Real-Time Salt Contamination Monitoring System and Method for Transmission Line Insulator Based on Artificial Intelligence

Abstract

Insulators on overhead power lines have long been exposed to the outdoors and are susceptible to pollution and salt contamination. Due to factors such as wind and gravity, pollution in the atmosphere gradually deposits on the surface of the insulator. In humid and windy conditions, conductive pollutants begin to dissolve in the water on the surface of the insulator, increasing the leakage current and affecting insulation performance. This study mainly uses a data acquisition system to measure the leakage current of the insulator and weather parameters (including temperature, relative humidity, pressure, wind speed, and ultraviolet) around the insulator. Artificial intelligence is then applied to establish a prediction model for leakage current based on weather parameters. The established model accurately predicts insulator leakage current through weather parameters. In order to observe the real-time status of the insulator, this study establishes a monitoring platform that integrates the predicted leakage current with weather parameters. It allows users or maintenance personnel to connect to the server through the network to observe the predicted results and weather parameters. The results can establish a real-time salt contamination monitoring system for insulators on transmission lines, enabling operation and maintenance personnel to understand the actual insulation situation of the insulator in real-time. This can not only prevent power outages due to salt contamination or pollution but also reduce the workload for maintenance personnel. Moreover, the maintenance strategy is upgraded from time-base maintenance to condition-base maintenance, significantly improving the efficiency of operation and maintenance for power lines.

1. Introduction

High-voltage insulators on overhead power lines have long been exposed to the outdoors and are susceptible to pollution and salt contamination. Salt contamination and industrial pollution may cause flashovers and line outages, thereby affecting the power supply and reducing the reliability of the power system. To prevent insulator flashovers, maintenance staff need to periodically wash the insulators [1,2].

Currently, various maintenance procedures are available to minimize flashovers caused by contamination and industrial pollution. Among the available maintenance procedures, periodic washing is the most common method to remove pollutants from the surface of insulators [3]. Furthermore, the equivalent salt deposit density (ESDD) and non-soluble deposit density (NSDD) are regularly used to assess the severity of site pollution [4].

In terms of maintenance, utilities often adopt periodic maintenance procedures when abnormal conditions are observed on the insulators. Although this method can effectively prevent flashovers due to contamination and industrial pollution, it requires a significant amount of manpower and cannot accurately ensure the insulation of the insulators [5,6]. Therefore, one of the challenges for maintenance is how to assess the condition of insulators and determine the washing schedule. Considering the variability of weather and pollution conditions, periodic washing may not be an efficient strategy.

Generally speaking, the progress of insulator discharge increases with the leakage current, which is observable when the surface of the insulator is contaminated and wet. Therefore, by observing the surface discharge and leakage current of the insulator, one can roughly understand its insulation and whether it has reached all the stages of the pollution flashover mechanism. In addition, it indicates how close the insulator string is to flashover. However, the leakage current is influenced by weather conditions such as temperature, relative humidity, pressure, wind speed, ultraviolet exposure, and the type and layers of pollution on the surface of the insulator [7,8].

The relationship between discharge and leakage current can be determined through experiments.

Table 1. Relationship between leakage current and surface discharge phenomena of the insulator.

Leakage Current Range	Surface Discharge Phenomena
10 μA~100 μA	The surface is dry with a light corona
100 μA~1 mA	The surface is dry with a light corona
1 mA~10 mA	The surface is wet with a light spark
10 mA~100 mA	Spark and arc occur in a partial string
100 mA~1 A	Extended spark and arc occur
1 A~10 A	Flashovers occur
10 A above	Fault and outages occur

illustrates the correlation between leakage current and surface discharge phenomena in the ceramic insulator. The experiment reveals that the leakage current is extremely low when the surface is clean and dry; 1 mA is the normal leakage current in the clean and dry state. When the relative humidity reaches a certain level, for example, 80% or more, the leakage current exceeds 1 mA, indicating that more obvious corona or sparks can be observed [9,10,11,12].

Therefore, both relative humidity and pollution conditions must combine to form a larger corona or spark. This leakage current typically initiates at 2 to 3 mA. When the leakage current exceeds 10 mA, due to the heat generated at the discharge roots, the pollution dries out in their vicinity, and the dry band is formed. Subsequently, a spark occurs, causing discharges along small portions of the insulation. If the leakage current exceeds 100 mA, an extended partial arc occurs. This arc discharge is highly unstable. Therefore, if an extended partial arc discharge, due to an extremely large leakage current, is observed, the insulator should be cleaned immediately.

As mentioned above, the leakage current is a more meaningful parameter as it provides information on all stages of the pollution flashover mechanism and indicates how close the insulator string is to flashover.

From a literature survey, it is evident that insulator contamination is related to weather parameters such as temperature, relative humidity, pressure, wind speed, and ultraviolet [13,14,15,16]. Since the leakage current on the surface of the insulator is affected by the material, surface contamination, and surrounding environment, as well as its nonlinear characteristics, artificial intelligence algorithms such as machine learning technology have been applied to analyze the leakage current or estimate contamination on the surface of insulators in some related studies [10,11,13,17,18,19,20,21,22]. For instance, artificial neural networks (ANNs) have been used to build a leakage current model [18,19], support vector machines (SVMs) to evaluate contamination degree [20,21], and random forests to predict equivalent salt deposit density (ESDD) based on parameters such as pollution and weather [22]. Although there have been numerous studies using machine learning and other algorithms for the leakage current of insulators, there is limited study on modeling analysis using long-term measurement data to build predictive models and evaluate the most effective prediction methods.

The study was conducted at a 161 kV test station located in a severely polluted industrial area on the western coast of Taiwan. In the test station, a data acquisition system has been built to measure the leakage current of insulators and atmospheric parameters (including temperature, relative humidity, pressure, wind speed, and ultraviolet) around the insulator. One insulator string was monitored under real operational conditions for 30 consecutive months.

This paper proposes a novel method to predict the leakage current using artificial intelligence algorithms and establishes a real-time salt contamination monitoring system. Firstly, this study takes silicon-grease-coated insulators into account and installs them in the test station. The leakage current and weather parameters are collected for an extended period (30 months). Then, artificial intelligence and machine learning techniques (such as support vector regression, gradient boosting regression, and long short-term memory neural networks) are applied to establish a prediction model for the leakage current of the insulators. The established model can accurately predict the leakage current of insulators based on weather parameters. Subsequently, the most effective prediction method is evaluated in terms of Mean Squared Error (MSE), Mean Absolute Error (MAE), and Explained Variance Score (EVS).

Additionally, to observe the real-time status of the insulator, this study establishes a monitoring platform that integrates the predicted leakage current, pollution level of the insulator, and weather parameters. It allows users or maintenance personnel to connect to the server through the network to observe the predicted results and weather parameters. The results can establish a real-time salt contamination monitoring system for insulators on transmission lines, enabling operation and maintenance personnel to realize the actual insulation situation of insulators in real time. This system not only helps prevent power outages due to salt contamination or pollution but also reduces the workload for maintenance personnel.

2. Field Monitoring Program and Data Collection

2.1. Test Station

The study was conducted at a 161 kV test station located in a severely polluted industrial area in Taiwan, very close to the Taiwan Strait. The test station is situated 1.60 km from the Taiwan Strait and 10 km from a coal-fired power plant, separately. The location of the test station is shown in Figure 1.

Therefore, in this challenging environment, the insulators tested at the test station suffer from heavy pollution and salt contamination. Especially in winter, strong occasional monsoons bring plenty of salt contamination and industrial pollution, making it suitable for studying the pollution performance of the insulators.

The study utilized a string of ceramic insulators coated with silicone grease because it can provide longer creepage and good anti-fog performance in the test station. Throughout the entire test period, the insulators were energized at 161 kV, and the leakage current of the string of testing insulators was recorded through a measurement device. The specifications of the leakage current measuring device are shown in

Table 2. Specifications of leakage current measuring devices.

Leakage Current Measuring Device: Measure Leakage Current from the Insulator
A. Range	20 μA to 100.0 mA.
B. Accuracy	≤1.5% ± 1 dig.
C. Communication	RS-232 or RS-485 or IC2 or TCP/IP, etc., analog to digital signal to prevent electromagnetic interference.
D. Protection	Metal shell and surge protection device.

. The test station is depicted in Figure 2.

In addition to establishing a leakage current measurement device at the test station, a weather collection system was built to record weather parameters, including temperature, humidity, wind direction, wind speed, and ultraviolet radiation. In this study, measurement data for a period of thirty consecutive months were recorded from June 2016 to December 2018. These data encompass not only different seasons and natural washing by rain but also various special conditions such as typhoons, salt contamination, and industrial pollution. Figure 3 illustrates the weather data collection system.

2.2. ESDD and NSDD Measurements

In this study, the pollution severity at the test station and the pollution performance of testing insulators were generally determined by measuring the equivalent salt deposit density (ESDD) and non-soluble material deposit density (NSDD) from a pilot insulator every month from February 2018 to March 2019. As shown in Figure 4, ESDD and NSDD increased significantly from September to March due to strong, occasional monsoons bringing plenty of salt contamination in the winter. During the monitoring period, the maximum ESDD was around 0.3 mg/cm². Figure 5 shows the high constituents of sea salt, including Cl⁻ (32%), SO₄²⁻ (24%), Na⁺; (19%), NO_3⁻ (10%), and Ca²⁺ (9%), with the rest being less than 4%. Figure 6 displays the ratio of the proportion of insoluble non-soluble components, with SiO₂ at 65%, Fe₂O₃ at 15%, Al₂O₃ at 14%, and the remaining components being less than 3%. Namely, this indicates that the main components, such as sand-like SiO₂ and high constituents of Al₂O₃, representing pollution in this area, include sea salt, sand, dust, and industrial pollution. The results correspond to salt contamination and industrial pollution, aligning with the location of the test station.

2.3. Leakage Current Measurement

The leakage current of the testing insulator string in the test station was measured and recorded from June 2016 to December 2018. According to the measurement results, the leakage current of the testing insulator string was around 0.1 mA~0.3 mA in the first 6 months (from June 2016 to December 2016) because the surface of the insulator was clean enough. However, the leakage current gradually increased in November and showed a significant positive correlation with humidity. Figure 7 illustrates the relationship between the leakage current and humidity.

Throughout the monitoring period, the long-term trend in leakage current reveals a significant and sharp variation from October to March of the following year. This phenomenon is attributed to occasional monsoons that introduce heavy salt contamination and industrial pollution. Moreover, the test station experienced low rainfall during this period, leading to the gradual accumulation of contamination and pollution on the insulator’s surface, resulting in an increase in leakage current. After March, the trend in leakage current stabilizes, indicating a substantial reduction when rain occurs due to the prevention of rain washing on the insulator’s surface. The primary reason for this stabilization is the low amount of salt contamination.

Examining the results of the leakage current, the maximum peak value was approximately 0.25 mA in 2016 (Figure 7); however, in 2017, the maximum peak value reached about 1.25 mA (Figure 8), a significant increase from 0.25 mA, demonstrating a gradual upward trend. Figure 7 and Figure 8 illustrate the variation in leakage current for the years 2016 and 2017.

2.4. Weather Data Measurement

Previous research has shown that the interaction between weather conditions and the performance of insulators is extremely dynamic and complex. In particular, strong winds can lead to the transportation and deposition of pollutants on the insulators, while high humidity can cause the formation of a conductive electrolytic layer on the insulator surface [23]. Both wind and humidity are significant factors contributing to the increase in leakage current. A strong correlation between humidity and leakage current is evident in Figure 7. Therefore, in this study, a set of weather stations was placed around the test station to measure and record local weather data, including temperature, relative humidity, pressure, wind speed, ultraviolet radiation, and other parameters, around the test insulator every minute.

3. Prediction Method of the Leakage Current

Generally, time series analysis involves methods for extracting meaningful statistics and other characteristics from time series data. This encompasses various techniques, such as time series models, exponential smoothing forecasts, or moving averages. This study primarily focuses on the analysis of long-term time series data, including leakage current, temperature, relative humidity, and pressure measured per minute. Regression model algorithms like support vector regression (SVR), gradient boosting regression (GBR), and long short-term memory neural network (LSTM) are applied for predicting the insulator’s leakage current based on on-site measured weather data. Additionally, evaluation metrics (mean squared error (MSE), mean absolute error (MAE), and explained variance score (EVS)) are used to assess the error between the actual and predicted values, providing an explanation of the models’ performance in estimating leakage current.

3.1. Time Series Analysis

Time series analysis is a statistical forecasting method that examines a series of data points indexed in chronological order. The primary objective of time series analysis is to scrutinize the changing trends in data over time and estimate future trends. The advantage of time series analysis is its ability to describe and predict variables as long as there is historical data available. However, its limitation becomes apparent when forecasting variables are influenced not only by time and changes but also by other factors, leading to a notable reduction in forecasting accuracy [24].

3.2. Support Vector Regression (SVR)

Support vector machine (SVM) was introduced by Cortes and Vapnik in 1995 [25]. It is a supervised learning model with associated learning algorithms used for data classification and regression analysis. SVM typically functions as a binary linear classifier, using a line to separate training data into two categories as effectively as possible. It then corresponds to a two-dimensional space for classification prediction. However, the data encountered in big data scenarios are often linear and inseparable. Hence, SVM can utilize various kernel functions to perform non-linear classification and map the data to a high-dimensional space. In cases where linear classification in two-dimensional space is not sufficient, the data are transformed into a high-dimensional space using a kernel function to identify the optimal classification hyperplane. Support vector regression is an extension of support vector machine, designed to handle regression problems.

3.3. Gradient Boosting Regression (GBR)

Gradient boosting is a machine learning technique used for regression, classification, and other tasks. It produces a prediction model in the form of an ensemble of weak prediction models, typically decision trees. Boosting is an algorithm that elevates a weak learner to a strong learner. The process begins by training a base learner from the initial training set and then adjusting the training sample distribution based on the performance of the base learner. Subsequently, the next base learner is trained using the adjusted sample distribution. This cycle repeats until the number of base learners reaches the pre-specified value of N. Finally, the N base learners are weighted and combined.

Gradient boosting, a type of boosting algorithm, differs from traditional boosting in that it does not assign weights to correct and incorrect samples. Instead, it calculates the difference between the predicted result and the sample (residual) and establishes a new learner to reduce the residual [25,26,27,28].

3.4. Long Short-Term Memory Neural Network (LSTM)

Long short-term memory (LSTM) is an artificial recurrent neural network (RNN) architecture commonly employed in the field of deep learning. It is a cyclic neural network well-suited for tasks such as data classification and time data prediction, as depicted in Figure 9. A typical LSTM unit comprises a cell, an input gate, an output gate, and a forget gate. During the transmission of neuron data, previous prediction data can be retained, while unimportant information is discarded to enhance the overall learning effect [29]. Subsequently, through the error back propagation (BP) process, weights are adjusted, and the model undergoes repeated analysis of big data and deep learning to establish prediction models.

3.5. Evaluation Metrics

This study adopts three evaluation metrics to assess prediction performance: mean squared error (MSE), mean absolute error (MAE), explained variance score (EVS). The mean squared error (MSE) is an estimator that measures the average of the squares of the errors. It can be used to evaluate the dispersion between individuals in the data. A smaller MSE value indicates better accuracy in the prediction model [30,31].

$$MSE=\frac{1}{n}\sum _{i=1}^n{(y_i-\widehat{y_i})}^2$$

(1)

The mean absolute error (MAE) is a measure of errors between paired observations expressing the same phenomenon. When the same physical quantity is measured multiple times, each measurement value and its absolute error will not be the same. It takes the absolute error of each measurement and then calculates the average value. The mean absolute error is a non-negative value, and the better the model, the closer the MAE is to zero [11].

$$MAE=\frac{1}{n}\sum _{i=1}^n\left|y_i-\widehat{y_i}\right|$$

(2)

The explained variance score (EVS) measures the dispersion of errors in a given dataset. It is calculated as Formula (3), where $Var(y_i-\widehat{y_i})$ is the variance of prediction errors and actual values, respectively. A result closer to 1 indicates that the independent variable can better explain the dependent variable, while a smaller value suggests a less effective model [32].

$$EVS=1-\frac{Var(y_i-\widehat{y_i})}{Var\left(y_i\right)}$$

(3)

4. Prediction Model of the Leakage Current and Example Analysis

4.1. Prediction Model

The purpose of this study is to predict the leakage current of insulators based on weather and environmental data using artificial intelligence methods. Due to the complex relationship between leakage current and weather conditions, artificial intelligence methods open up new possibilities for addressing this issue. These algorithms learn from the data and continuously improve their performance. In order to predict the leakage current of the test insulator string at the test station, this study employs time series data modeling. The initial step involves using leakage current and weather data (temperature, relative humidity, wind speed, ultraviolet, and air pressure) from June 2016 to December 2018 (a total of 30 months). The data are then filtered to remove any anomalies during the measurement process. Subsequently, the processed data are utilized for modeling with SVR, GBR, and LSTM methods. Once the models are established, real-time measured weather parameters, including temperature, relative humidity, wind speed, and ultraviolet, can be employed to estimate the leakage current of the insulator. The proposed model is validated using collected measurement data from 2019 at the test station. The output results are evaluated using MSE, MAE, and EVS. Finally, this paper introduces a cloud monitoring platform based on estimated leakage current and weather parameters to assess the pollution level of the insulator. The pollution levels are categorized as Normal, Warning, and Danger, distinguished by colors. The flowchart of the analysis and algorithm modeling is presented in Figure 10.

4.2. Verification Analysis and Result

Many factors, such as humidity, wind speed, temperature, and atmospheric pressure, influence the leakage current on the surface of the insulator. Changes in weather parameters can result in uncertain and irregular behavior of the leakage current. Therefore, to produce more accurate predictions, it is crucial to employ advanced and suitable models.

In order to verify and analyze the predicted results of leakage current, this study applied measurement data collected in 2019 from the test station for the simulation. The input data, which can be divided into three case studies based on different time segments, are as follows:

• Case study 1: 1 day (1 February 2019)
• Case study 2: 5 days (1 March 2019~5 March 2019)
• Case study 3: 6 days (16 June 2019~22 June 2019)

This study proposes three prediction models (SVR, GBR, and LSTM). The input data for each case study include temperature, humidity, wind speed, air pressure, and ultraviolet per minute during the time period. The output data are predicted leakage currents, which are verified by evaluation metrics (MSE, MAE, and EVS). From verification and analysis, the LSTM algorithm has the best performance.

4.2.1. Case Study 1

Case study 1 utilized the data from the entire day of 1 February 2019 as input. Figure 11, Figure 12 and Figure 13 show the predicted leakage current through the LSTM, GBR, and SVR models and the actual measured leakage current. The blue line is the predicted result of leakage current, and the red line is the actual measured leakage current.

Table 3. Case 1: Prediction errors of different algorithms.

Model	MAE	MSE	EVS
LSTM	0.0780	0.0065	0.8349
GRB	0.0889	0.0088	0.6534
SVR	0.1014	0.0112	0.6322

shows the prediction errors of different algorithms.

The proposed LSTM demonstrates better predictive performance for case study 1, with a mean absolute error (MAE) of 0.0780, a mean squared error (MSE) of 0.0065, and an explained variance score (EVS) of 0.8349, as compared to SVR and GBR. The analysis results show that when using mean squared error (MSE) and mean absolute error (MAE) to evaluate the performance of LSTM, the model’s prediction errors are very small. An EVS of 0.8349 is considered a good result, indicating that the model can explain a significant portion of the variance in the target variable, suggesting that the model’s predictions are quite accurate.

Overall, based on the results of MSE, MAE, and EVS, it can be concluded that the LSTM model performs well in predictions, with very small errors between predicted and actual values, and it exhibits good explanatory capabilities.

4.2.2. Case Study 2

The case study 2 applied a 5-day period from 1 March 2019 to 5 March 2019 as input data. Figure 14, Figure 15 and Figure 16 display the predicted leakage current through the LSTM, GBR, and SVR models alongside the actual measured leakage current. The blue line represents the predicted leakage current, while the red line represents the actual measured leakage current.

Table 4. Case 2: Prediction errors of different algorithms.

Model	MAE	MSE	EVS
LSTM	0.0857	0.0276	0.2116
GRB	0.1346	0.0306	0.4144
SVR	0.2913	0.1177	−2.0080

presents the prediction errors of different algorithms.

In case study 2, this study examines the prediction of leakage current for a period of 5 days. The results indicate that the proposed LSTM exhibits a lower MAE of 0.0857 and a MSE of 0.0276, while GBR shows a higher EVS of 0.4144. The LSTM demonstration reveals that MSE and MAE approaching zero indicate small errors in predictions. However, the EVS result is 0.2116, which is relatively low. This indicates that the model has a limited ability to explain the variance in the target variable, suggesting that there might be some variability that the model cannot capture. While the results of MSE and MAE suggest good performance in prediction, it is essential to enhance the LSTM model’s explanatory power for the target variable to further improve overall performance.

4.2.3. Case Study 3

Additionally, to assess the adaptability and flexibility of the proposed model, we examined the leakage current prediction for a 6-day period 16 June 2019 to 22 June 2019 in case study 3. Figure 17, Figure 18 and Figure 19 show the predicted leakage current through the LSTM, GBR, and SVR models compared to the actual measured leakage current. The blue line represents the predicted leakage current, and the red line represents the actual leakage current.

Table 5. Case 3: Prediction errors of different algorithms.

Model	MAE	MSE	EVS
LSTM	0.0897	0.011	0.3025
GRB	0.2137	0.0802	−0.446
SVR	0.7113	0.5482	−2.166

displays the prediction errors of different algorithms.

This study examines the prediction of leakage current for a period of 6 days. The results indicate that the proposed LSTM exhibits a lower MAE of 0.0897, a MSE of 0.011, and a relatively higher EVS of 0.3025. Similarly, it demonstrates that LSTM yields lower MSE and MAE values while evaluating performance, and the LSTM model exhibits very small prediction errors. However, with the relatively low EVS, there is room for improvement in explaining the variance of the target variable.

4.2.4. Performance Comparison

This study divides the accumulated data from the test station into training and validation sets to evaluate the performance of various artificial intelligence methods. The LSTM model achieved a precision of a MAE of 0.0780 and a MSE of 0.0065 for validating data in case study 1; a MAE of 0.0857 and a MSE of 0.0276 for case study 2; and a MAE of 0.0897 and a MSE of 0.011 for case study 3. However, both SVR and GBR could not achieve the optimum evaluation metrics compared with LSTM, indicating the limitations of SVR and GBR in predicting leakage current. Compared to SVM and GBR, the LSTM model exhibits superior overall performance. LSTM incorporates gating mechanisms that control the flow of information in memory units, enabling it to better capture and retain long-term patterns in leakage current and weather parameters, significantly enhancing performance in processing time series data. In contrast to the prediction methods SVM and GBR, the proposed LSTM model is computationally more efficient and provides more accurate predictions. Throughout the study, LSTM attains the highest accuracy in learning operations, as evident in the case studies, where the evaluation metrics reveal lower MSE and MAE values while evaluating performance. The LSTM model exhibits very small prediction errors. Despite progress through validation analysis, some improvements are still needed. The infrequent occurrence of peak leakage current values results in the underprediction of these peaks. The LSTM model, along with other models, cannot achieve high accuracy in the validation data, mainly due to imbalanced factors. The accuracy of multiclass problems is influenced by the dominant class in the imbalanced dataset. While the results of MSE and MAE suggest good predictive performance, the EVS, however, is relatively low. This indicates that the model has limited ability to explain the variance in the target variable, suggesting that there might be some variability that the model cannot capture. However, LSTM demonstrates higher capability and precision in handling the imbalanced frequency of leakage current data. Hence, continuous data collection over an extended period may be required to improve the predictive model.

5. Establishment of a Real-Time Salt Contamination Monitoring System

5.1. System Architecture

Based on previous verification and analysis, the LSTM algorithm exhibits the best performance. Therefore, to establish a real-time salt contamination monitoring system for insulators on the transmission line, this study utilizes the LSTM algorithm to create a cloud monitoring platform for leakage currents in insulators. The platform architecture allows the saving of weather parameters to the database. After storage, the real-time leakage current can be estimated. Subsequently, the real-time weather and predicted leakage current are transmitted to the monitoring webpage. This provides maintenance personnel with the ability to observe and monitor estimated leakage current results through computers or mobile phones. In this study, an economical and effective maintenance strategy can be formulated based on the predicted leakage current, leading to significant savings in maintenance and labor costs.

Furthermore, to assess the salt contamination or pollution level of the insulator, this study examines the relationship between leakage current and discharge phenomena when the insulator surface is contaminated. As depicted in Figure 1, when the leakage current exceeds 1 mA, the insulator’s surface initiates a small spark discharge. Therefore, the proposed monitoring system’s threshold values can be set at 0.5 mA and 1 mA. Additionally, in accordance with the estimated leakage current, this study utilizes a 30-min interval. The system calculates the number of leakage current pulses surpassing 0.5 mA and 1 mA within this time frame. If there are more than 10 instances of leakage current pulses exceeding 0.5 mA in a 30-min interval, it indicates a gradual deterioration in the insulator’s insulation condition. Although there is no immediate risk of flashover, from a maintenance standpoint, the pollution level of the insulator will transition from Normal to Warning.

When there are more than five instances of leakage current pulse greater than 1 mA within 30 min, it indicates that the insulator has poor insulation and is beginning to generate small spark discharge. If the current pollution or environmental conditions are more severe, a flashover is very likely to occur. The pollution level of the insulator will then change to Danger. At this point, washing of the insulators or other maintenance is recommended.

Table 6. The predicted leakage currents (LCMs) and pollution levels of the insulator.

Output	Pollution Level	Section (30 min)	LCM (mA)	Pulse Number
Strategy	Normal	30	<0.5	-
	Warning	30	>0.5	>10
	Danger	30	>1	>5

shows the predicted leakage current and pollution level of the insulator.

5.2. Monitoring System Webpage Planning

This study proposes a monitoring system platform to assess the salt contamination or pollution level of insulators. The platform displays various parameters, including insulator leakage current and environmental parameters. This information assists maintenance personnel in gaining insights into the insulation condition of insulators, facilitating more efficient maintenance procedures.

Figure 20 illustrates the planning of the monitoring system platform. The monitoring system is presented in the form of a visual webpage, with four main sections: (1) real-time weather parameters around the insulator, including pressure, temperature, relative humidity, wind speed, and ultraviolet (as shown in Figure 20); (2) the location of the insulator, including the tower’s location, voltage level, and other relevant information; (3) insulator pollution levels, categorized as Normal, Warning, and Danger, and distinguished by different colors; and (4) the number of leakage currents above 0.5 mA and 1 mA (as shown in Figure 21).

To validate the proposed monitoring system, this study employs the measured real-time weather parameters from 10:00 to 11:02 on 3 March 2019, as examples. The real-time weather parameters are shown in Figure 20.

Subsequently, the prediction of leakage current and pollution level can be analyzed through the monitoring system.

The leakage current is assessed using LSTM, calculating the number of pulses surpassing 0.5 mA and 1 mA within a 30-min interval to analyze the pollution level. If there are more than 10 instances of leakage current pulses exceeding 0.5 mA within this period, it indicates a gradual deterioration in the insulator’s insulation condition, causing the pollution level to transition from Normal to Warning. When there are more than five instances of leakage current pulses exceeding 1 mA within a 30-min interval, the pollution level of the insulator will change to Danger.

Figure 21, Figure 22 and Figure 23 illustrate the results of the prediction of leakage current and pollution level. It is evident that the leakage current gradually increases from 10:00 to 10:29, and the pollution level remains Normal during this period (as shown in Figure 21).

As the leakage current continued to rise, reaching 0.5 mA at 10:34, the situation escalated further. By 10:49, the instances surpassing the 0.5 mA threshold had totaled 15 within a 30-min interval, signaling a gradual deterioration in the insulator’s insulation condition. Consequently, the pollution level transitioned from Normal to Warning (as shown in Figure 22).

At 11:02, the leakage current persisted without a decrease, surpassing the 0.5 mA threshold 28 times within a 30-min interval, with 13 instances exceeding the 1 mA threshold. This indicates a severe pollution or environmental condition, increasing the likelihood of a flashover in the insulator string. As a result, the pollution level transitions from Warning to Danger (as shown in Figure 23).

6. Conclusions

This paper presents a novel method for predicting insulator leakage current on transmission lines using artificial intelligence and proposes an early warning strategy to establish a real-time salt contamination monitoring system, providing insights for pollution prevention for transmission lines. The study is based on data collected over thirty consecutive months from a severely polluted outdoor test station in Taiwan. The main findings are concluded below:

• The leakage current of insulators is closely linked to environmental parameters, serving as an indicator of insulator pollution levels.
• Artificial intelligence algorithms can be applied to estimate leakage currents because they are monitored in real-time consecutively.
• The study proposes three prediction models (SVR, GBR, and LSTM) to predict the leakage current of insulators. In this study, LSTM shows effectiveness in predictive tasks involving sequential data due to its ability to capture long-range dependencies and mitigate the vanishing gradient problem. Verification analysis and results reveal that LSTM attains the highest accuracy in learning operations, as evidenced by the case studies where evaluation metrics indicate lower MSE and MAE values, suggesting small errors in predictions.
• The data used in the prediction model are from the measured data of the test station in actual operation. The results show that the proposed method is feasible.
• In this study, despite the progress achieved through verification analysis, some enhancements are still necessary. From the results, the frequency of occurrence of leakage current peaks is relatively rare, resulting in minority peaks being seldom predicted. Thus, it might need to continuously collect data for long periods to improve the prediction model.
• The proposed monitoring system offers real-time information for operational and maintenance personnel, facilitating cost-effective maintenance planning. It can prevent power outages due to salt contamination or pollution and reduce the workload for maintenance personnel. Moreover, upgrading the maintenance strategy from condition-based maintenance to time-based maintenance may greatly improve the efficiency of operation and maintenance for power lines.
• The proposed novel method has the potential for application in critical transmission lines or towers with severe salt contamination, utilizing low-cost weather sensors to construct a real-time salt contamination monitoring system in the power grid.