A salt mist pollution risk grading early warning and operation and maintenance disposal method for coastal power distribution equipment

Abstract

The application provides a kind of salt fog pollution risk grading early warning and operation and maintenance method for coastal power distribution equipment, obtains coastal power distribution equipment environmental monitoring, operation and disposal feedback data;Build a scouring retention factor containing rainfall and artificial reduction coefficient;Combined with the environmental exposure intensity recursive calculation pollution accumulation and mapping as salt fog pollution risk index;Extract wind power output fluctuation and anti-send state to build wind power operation disturbance auxiliary index;Fusion above-mentioned risk index and auxiliary index get comprehensive alarm score;Using threshold value comparison comprehensive alarm score with delay judgment rule to determine risk level;When risk is serious and operation is nervous, access support device to perform physical adjustment.The application can accurately evaluate the dynamic accumulation state of equipment salt fog and reasonably schedule limited operation and maintenance resources, improve the anti-pollution flashover capability and operation safety of distribution network.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation status monitoring and intelligent equipment operation and maintenance technology, and in particular to a method for risk classification, early warning and operation and maintenance of salt spray pollution for coastal power distribution equipment. Background Technology

[0002] In coastal areas, power distribution equipment is exposed to high-salt and high-humidity environments for extended periods, making its insulation surfaces prone to salt spray contamination. With prolonged accumulation and fluctuating weather conditions, the insulation level affected by salt spray gradually declines, potentially leading to surface discharge or even flashover-induced tripping faults. Current equipment condition monitoring methods largely rely on periodic manual inspections or alarms based on single environmental parameter thresholds. The accuracy in assessing the interaction of multiple physical factors, including dynamic salt spray adhesion, rain erosion, and the mitigating effects of manual cleaning, needs further optimization.

[0003] On the other hand, coastal power distribution networks typically integrate a large number of distributed wind power and other renewable energy resources. Frequent fluctuations in wind power output and potential active power backflow can generate transient overvoltages or inrush currents in localized areas of the distribution network. When these high-frequency electrical disturbances are superimposed on distribution equipment with insulation damage caused by salt spray pollution, the probability of flashover faults is further increased. Current early warning systems rarely conduct multi-dimensional, in-depth integrated assessments of the dynamic physical trends of pollution development and the electrical operational risks of localized renewable energy fluctuations in the power grid.

[0004] When severe weather conditions cause widespread equipment malfunctions, the dispatching system faces immense pressure from a large volume of concurrent dispatch orders due to limited on-site repair personnel and cleanup vehicles. Existing emergency dispatch mechanisms often lack a comprehensive consideration of equipment importance, on-site accessibility, and resource constraints, and are prone to generating frequent duplicate dispatches when environmental parameters fluctuate critically. Furthermore, the lack of quantitative data feedback on cleanup effectiveness and risk status reset after maintenance operations indicate room for improvement in the overall resource allocation efficiency and closed-loop management of the maintenance strategy.

[0005] Furthermore, existing condition assessment models have significant drawbacks in data processing and algorithmic aspects. On the one hand, the raw physical signals (such as voltage, current, and salt spray concentration) collected by monitoring systems are often mixed with Gaussian white noise and industrial impulse interference. Traditional mean filtering tends to smooth out transient abrupt changes in pre-fault characteristics during denoising and has poor interpolation performance for missing continuous data, leading to signal distortion. On the other hand, existing multi-parameter fusion early warning models mostly employ fixed-weight linear weighting methods, assuming that the contributions of environmental and electrical factors to insulation degradation are linearly independent. However, in physical reality, multi-dimensional factors exhibit complex nonlinear coupling relationships (for example, high humidity only significantly accelerates salt spray adhesion within specific temperature and wind speed ranges). Fixed-weight models cannot capture such dynamic interactions, resulting in insufficient early warning accuracy under complex operating conditions. Summary of the Invention

[0006] The purpose of this invention is to provide a method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment. It aims to solve the technical problems in the existing technology, such as the single dimension of risk assessment for insulation degradation of power distribution equipment under the superposition of high salt spray exposure environment and high frequency fluctuations of distributed wind power, the easy generation of false alarms due to high frequency fluctuations, and the lack of priority closed-loop scheduling strategy under limited operation and maintenance resources.

[0007] This invention provides a method for risk classification, early warning, and operation and maintenance management of salt spray pollution in coastal power distribution equipment, including: Acquire environmental monitoring data, equipment operation data, and response feedback data for coastal power distribution equipment; Construct a scour retention factor that includes rainfall scour coefficient and human intervention reduction coefficient; Based on the environmental monitoring data, the environmental exposure intensity is determined. The scouring and retention factor is then used for discrete recursive updates to calculate the pollution accumulation amount, which characterizes the continuous deposition and reduction state of salt spray. The pollution accumulation amount is then mapped to a salt spray pollution risk index. Extract the wind power output fluctuation characteristics and equivalent active power feedback status from the equipment operation data to construct auxiliary indicators for wind power operation disturbances; The salt spray pollution risk index is combined with the wind power operation disturbance auxiliary index to calculate a comprehensive alarm score; The risk level of the coastal power distribution equipment is determined by comparing the comprehensive alarm score using a preset hysteresis judgment rule that includes rising and falling thresholds. When the risk level triggers a severe warning and the voltage deviation of the associated nodes and the line load rate exceed the set operating limits, the coastal power distribution equipment is controlled to connect to a reactive power support device or a mobile temporary support device to perform physical support adjustments.

[0008] Optionally, after acquiring the environmental monitoring data and the equipment operation data, the method further includes: Sliding window smoothing is applied to data with missing samples; When the missing proportion is lower than the first set threshold, the data is completed using the effective mean. When the consecutive missing proportion is higher than the first set threshold, the missing flag is recorded and the data reliability is calculated. When the data credibility is lower than the second set threshold, a verification flag is output to intercept the data.

[0009] Optionally, determining the environmental exposure intensity specifically includes: The exposure factor of the equipment location is calculated based on the distance from the sea, windward conditions, and shielding conditions of the area where the coastal power distribution equipment is located. After normalizing the wind speed, humidity, temperature, and salt spray concentration in the environmental monitoring data, the environmental exposure intensity is obtained by weighted summation based on the equipment location exposure factor.

[0010] Optionally, the specific method for performing discrete recursive updates in conjunction with the scour retention factor is as follows: The pollution accumulation of the previous period is multiplied by a preset pollution retention coefficient according to the sampling period, and the product of the environmental exposure intensity of the current period and the scour retention factor is added to achieve dynamic tracking and calculation of rainfall reduction, artificial reduction and salt spray accumulation.

[0011] Optionally, in the step of calculating the comprehensive alarm score, a voltage deviation auxiliary index calculated based on the voltage deviation of the device-associated node and a line load auxiliary index calculated based on the line load rate are also incorporated.

[0012] Optionally, the step of using a preset hysteresis judgment rule that includes an upward threshold and a downward threshold to compare the comprehensive alarm score to determine the risk level specifically includes: To avoid frequent changes in risk level, the rising threshold is configured to be greater than the falling threshold to form a hysteresis interval; The risk level will be raised by one level only if the comprehensive alarm score is higher than the rising threshold for a consecutive preset number of sampling periods; The risk level will be downgraded by one level only if the comprehensive alarm score is lower than the decrease threshold for a consecutive preset number of sampling periods.

[0013] Optionally, based on the risk level, the system may also include dispatcher-level handling operations: The priority score is calculated by combining the risk level, equipment importance, road accessibility, and estimated arrival time, and an on-site disposal list including on-site verification or pollution cleanup actions is generated. When the available resources in the current cycle are insufficient to cover the on-site disposal list, actions that cannot be performed at present are filtered out based on resource constraints, and coastal power distribution equipment with a severe risk level located on the main line or important user side is forcibly prioritized for disposal.

[0014] Optionally, when determining the pollution removal action, the expected risk reduction, the matching degree between the execution cost and the risk level are calculated, and the final disposal action is selected based on the principle of maximizing recommended utility.

[0015] Optionally, for coastal power distribution equipment that has reached a critical state, individual inspection actions are prohibited from being issued in the on-site handling list. Instead, a continuous process instruction of flushing-retesting-inspection and repair is forcibly generated and issued.

[0016] Optionally, after receiving the processing feedback data transmitted from the field, the following closed-loop update operation is performed: Based on the inspection results, salt density retest value, and flushing completion sign, the corresponding manual treatment reduction coefficient is used to update the pollution accumulation again; If the retest results are satisfactory, the risk level of the control equipment will be downgraded and the system will enter a safe observation state. If the retest result is unsatisfactory, the risk level will be locked and upgraded, and a secondary treatment suggestion will be automatically generated. Based on the early warning data and the response feedback data for each period, the hit rate and alarm lead time of the current early warning strategy are evaluated and updated.

[0017] The present invention has achieved the following beneficial effects: The method provided by this invention achieves dynamic quantitative tracking of salt spray accumulation and stripping status on equipment surfaces by coordinating environmental exposure intensity and erosion retention factors, overcoming the limitations of single-parameter monitoring. Simultaneously, it extracts wind power output fluctuations and backfeed status as auxiliary indicators, and, combined with voltage and load characteristics, objectively assesses the composite risks arising from the superposition of salt spray degradation and electrical disturbances. It introduces hysteresis judgment rules including rising and falling thresholds, effectively reducing frequent repetitive alarm work orders under critical conditions. In the operation and maintenance scheduling stage, it fully considers road accessibility and on-site available resource constraints to calculate priority scores, optimizing the allocation order of operation and maintenance resources under multi-point concurrent alarms. After completing on-site cleaning, it updates the pollution accumulation based on measured feedback data in a closed loop, improving the accuracy of pollution flashover warnings and the overall scheduling efficiency of coastal power distribution networks.

[0018] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram illustrating the spatial distribution and location exposure risk assessment of coastal power distribution equipment in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the dynamic change of the integrated salt spray pollution risk value over time and the graded early warning threshold in an embodiment of the present invention; Figure 3 This is a radar chart comparing the multi-dimensional effectiveness of the graded early warning closed-loop processing method with other early warning methods in this embodiment of the invention. Detailed Implementation

[0021] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0022] A specific implementation environment for a salt spray pollution risk classification, early warning, and operation and maintenance management method for coastal power distribution equipment includes a hardware monitoring network, a power distribution automation master station system, and an operation and maintenance dispatch server deployed in the coastal power supply area. The hardware monitoring network includes micro-meteorological sensor groups, salt spray concentration monitors, and electrical quantity acquisition terminals installed at coastal power distribution substations, power line poles, and around key power distribution equipment. The micro-meteorological sensor groups are configured to continuously collect meteorological parameters, including ambient wind speed, relative humidity, ambient temperature, and rainfall, according to a preset sampling period. The electrical quantity acquisition terminals include voltage transformers and current transformers, used to collect real-time voltage, line current, and active power flow data at each power grid node. The operation and maintenance dispatch server includes a communication interface, a processor, and a memory, and receives the environmental monitoring data, equipment operation data, and handling feedback data transmitted from mobile operation terminals via fiber optic networks or wireless communication links.

[0023] In the initial stage of data acquisition and processing, the communication interface continuously receives raw data streams from various sensors. This includes acquiring environmental monitoring data, equipment operation data, equipment location data, and response feedback data for the coastal power distribution equipment. The environmental monitoring data specifically covers wind speed, relative humidity, temperature, salt spray concentration, and rainfall and duration sequences. The equipment operation data includes node voltage, line load rate, current data, switch status signals, alarm records, and load fluctuation power values ​​for nodes physically connected to the coastal power distribution equipment. The equipment location data is obtained by querying a static equipment ledger database and includes equipment type code, years of operation, geographical coordinates of the equipment's location, straight-line distance from the coastline, and installation altitude. The response feedback data is generated and uploaded by on-site maintenance personnel via their mobile terminals and includes manually entered inspection result text, sensor-measured salt density retest results, confirmation flags for manual flushing operations, records of repaired and replaced parts, and Boolean flags indicating successful retesting.

[0024] After acquiring the environmental monitoring data and the equipment operation data, the processor performs data preprocessing and reliability verification steps for various types of underlying data streams. Because field sensors may experience data jumps or short-term communication interruptions under complex electromagnetic environments and extreme weather conditions, the processor performs sliding window smoothing on the received m-th type of raw sample time series. The mathematical model configuration for the sliding window smoothing is as follows: ; In this formula, Indicates the current sampling time node The output after smoothing calculation using a sliding window is the first... Smoothing data values ​​by sampling quantity; The integer length of the sliding window pre-set in memory represents the number of historical valid sampling points included in a single smoothing operation; Indicates at time node The data is collected and recorded in the register or cache at the location. Similar to the original sample size value; The summation index parameter for sliding calculations is set to a range from... Increment to .

[0025] During the sliding window smoothing calculation, the processor synchronously monitors the continuity of valid data points within the window. When a missing sampling point is detected within the window, and the number of missing sampling points is proportional to the length of the sliding window... When the ratio is lower than the first set threshold, the processor calculates the arithmetic mean of all valid sampled points within the current window and uses this valid mean to fill in the missing data points in memory, then continues to perform smoothing operations. When consecutive missing sampled points are detected and their missing ratio is equal to or higher than the first set threshold, the processor stops the mean filling operation, uses the valid value recorded in the previous normal cycle of the data channel as the current value, and records the corresponding missing flag for the data in the system log for that time period. As a preferred and advanced implementation of the above sliding window smoothing process, for complex electrical quantity data with both high-frequency noise and signal abrupt changes, this embodiment introduces a multi-scale denoising and reconstruction method based on wavelet transform. Specifically, the processor performs multi-scale wavelet decomposition on the time series, separating high-frequency noise components and low-frequency signal components at different scales. For data segments marked as missing, the system does not use simple mean filling, but extracts the valid wavelet coefficients of adjacent scales for nonlinear interpolation reconstruction. This method effectively improves the suppression of mixed noise while preserving key transient change features, thus enhancing the signal-to-noise ratio and fidelity of the preprocessed data and laying a high-precision data foundation for subsequent risk assessment. To quantify data quality, the processor calculates the data reliability of the time series based on the number of missing markers. ; In this formula, Represents the calculated first Class data at time nodes The reliability metrics for data within the evaluation period; This indicates that the current length is The total number of sampling points recorded and marked as missing within the sliding window; This represents the fixed length value of the sliding window.

[0026] The processor compares the calculated data reliability with a preset second threshold. During calculation cycles where the data reliability is lower than the second threshold, the processor triggers and generates a verification flag via the system output. This verification flag is used to logically intercept the computational flow associated with the device in subsequent comprehensive alarm scoring and dispatch calculation stages, preventing the system from outputting erroneous operation and maintenance control commands based on low-quality or long-term missing distorted data.

[0027] After smoothing and reliability verification, for multi-source data with different units and dimensions, the processor performs upper and lower limit normalization operations to map various parameters to a unified numerical range. The normalization logic is configured as follows: ; In this formula, Indicates the first Class sampling data at time nodes The standard dimensionless value generated after normalization; This represents the data value output after the above smoothing process; This indicates that the system database is the first... The upper limit benchmark value for engineering physics set by the class data; This represents the lower limit benchmark value for engineering physics set for this type of data; This is a preset anti-overflow constant for the system. Its value is a definite, extremely small positive real number used to prevent division by zero errors that occur when the upper and lower limits of the project are equal or the difference is extremely small.

[0028] Through the above calculations, environmental monitoring data and some equipment operation data are converted into standard values. Furthermore, the processor sequentially constructs structured equipment feature vectors for candidate coastal power distribution equipment within the target jurisdiction, which are then used for subsequent high-dimensional state matrix operations. The memory construction form of the feature vector is defined as follows: ; In this formula, This indicates that for the first [unit / item] in the power grid topology The candidate coastal power distribution equipment is at the following time point. The generated feature vector; , , , , These represent the standardized numerical elements of the current wind speed, relative humidity, temperature, salt spray concentration, and rainfall, respectively, output by calling the normalization module. This indicates that the data obtained by extracting electrical quantity acquisition terminal data is related to the equipment. The numerical element representing the node voltage deviation associated with the location; This represents the equipment location exposure factor element calculated based on equipment coordinates and terrain parameters; This represents the equipment commissioning period factor element, calculated and normalized from the equipment commissioning date. This represents the historical defect factor elements obtained from the retrieval and statistics of the equipment's historical defect database; This indicates the handling indicator elements extracted based on the latest handling feedback data, reflecting the status of flushing, inspection, or maintenance.

[0029] The processor calculates the environmental exposure intensity based on the elements contained in the device's feature vector. Since the adhesion rate of salt spray pollution to physical surfaces is controlled by a combination of meteorological and geographical factors, the system measures the contribution of multiple meteorological dimensions by setting independent weighting coefficients. The specific calculation logic is as follows: ; In this formula, Represents the calculated first... At the current time point, the coastal power distribution equipment The environmental exposure intensity value; , , , These represent the normalized values ​​of wind speed, salt spray concentration, relative humidity, and temperature extracted from the feature vector, respectively. This represents the exposure factor of the device location; , , , , These represent the fixed weighting coefficients determined by the system during the configuration phase, corresponding to wind speed, salt spray concentration, relative humidity, temperature, and location factors, respectively.

[0030] Regarding the equipment location exposure factor used in the above formula, the processor performs a comprehensive quantification of spatial disadvantages by extracting the equipment's two-dimensional geographic information data and digital elevation model data. The calculation process includes measuring the Euclidean straight-line distance from the equipment installation point to the nearest coastline, analyzing the windward cross-section of that point under the prevailing local wind direction, and measuring the average height and obstruction angle of physical obstacles within a fixed radius around the equipment. Figure 1 As shown. The formula for calculating the location exposure factor is: ; In this formula, Indicates that it is for the equipment Calculated equipment location exposure factor; This represents the distance from the sea to the equipment installation location after normalization. This represents the windward exposure calculated based on topographic features and the prevailing wind field vector. This represents the unobstructed condition parameter calculated based on the distribution of surrounding obstacles; , , These represent the spatial weighting coefficients corresponding to distance from the sea, windward features, and unobstructed features, respectively.

[0031] Based on the obtained environmental exposure intensity, and considering the physical fact that salt spray accumulation on the insulation surface of power distribution equipment is not unidirectionally increasing but is subject to the dual stripping effects of rainfall and manual cleaning, the system constructs a scour retention factor that includes a rainfall scour coefficient and a manual treatment reduction coefficient. This factor characterizes the proportion of existing contaminants remaining on the insulation surface within one calculation cycle. Its value is updated by the processor according to the following formula: ; In this formula, The device obtained from the calculation At the time point The scouring retention factor value; This represents the normalized rainfall value extracted at the current time point; The rainfall scouring coefficient represents the ability of a unit amount of rainfall to remove salt pollution. A binary indicator showing whether manual decontamination was performed during the current period; The manual removal reduction coefficient represents the degree of thoroughness of a specific manual cleaning action.

[0032] Based on the environmental exposure intensity and erosion retention factor obtained above, the processor performs discrete recursive calculations of the salt spray load on the equipment surface over time. The system maintains a historical cumulative state register in memory for each monitored device. Upon arrival of each sampling clock signal, the historical values ​​from the previous period are extracted and superimposed with the current period's comprehensive impact increment. This discrete recursive update is executed according to the following formula: ; In this formula, Indicates the target device At the current time point The calculated updated value of cumulative pollution; This indicates that the device was at the previous discrete sampling time node. Historical values ​​of accumulated contamination stored in memory; This represents the system's preset contamination retention coefficient; This represents the calculated environmental exposure intensity within the current period; This represents the scouring retention factor within the current cycle; This represents the fixed time step between two adjacent operation cycles. Understandably, for coastal power distribution equipment marked in the ledger as being in a high salt spray exposure area, having a large number of historical defect data records, or having a long time since the last cleaning cycle, the processor will assign it a higher pollution retention coefficient during the initialization phase to match the physical deterioration reality that salt stains are more likely to adhere and be difficult to remove after the surface material ages.

[0033] Subsequently, the system converts the calculated pollution accumulation in the non-standard range into a risk index characterizing the equipment failure tendency. This mapping process employs a continuous nonlinear transformation to smoothly converge the physical accumulation values ​​to the standardized range. The specific mapping formula is as follows: ; In this formula, Indicates the equipment At the time point The calculated output is a salt spray pollution risk index between zero and one. This represents the cumulative amount of pollution obtained from the above discrete recursive calculation; Represented as device The configured equipment sensitivity factor, which is related to the equipment's insulation creepage distance and rated withstand voltage class; natural constant. It is the base for exponentiation.

[0034] Besides the risk of static pollution, the overall security of coastal power distribution networks is also constrained by the electrical operating status of equipment. The processor performs parallel calculations to extract auxiliary electrical indicators from equipment operating data. First, it extracts the effective values ​​of the three-phase voltages of network nodes directly physically connected to candidate equipment and calculates the degree to which they deviate from the normal operating range, i.e., the voltage deviation auxiliary indicator. ; In this formula, Indicates at a time node Calculated and related equipment Related voltage deviation auxiliary indicators; This represents the actual operating voltage value of the node reported in real time by the acquisition terminal. This represents the nominal rated voltage constant of the distribution node; This indicates the maximum allowable voltage deviation range threshold set according to the power grid operation regulations for this node.

[0035] Furthermore, the processor extracts the load current data of the feeder or branch line where the target device is located and calculates the auxiliary line load index of the device. This calculation focuses on evaluating the accelerated thermal breakdown effect that may be caused by surface salt contamination under heavy load and heating conditions. The specific calculation logic is as follows: ; In this formula, Indicates at a time node Calculated and related equipment Related line load auxiliary indicators; This indicates the real-time line load rate of the line where the device is located, calculated at the current moment. This indicates the set overload startup threshold that triggers the evaluation of this metric; This represents the load sensitivity coefficient used to adjust the weight of this indicator on the final score; The function is used to ensure that when the line load rate is lower than the heavy load start threshold, the output of this auxiliary indicator is always zero.

[0036] For distributed wind power resources with a high proportion of grid connection in coastal areas, high-frequency fluctuations in wind power output can easily generate transient overvoltages or inrush currents in local distribution networks. When such electrical disturbances affect distribution equipment with degraded insulation due to salt spray adhesion, they can easily induce surface flashover faults. To address this, the processor is equipped with a dedicated data extraction and identification module to extract wind power output fluctuation characteristics and equivalent active power feedback status from the equipment operation data, constructing an auxiliary index for wind power operation disturbances. The processor periodically reads the total active power output of the wind farm, calculates the output differential rate within a unit time step, and simultaneously reads the active power direction vector on the substation incoming side. When power is detected flowing in the reverse direction from the distribution network to the main grid, a feedback flag is activated. The specific calculation formula for the auxiliary index for wind power operation disturbances is as follows: ; In this formula, Indicates at a time node Calculated auxiliary indicators of wind power operation disturbance in the system's location area; Indicates in The absolute change difference in regional wind power active power output collected within the time step; This indicates the rated total installed capacity of grid-connected wind turbine units in the region; This indicates the status flag for the back transmission of equivalent active power recorded at the current time point; This represents the weight scalar assigned to the rapid fluctuation characteristics of wind power output; This represents the weight scalar assigned to the active power feedback state.

[0037] For the equivalent active power feedback status flag used in the formula, the processor performs binarization judgment based on the equivalent active power symbol directly reported by the acquisition terminal from the substation side. The mathematical expression of its judgment logic is as follows: ; In this formula, This is the equivalent active power feedback status flag quantity that we are looking for; Indicates at a time node The instantaneous value of equivalent active power measured by the power meter at the substation is defined as positive when flowing into the distribution network and negative when flowing out of the distribution network and being transmitted to the main grid.

[0038] In addition to a static snapshot of the risk at the current moment, the development trend of pollution is crucial for early warning and assessment. The processor retrieves data from devices in memory. In a historical cohort of salt spray pollution risk indices from different sampling periods, the risk growth gradient difference between the current moment and a specific historical moment is calculated, i.e., the risk growth trend: ; In this formula, The device obtained from the calculation At the time point Risk growth trend data value; This represents the salt spray pollution risk index calculated at the current time point; This indicates looking back at history on a timeline. The historical value of the salt spray pollution risk index recorded by the device during each sampling period; The number of time span cycles preset for the system.

[0039] Subsequently, the system performs multi-dimensional risk aggregation calculations. It integrates the salt spray pollution risk index, voltage deviation auxiliary index, line load auxiliary index, and wind power operation disturbance auxiliary index, along with risk growth trend data characterizing dynamic degradation, to calculate a comprehensive alarm score used for final equipment status assessment. This score aggregation is accomplished through a pre-set array of normalized multipliers. ; In this formula, Indicates the target device At the time point Summarize the generated comprehensive alarm score; , , , , These correspond to the salt spray pollution risk index, voltage deviation auxiliary index, line load auxiliary index, wind power operation disturbance auxiliary index, and risk growth trend calculated and output in the previous steps, respectively. , , , , The system console assigns positive real-valued feature weight parameters based on the importance of different indicators. Furthermore, to overcome the limitation of fixed linear weighting in capturing complex nonlinear relationships among multiple variables, this system constructs a deep temporal fusion network based on an attention mechanism to replace the aforementioned linear summation operation in a more advanced implementation architecture. The system uses the time series of the salt spray pollution risk index, voltage deviation auxiliary indicator, line load auxiliary indicator, and wind power operation disturbance auxiliary indicator as multi-channel inputs (the total number of input features is M). In the deep temporal fusion network, the data from each channel first passes through a bidirectional LSTM encoder to generate hidden state representations. Then, the self-attention mechanism layer is entered, and the specific mathematical model for calculating the comprehensive alarm score is as follows: Dynamic weights Calculated from the attention scoring function: raw fractions In this model, the query vector The key vector represents the global context state at the current moment. Representing single-item feature information, this network achieves adaptive and dynamic weighting of each input feature weight, accurately capturing the nonlinear coupling effect between salt spray accumulation and electrical disturbance.

[0040] Based on the generated comprehensive alarm score sequence, the system needs to discretize it into executable categorized early warning instructions. The processor uses a preset hysteresis judgment rule that includes rising and falling thresholds to compare the comprehensive alarm scores and determine the risk level of the coastal power distribution equipment. Within the storage area, the system divides the equipment's risk level into four physical levels from low to high: safe zone, attention zone, early warning zone, and severe zone. Figure 2 As shown. The basic conditions for determining the grade interval are shown in the following series of inequalities: ; ; ; ; In this set of defining formulas, Indicates assigning to the device Risk level state variables; , , , Fixed identifiers representing the safe zone, the attention zone, the warning zone, and the critical zone, respectively; This indicates the current overall alarm score of the device; , , Let each represent a basic cascaded threshold constant required to divide the above four intervals, and satisfy the following conditions: The numerical progression relationship.

[0041] To eliminate the generation of frequently recurring work orders caused by minor oscillations in the comprehensive alarm score around a certain threshold due to critical environmental parameters, the processor deploys a mandatory hysteresis judgment rule in the level switching control logic. The system independently decomposes and configures corresponding rising and falling thresholds for each boundary cascaded threshold. The specific triggering condition for an upward jump in equipment risk level is as follows: ; In this state transition formula This indicates the calculated new risk level of the device at the current time point; Indicates the risk level of the previous period; In continuous time series The comprehensive alarm score sample value within; Indicates the current level The corresponding specific rise threshold; This parameter represents the number of sampling periods that the system is required to continuously meet the condition. This rule mandates that the comprehensive alarm score must be continuously satisfied only if the condition is met. The processor will only raise the risk level by one level and update the status register when all sampling periods are strictly higher than the rising threshold.

[0042] Similarly, the safety trigger condition for the equipment risk level to fall back is configured as follows: ; In this formula, Indicates the current level The threshold for transitioning to a lower level; the physical meanings of the remaining parameters are the same as those in the rising rule formula. This rule requires that the overall alarm score be continuous only if... The processor will only perform a risk level downgrade operation when all sampling periods are strictly below the aforementioned drop threshold.

[0043] In the configuration and implementation of the above hysteresis judgment rules, the rising threshold and falling threshold set by the system hardware must satisfy the following strict inequality relationship to form an effective hysteresis dead zone: ; In this formula, This represents the value of the rising threshold configured for the same risk state switching boundary; This represents the descent threshold value configured for the same boundary. The absolute value of the difference between these two values ​​constitutes the hysteresis band used for wave absorption and oscillation prevention in the control algorithm.

[0044] When the system identifies that the risk level of some coastal power distribution equipment has exceeded the limit and entered the attention, warning, or severe level state, the processor initiates dispatching operations for all candidate devices that have triggered alarms. To ensure the maximum utilization of limited on-site resources, the processor calculates a priority score for each abnormal device based on the comprehensive alarm score, equipment importance coefficient, road and on-site accessibility indicators, available resource matching coefficient, estimated arrival time, and historical defect factors. The calculation structure is as follows: ; In this formula, This indicates that the system is a specific abnormal device. The generated priority scoring results for work order sorting; This is the overall alarm score for the device; This represents the equipment importance coefficient, read from the equipment attribute library, indicating the criticality of this power distribution node. This indicates the road and on-site accessibility sign parameters obtained by calling an external traffic condition interface; This indicates the matching coefficient between the available washing vehicles and the types of disposable resources required by the equipment in the current work area; This represents the estimated arrival time of the repair team, as predicted by the path planning algorithm. Historical defect factors; to This represents the scheduling weight scalar that the system assigns to various constraints and state parameters to balance various indicators.

[0045] Based on the priority scores ranked in descending order, the system generates a preliminary list of on-site handling procedures. Subsequently, when determining the specific action instructions for on-site verification or contamination removal, the processor traverses the built-in dictionary of four standard actions: enhanced monitoring, on-site verification, contamination removal, and handling-related alarm cancellation. For each candidate action, the system calculates the theoretically recommended utility that can be generated by applying the action to the device. The calculation logic for this utility is directly proportional to the expected benefit and inversely proportional to the execution resistance. ; In this formula, This indicates the calculated execution of a specific action. For equipment Recommended utility assessment value; This indicates that the action was recorded in the statistical database. The parameter representing the expected reduction in risk that historical averages can bring; This indicates the availability parameters of consumables and working hours required to perform the action at the current moment; This indicates the action Specific nature and current equipment risk level The rule matching degree coefficient in the knowledge base; This represents the estimated cost of deploying personnel and equipment, calculated accordingly. This indicates the estimated response time required to complete the operation.

[0046] Based on the matrix sequence obtained through traversal calculation, the system selects the final action as the work order output according to the principle of maximizing recommendation utility. ; In this formula, The system is indicated as a device. The selected optimal final recommended action identifier; The function is represented in the dictionary of the entire set of standard actions. Retrieve and extract the aforementioned recommendation utility function The specific action variable that obtains the maximum value of the entire domain. .

[0047] Before officially issuing the on-site handling list, the processor performs strict resource constraint boundary checks. Given the limited number of physical entities with available resources for the current period, the system filters out actions that cannot be executed at present based on resource constraints. The system constructs a system of linear constraint inequalities consisting of binary variables to limit the total resource consumption of all dispatched work orders: ; ; In these two sets of formulas, This indicates that the current batch plan will target all devices that triggered the alarm. Perform specific actions The required quantities are summed and accumulated. Whether the representation system ultimately determines the device Arrange and distribute disposal actions A binary variable whose range is strictly limited to the set {0, 1}, where 1 represents the execution of dispatch. Indicates cancellation or suspension; This indicates the action to be performed within the current operating cycle, obtained through the inventory query interface. The upper limit constant for the number of available resources, including actual on-site personnel, specialized vehicles, or insulating flushing equipment.

[0048] When the system finds that the available resources for the current period are insufficient to cover the initially generated on-site handling list while solving the above-mentioned set of constraint inequalities, the processor, in adjusting and optimizing its strategy, forces priority scheduling of coastal power distribution equipment with a severe risk level and located on the main line or important user side, and locks the resource allocation status for these high-priority objects. For coastal power distribution equipment that has reached a severe state, the processor, in its instruction generation logic, shields and prohibits the issuance of on-site verification or inspection actions that are merely observational in nature. Instead, it forces the system to generate and issue a set of continuous long-range instruction codes for flushing, retesting, inspection, repair, and verification that are inseparable from the work order system based on a fixed template.

[0049] When the aforementioned risk level triggers a severe warning and the equipment's operating status is extremely strained—that is, when electrical parameters such as voltage deviation and load rate have approached or even exceeded safe operating limits—the processor, without waiting for manual dispatch, directly enters the physical intervention protection program. The processor sends hard-contact control frames directly to the substation's monitoring and control devices and line automation terminals via a dedicated industrial control LAN. It controls the busbar where the coastal power distribution equipment is located to directly connect to the reactive power support device, or remotely close and connect to a mobile temporary support device to perform physical support adjustments. By injecting reactive current into the local network, it stabilizes the voltage distribution and delays large-scale power outages caused by localized flashover. This physical support adjustment has a clear electrophysical mechanism: when the surface of the coastal power distribution equipment is severely contaminated with salt and under heavy load, the leakage current on the surface of the local insulating medium increases significantly, triggering a localized arc (i.e., a "dry-line arc"). If the node voltage drops due to wind power disturbance or heavy load, it will cause extreme distortion of the electric field distribution, thereby accelerating the penetration of the dry-line arc. By forcibly connecting the reactive power support device, reactive current can be rapidly injected into local nodes and the node voltage can be increased, changing the gradient of the electric field distribution along the surface. This physically suppresses the arc elongation trend of specific weak insulation points and effectively reduces the probability of flashover.

[0050] After the on-site maintenance team completes the continuous process of flushing-retesting-inspection and repair, or other routine cleaning tasks, the frontline mobile terminal packages and compresses the recorded data and transmits it back to the central processing unit via a wireless communication base station. Upon receiving the feedback data containing the inspection conclusion, salt tightness retest value, and flushing completion indicator, the processor activates the status reset and closed-loop update operation module. Based on the transmitted operation code, the system calls the corresponding manual treatment reduction coefficient, performs blocking and attenuation operations on the historical accumulated amount of the device in the memory, and updates the value of the pollution accumulation register again. The update calculation uses the following attenuation model: ; In this formula, This indicates that after the system confirms receipt and completes processing of the on-site handling feedback data, the equipment... The latest cumulative pollution value updated and covered at this current moment; This represents the old value of the accumulated pollution temporarily stored in the system cache at the moment before the system receives the feedback data; This indicates that the flushing completion flag, extracted from the feedback data packet, has been fully executed, and its value is [value]. or Logical variables; This represents the pollution reduction coefficient scalar retrieved from a fixed knowledge base, corresponding to the specific action being performed. This indicates that the measured salt density retest value, based on data returned by on-site personnel, is converted into a fine-tuning correction amount for the residual pollution baseline of the system after inverse calculation.

[0051] Based on the verification of the returned salt density retest values, the processor strictly classifies and controls the subsequent status of the equipment. If the retest result is qualified, the risk level parameter in the corresponding data record structure of the equipment is overwritten, causing it to be removed from the high-risk monitoring list and enter the safe observation state pool. If the retest result is unqualified, it indicates that the routine on-site flushing has failed to remove stubborn salt contamination or that the insulation material has undergone deep irreversible carbonization. The system immediately performs protection locking, maintaining or forcibly increasing its risk level, and triggers the background timed task component to automatically generate suggestions for in-depth secondary maintenance and replacement of the equipment.

[0052] After several sampling cycles and iterative runs of handling events, the processor initiates a background data analysis process. Based on the warning timestamps recorded in each cycle, historical alarm score sequences, and the actual physical results in the handling feedback data, it statistically calculates several key performance indicators to evaluate and update the current warning strategy. The processor extracts data from the historical database in batches and calculates the system alarm lead time. ; In this formula, This represents the average alarm lead time parameter obtained by the system through statistical calculation for all valid early warning events within a specific analysis period; This represents the total number of independent abnormal events recorded during the analysis period that were ultimately verified on-site to have potential faults. This indicates the number recorded in the on-site maintenance log. The physical confirmation timestamp of the coastal power distribution equipment being determined to have reached a critical fault state in this incident; This indicates that the system is targeting the first... The first time an event generates an internal runtime timestamp that includes a comprehensive alarm score at the level of attention or above and issues an early warning notification.

[0053] Next, the processor verifies the false alarm control capability of the early warning algorithm using the actual verification status of the equipment, and generates a system false alarm rate indicator: ; In this formula, This represents the system false alarm rate, a value used to evaluate the probability of an algorithm model generating unnecessary alarms under non-fault conditions. This indicates the total number of false alarm events in which the system issued on-site warnings and handling orders during the analysis period, but the returned inspection conclusions and salt tightness retest values ​​confirmed that the equipment was in a completely normal state. This indicates the total number of warning actions of all levels issued by the system based on the comprehensive alarm scoring sequence within this time period.

[0054] Meanwhile, the processor monitors the risk coverage capability of the algorithm model and calculates the system's false alarm rate based on the external power grid outage event database and internal logs. ; In this formula, This represents the false negative rate assessment value indicating that the characterization system failed to identify risks in a timely manner; This represents the total number of missed events during the period when the system did not issue any level of warning, but the corresponding coastal power distribution equipment actually experienced physical faults such as surface flashover or insulation breakdown. This represents the absolute total number of all faults of this type of power distribution equipment recorded during the specified analysis period.

[0055] Taking into account both false alarm control and risk coverage, the processor achieves a high hit rate in the overall logic computing system based on precision and recall metrics. ; In this formula, The hit rate parameter represents the accuracy of the system's early warning commands. The higher the calculated value, the more accurate the system's algorithm criteria are. This indicates the number of valid early warning events confirmed by inspections and retests in the feedback data after the system issues an early warning command, indicating that the target equipment does indeed have partial discharge, insulation degradation, or severe salt deposits. This is a constant representing the total number of all early warning documents issued by the system.

[0056] In addition to the above assessment of the early warning model's triggering capability, the processor also evaluates the effectiveness of on-site intervention measures in improving the overall security situation of the system based on risk tracking sequence data, and calculates the risk reduction rate index after the intervention: ; In this formula, This represents the quantitative value of the average post-treatment risk reduction rate resulting from all completed physical treatment actions statistically analyzed by the system. This indicates the cumulative total number of batches of all on-site handling actions that were fully implemented within the assessment period; Indicates the first Before the execution of the next on-site action, the system calculates and records the highest comprehensive alarm score for the target device. Indicates the first After the on-site action is completed and the retest data is returned as qualified, the system recalculates and updates the steady-state comprehensive alarm score.

[0057] Finally, the system scans the scheduling queue and work order receipt state machine to statistically measure the completion rate of maintenance scheduling instructions within the specified time limit, i.e., the closed-loop completion rate metric. ; In this formula, A numerical value representing the system closed-loop completion rate, which reflects the efficiency of the execution process. This indicates the total number of work orders that have been generated in the entire on-site handling list generated by the system, including not only dispatch instructions but also complete handling feedback data containing inspection, retesting, and maintenance information, which have been successfully received and whose target risk level has been downgraded back to safe or observation status. This indicates the total number of on-site handling lists successfully pushed to front-line operation terminals by the system based on priority calculation logic within this statistical period. The system stores the indicators generated by the feedback loop into a historical database, completing the full cycle of risk identification, algorithm control, and operation and maintenance handling for coastal power distribution equipment in this iterative cycle. Based on a comprehensive comparison of various statistical indicators, the hierarchical early warning closed-loop handling method proposed in this invention demonstrates superior comprehensive performance in multiple dimensions such as alarm lead time and hit accuracy. Figure 3 As shown in the diagram. Further comparative experimental data demonstrate that, compared to traditional fixed threshold and linear weighted early warning methods, the technical solution based on the attention mechanism-driven deep temporal fusion network of this invention significantly improves the prediction accuracy and risk recall rate of severe salt spray pollution events under complex conditions involving high-frequency fluctuations in wind power coupled with extreme high humidity. This system can identify the risk of sudden partial discharge caused by the nonlinear coupling of multiple meteorological and electrical factors earlier, significantly reducing the total number of false alarm events and ensuring the efficient and accurate allocation of operation and maintenance resources at the dispatching end.

[0058] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for risk classification, early warning, and operation and maintenance management of salt spray pollution for coastal power distribution equipment, characterized in that, include: Acquire environmental monitoring data, equipment operation data, and response feedback data for coastal power distribution equipment; Construct a scour retention factor that includes rainfall scour coefficient and human intervention reduction coefficient; Based on the environmental monitoring data, the environmental exposure intensity is determined. The scouring and retention factor is then used for discrete recursive updates to calculate the pollution accumulation amount, which characterizes the continuous deposition and reduction state of salt spray. The pollution accumulation amount is then mapped to a salt spray pollution risk index. Extract the wind power output fluctuation characteristics and equivalent active power feedback status from the equipment operation data to construct auxiliary indicators for wind power operation disturbances; The salt spray pollution risk index is combined with the wind power operation disturbance auxiliary index to calculate a comprehensive alarm score; The risk level of the coastal power distribution equipment is determined by comparing the comprehensive alarm score using a preset hysteresis judgment rule that includes rising and falling thresholds. When the risk level triggers a severe warning and the voltage deviation of the associated nodes and the line load rate exceed the set operating limits, the coastal power distribution equipment is controlled to connect to a reactive power support device or a mobile temporary support device to perform physical support adjustments.

2. The method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment according to claim 1, characterized in that, After acquiring the environmental monitoring data and the equipment operation data, the process also includes: Sliding window smoothing is applied to data with missing samples; When the missing proportion is lower than the first set threshold, the data is completed using the effective mean. When the consecutive missing proportion is higher than the first set threshold, the missing flag is recorded and the data reliability is calculated. When the data credibility is lower than the second set threshold, a verification flag is output to intercept the data.

3. The method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment according to claim 1, characterized in that, Determining the environmental exposure intensity specifically includes: The exposure factor of the equipment location is calculated based on the distance from the sea, windward conditions, and shielding conditions of the area where the coastal power distribution equipment is located. After normalizing the wind speed, humidity, temperature, and salt spray concentration in the environmental monitoring data, the environmental exposure intensity is obtained by weighted summation based on the equipment location exposure factor.

4. The method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment according to claim 1, characterized in that, The specific method for performing discrete recursive updates in conjunction with the scour retention factor is as follows: The pollution accumulation of the previous period is multiplied by a preset pollution retention coefficient according to the sampling period, and the product of the environmental exposure intensity of the current period and the scour retention factor is added to achieve dynamic tracking and calculation of rainfall reduction, artificial reduction and salt spray accumulation.

5. A method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment according to claim 1, characterized in that, In the step of calculating the comprehensive alarm score, a voltage deviation auxiliary index calculated based on the voltage deviation of the device-associated node and a line load auxiliary index calculated based on the line load rate are also incorporated.

6. A method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment according to claim 1, characterized in that, The method of determining the risk level by comparing the comprehensive alarm score with a preset hysteresis judgment rule that includes rising and falling thresholds specifically includes: To avoid frequent changes in risk level, the rising threshold is configured to be greater than the falling threshold to form a hysteresis interval; The risk level will be raised by one level only if the comprehensive alarm score is higher than the rising threshold for a consecutive preset number of sampling periods; The risk level will be downgraded by one level only if the comprehensive alarm score is lower than the decrease threshold for a consecutive preset number of sampling periods.

7. The method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment according to claim 1, characterized in that, Based on the aforementioned risk level, the following dispatching operations are also included: The priority score is calculated by combining the risk level, equipment importance, road accessibility, and estimated arrival time, and an on-site disposal list including on-site verification or pollution cleanup actions is generated. When the available resources in the current cycle are insufficient to cover the on-site disposal list, actions that cannot be performed at present are screened out based on resource constraints, and coastal power distribution equipment with a severe risk level located on the main line or important user side is forcibly prioritized for disposal.

8. A method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment according to claim 7, characterized in that, When determining the pollution cleanup action, the expected risk reduction, execution cost, and risk level are matched, and the final disposal action is selected based on the principle of maximizing recommended utility.

9. A method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment according to claim 7, characterized in that, For coastal power distribution equipment that has reached a critical state, the on-site handling list prohibits the issuance of individual inspection actions. Instead, a continuous process instruction of flushing, retesting, and inspection and repair is forcibly generated and issued.

10. A method for risk classification, early warning, and operation and maintenance of salt spray pollution for coastal power distribution equipment according to claim 1, characterized in that, After receiving the feedback data from the field, the following closed-loop update operation is performed: Based on the inspection results, salt density retest value, and flushing completion sign, the corresponding manual treatment reduction coefficient is used to update the pollution accumulation again; If the retest results are satisfactory, the risk level of the control equipment will be downgraded and the system will enter a safe observation state. If the retest result is unsatisfactory, the risk level will be locked and upgraded, and a secondary treatment suggestion will be automatically generated. Based on the early warning data and the response feedback data for each period, the hit rate and alarm lead time of the current early warning strategy are evaluated and updated.

Images