is obtained through forward integration, and then the adjoint trajectory is obtained through backward integration, thus yielding the gradient of the time-varying trajectory parameters of the counter-electrode system, which can be used for line search or quasi-Newton updates. This process is aligned item by item with the data fields under a unified time reference, ensuring that each update can be replayed and verified.

[0113] The contaminant spectrum parameter set includes particle size distribution parameters, charge-to-mass ratio, volume fraction of soluble contaminants, and volume fraction of insoluble particles. Feasibility region constraints are imposed within the experimental safety boundary. , ,in This is a time series of leakage current. This is the upper limit of leakage current. The upper limit of the leakage current change rate is defined. A surrogate model of objectives and constraints is constructed based on a Gaussian process. The next set of spectral parameters is selected within the feasible region using a constrained expected improvement criterion, and iterative updates are performed after observing the response. By using an acquisition function that multiplies the feasible probability by the expected improvement, both safety and performance are incorporated into the decision-making process.

[0114] Example 12 presents the gradient update of the time-varying trajectory of the electrode system. Taking the outer edge strip target as an example, the initial tangential field ratio is too large, resulting in a high conductivity connectivity difference index. After calculating the gradient according to the above-mentioned accompanying steps, the phase difference between the two main channels is reduced and the dwell time in the high field region is shortened. The next round of forward prediction shows that the bottleneck distance decreases, the leakage current segment energy decreases, and the median flashover voltage converges towards the target range.

[0115] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for artificial contamination testing under high pressure, characterized in that, Includes the following steps: A parameterized mesh for the insulator surface is established, the target deposition distribution function and target response set are generated, and the environmental boundary condition time series set is merged to obtain the target deposition and response dataset; In the energized state, a time-varying electric field distribution is formed through the electrode system, soluble pollutant aerosols and insoluble particulate aerosols are supplied to the surface of the insulator and particulate charge control is implemented. Based on the time series set of environmental boundary conditions and the supply switching, a layered pollutant deposition structure is constructed, surface state field data is collected, and an in-situ deposition state dataset is generated. Loading is performed in a segmented manner based on the time series set of environmental boundary conditions and voltage condition changes. The loading sequence includes at least three types of segments: relative humidity step, temperature step, wind speed change, pollutant deposition flux pulse, and voltage transient. The start and end times of each segment are recorded with time stamps and synchronized with the timestamp of the acquisition channel. The electrical response time series and the surface state field time series are acquired. The surface state field time series includes at least the water film thickness field and the surface equivalent conductivity field time series. The median flashover voltage is obtained by flashover test and statistical methods. The leakage current segment energy, the surface equivalent conductivity field increment, and the water film thickness field increment are calculated to generate an electrophysiological response dataset. Within the leakage current constraint, based on the electrophysiological response dataset and the target deposition and response dataset, a dynamic model of surface state and migration behavior is established with the deviation indices of deposition configuration and conductive connectivity, and electrical response and median flashover voltage as optimization objectives. The time series set of time-varying trajectory of electrode system and environmental boundary conditions is updated, the pollutant spectrum parameter set is adjusted, and the convergence parameter set is output.

2. The method according to claim 1, characterized in that, The environmental boundary condition time series set consists of relative humidity time series, temperature time series, wind speed time series and pollutant deposition flux time series, and is sampled and aligned with a unified time reference as input to the target deposition and response dataset.

3. The method according to claim 1, characterized in that, The electrode system is a multi-electrode shaping array. The phase and amplitude of the electrode channels are set independently. The electrode channels are arranged with a fixed geometric spacing. A time-varying electric field distribution that moves along the axial and radial directions of the insulator is formed by sequential driving.

4. The method according to claim 1, characterized in that, Particle charge control employs corona charging, with the charge polarity being either unipolar positive or unipolar negative. The particle charge is measured using a Faraday cylinder in conjunction with an electrostatic meter, and the measured values ​​are used to correct the aerosol supply parameters.

5. The method according to claim 1, characterized in that, The construction sequence of the stratified pollutant deposition structure is as follows: first, insoluble particulate aerosols are supplied to form a particulate skeleton; then, in the wetting stage, the relative humidity is increased and soluble pollutant aerosols are supplied to form a coating layer; and finally, in the drying stage, the relative humidity is reduced to stabilize the stratified structure.

6. The method according to claim 1, characterized in that, The surface state field data includes the water film thickness field, the surface equivalent conductivity field, and the soluble pollutant surface concentration field. The water film thickness field is obtained by conversion through non-contact optical measurement, the surface equivalent conductivity field is obtained by segmented measurement through a multi-point electrode array and inversion, and the soluble pollutant surface concentration field is obtained by either micro-elution conductivity measurement or spectral measurement.

7. The method according to claim 1, characterized in that, The deviation between deposition configuration and electrical connectivity is measured based on the conduction map obtained by thresholding the surface equivalent conductivity field. The difference in deposition configuration is measured by the optimal transmission distance, and the difference in electrical connectivity is measured in the following way: A topology evolution is constructed using a super-level set filter of the surface equivalent conductivity field. The threshold is gradually swept from high to low, and the connectivity components and holes of the conduction region under each threshold are calculated. The persistence map of the measured field and the persistence map corresponding to the target connectivity template are obtained respectively. The bottleneck distance between the two persistence maps is used to characterize the difference in conductivity connectivity.

8. The method according to claim 1, characterized in that, The dynamic model of surface state and migration behavior is established through sparse structure identification. The candidate basis function set and the state derivative are sparsely regressed to obtain the analytical form. The time-varying trajectory of the electrode system is updated after obtaining the gradient by the adjoint method. The set of pollutant spectrum parameters is adjusted by introducing feasible region constraints of upper limit of leakage current and upper limit of leakage current change rate and using Bayesian optimization.

Images