Intelligent insulator monitoring and fault detection system for high-voltage power grid

Abstract

The invention relates to the technical field of power systems, and discloses an intelligent insulator monitoring and fault detection system for a high-voltage power grid, which comprises the following steps: S1, a multi-source sensor measurement module performs multi-source information acquisition and breaks through the limitation of a single parameter; s2, a multi-source information fusion processing module performs multi-source data fusion processing and constructs global state evaluation; according to the method, innovation and breakthrough are achieved through cooperative collection and data fusion processing of the multi-source sensor, multi-dimensional data such as the voltage, the electric field, vibration and the environment are obtained in real time, the filtering, fusion and global state evaluation technology is combined, and the real-time analysis of the multi-dimensional data is achieved. The data reliability and the sensitive identification capability for tiny faults and sudden fluctuations are effectively improved, and the spanning from single parameter judgment to multi-dimensional data collaborative decision is realized.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power systems, in particular to a high-voltage power grid intelligent insulator monitoring and fault detection system. BACKGROUND

[0002] The existing high-voltage power grid insulator monitoring system generally relies on a single voltage parameter or an experience threshold to make an abnormality judgment, and it is difficult to accurately reflect the real operating state under a complex power grid environment. Since there is a lack of a multi-dimensional data analysis mechanism, the system has weak recognition ability for local subtle faults, voltage nonlinear changes and sudden short-time fluctuations, and false negatives or false positives often occur. In addition, the fixed threshold method is difficult to adapt to different regions and complex electric field distribution, resulting in insufficient fault positioning accuracy and low system robustness, which cannot meet the demand of modern power systems for high-precision monitoring and rapid fault response. Therefore, there is an urgent need for a new intelligent insulator monitoring technology with multi-source information fusion capability and improved fault recognition and positioning accuracy.

[0003] Therefore, the high-voltage power grid intelligent insulator monitoring and fault detection system is provided by the person skilled in the art to solve the problems mentioned above. SUMMARY

[0004] In view of the deficiencies of the prior art, the high-voltage power grid intelligent insulator monitoring and fault detection system is provided to solve the problems mentioned in the background.

[0005] To achieve the above purpose, the following technical solutions are used: the high-voltage power grid intelligent insulator monitoring and fault detection system comprises the following steps:

[0006] S1: Multi-source sensor measurement module, multi-source information acquisition, breaking through the limitation of single parameter;

[0007] S2: Multi-source information fusion processing module, multi-source data fusion processing, and global state evaluation is constructed;

[0008] S3: Self-adaptive nonlinear electric field modeling module, self-adaptive nonlinear electric field model is constructed, and the adaptability to complex electric field is improved;

[0009] S4: Intelligent fault diagnosis and positioning module, intelligent fault diagnosis and accurate positioning, and monitoring accuracy is improved;

[0010] S5: Self-powered and low-power communication module, self-powered and low-power communication design, and system robustness is enhanced;

[0011] S6: Remote monitoring and feedback platform, cloud platform real-time monitoring, and rapid response is realized.

[0012] Preferably, the step 1 comprises:

[0013] The multi-source sensor measurement module comprises a voltage sensor, an electric field intensity sensor, a leakage current sensor, a temperature and humidity sensor, and a vibration sensor, and is used for synchronously collecting multi-dimensional parameters related to the operation state of the insulator, to form a multi-source original data set for subsequent data fusion processing.

[0014] Step 1.1 Multi-dimensional electric-field-humidity coupling index, which is used for reflecting the coupling characteristics among the electric field intensity, the humidity and the leakage current, and can be used for early identification of insulation degradation, and the formula is as follows:

[0015] ,

[0016] : the electric field value measured by the electric field intensity sensor, : the leakage current, : the relative humidity, : the coupling weight coefficient of the electric field, the current and the humidity;

[0017] Step 1.2 Insulator dynamic vibration-current combined characteristic quantity, and the formula is as follows:

[0018] ,

[0019] : the acceleration data point sampled by the vibration sensor, : the average acceleration, : the number of vibration data sampling points, : the leakage current;

[0020] Step 1.3 Multi-source parameter normalized state energy, and the formula is as follows:

[0021] ,

[0022] : the parameter measured by the kth sensor, : the long-term average of the kth parameter, : the standard deviation of the kth parameter, : the weight;

[0023] Step 1.4 Nonlinear voltage-electric field response deviation, and the formula is as follows:

[0024] ,

[0025] : the measured electric field value, : the voltage measured by the voltage sensor, : the voltage-electric field nonlinear coefficient, which is obtained by fitting;

[0026] Step 1.5 Multi-source original data entropy, and the formula is as follows:

[0027] ,

[0028] : probability of the multi-source data after quantization, : number of data bins.

[0029] Preferably, the step 2 comprises:

[0030] a data fusion unit for fusing multi-dimensional data from multiple sensors, using Kalman filtering, particle filtering, Bayesian inference and other data fusion algorithms to convert the raw data collected by each sensor into unified power grid state evaluation information;

[0031] Step 2.1 Kalman filtering algorithm is used to extract and estimate the true state of the system from noise. In power grid monitoring, Kalman filtering can be applied to the fusion of sensor data such as voltage, electric field, humidity, etc. Its formula is:

[0032] ,

[0033] : state estimation value at the current time k, : state estimation value at the previous time k-1, : Kalman gain, representing the trade-off between measurement and prediction, : sensor measurement value at the current time k, : observation matrix, used to map state variables to measurement space; Step 2.2 Particle filtering is suitable for nonlinear systems and can estimate the probability distribution of system state in the fusion of multi-source sensor data. Its formula is:

[0034]

[0035] ,

[0036] : state estimation at the current time k, : i-th particle state in particle filtering, : weight of particle i, adjusted according to the matching degree of measurement, N: number of particles, used to approximate the distribution of system state;

[0037] Step 2.3 Bayesian inference is based on probability theory and estimates posterior probability by updating prior probability. It is used to fuse the results of different sensors. Its formula is:

[0038] ,

[0039] ​​​ State variables The posterior probability, Observational data The conditional probability, State variables The prior probability, Observational data The marginal likelihood.

[0040] Preferably, step 2 further includes:

[0041] The status assessment module is used to build a global status evaluation model of the power grid operation based on the fused data, and output the power grid health index and fault early warning information.

[0042] Step 2.4 After multi-source data fusion, the power grid health index is calculated using a global state evaluation model to provide a comprehensive assessment of the power grid's operating status. The formula is as follows:

[0043] ,

[0044] The power grid health index measures the overall health of the power grid. : No. The parameters measured by each sensor : No. Long-term average of each sensor, : No. The standard deviation of each sensor, : No. The weight of each sensor, Number of sensors;

[0045] Step 2.5 Based on the global state evaluation model, fault early warning information can be issued according to the changing trend of the health index, and the formula is as follows:

[0046] ,

[0047] : Fault warning information; if the value falls below a threshold, an alarm will be generated. A set health threshold is set; a value below this threshold indicates that a power grid failure may occur.

[0048] Step 2.6 uses time series analysis to assess the trend of the health status of multidimensional parameters, assisting in the real-time monitoring and fault prediction of the decision support system. The formula is as follows:

[0049] Trend Analysis = ∑i=1N (wi · d(xi(t)) / dt),

[0050] Trend Analysis: Multidimensional health trend analysis results, d(xi(t)) / dt: the rate of change of the i-th sensor over time, representing the trend of health change, xi(t): the measurement value of the i-th sensor over time, wi: the weight of the i-th sensor.

[0051] Preferably, step 3 includes:

[0052] The electric field modeling unit is used to construct a nonlinear mathematical model of the electric field based on voltage, electric field and other sensor data. The model can automatically adjust its parameters to adapt to different electric field distributions according to changes in the power grid operating environment.

[0053] Step 3.1 To accurately describe the distribution of the electric field in a complex power grid environment, a nonlinear model is usually required, and its formula is as follows:

[0054] ,

[0055] Nonlinear electric field intensity :Voltage, Electric field intensity coefficient The nonlinear exponent of voltage with respect to electric field strength Leakage current, The coefficient representing the effect of leakage current on the electric field. : Nonlinear exponent of leakage current;

[0056] Step 3.2 To improve the adaptability of the electric field model, an adaptive algorithm is adopted, the formula of which is:

[0057] ,

[0058] The electric field strength after adaptive adjustment. : Initial electric field model output, Current moment The measured electric field data, Adaptive step size controls the model's update speed. The previous moment The estimated value of the electric field;

[0059] The model update mechanism is used to update the model online based on real-time monitoring data during power grid operation. It optimizes the electric field modeling results through adaptive algorithms to improve the model's adaptability to complex electric field environments.

[0060] Step 3.3 Formula for updating the electric field model using the adaptive filtering algorithm:

[0061] ,

[0062] Current moment Model parameters, The previous moment Model parameters, Adaptive step size, used to adjust the update amplitude. Current moment Measurement data, Observation matrix;

[0063] Step 3.4 Least squares model optimization is used to improve the prediction accuracy of the electric field model by reducing model error. The formula is as follows:

[0064] ,

[0065] : No. The actual electric field strength measurement values ​​of each data point Model-based and parameters Predicted electric field strength : No. Input features of each data point Total number of data points.

[0066] Preferably, step 3 further includes:

[0067] The fault prediction module predicts and locates electric field anomalies and faults based on the real-time output of the nonlinear electric field model.

[0068] Step 3.5 Based on the nonlinear electric field model, statistical methods are used to detect abnormal fluctuations in the electric field. The formula is as follows:

[0069] Anomaly Score = ,

[0070] Anomaly Score: A higher value indicates a greater degree of electric field anomaly. The measured electric field strength, The electric field strength calculated based on the electric field model. The standard deviation of the electric field strength serves as a measure of electric field fluctuations.

[0071] Step 3.6 uses a prediction algorithm to predict potential faults based on electric field data. The formula is as follows:

[0072] ,

[0073] Fault prediction probability, representing the likelihood of a fault occurring. : No. The weights of each input feature, : No. Each input feature Bias term, used to adjust the model output. Activation function The total number of input features;

[0074] Step 3.7 Combines data from multiple sensors and uses multidimensional features for fault diagnosis, the formula of which is:

[0075] ,

[0076] Fault diagnosis results : No. The weight of each sensor data, : No. A function for processing sensor data.

[0077] Preferably, step 4 includes:

[0078] The intelligent diagnostic unit is used to analyze the operating status of insulators in real time based on the aforementioned multi-source information fusion results and the output of the adaptive nonlinear electric field model, and to identify different types of faults such as flashover, partial discharge, and mechanical loosening by using intelligent diagnostic methods such as pattern recognition algorithms, deep neural networks, or support vector machines.

[0079] Step 4.1 The Support Vector Machine (SVM) finds the optimal hyperplane for classification by maximizing the classification margin. The formula is as follows:

[0080] ,

[0081] Input data The classification results Lagrange multipliers in Support Vector Machines Support vectors' category labels. Kernel functions are used to calculate the similarity between input data and support vectors. Support vectors The bias term determines the position of the classification plane;

[0082] Step 4.2 Deep neural networks learn features through multiple layers of neurons, and the formula is as follows:

[0083] ,

[0084] y. Network output, : No. The weight matrix of the layer, : No. Layer bias terms, Activation function Input data;

[0085] Step 4.3 For different types of faults, a multi-class classification model is used for judgment, and the formula is as follows:

[0086] ,

[0087] Predicted fault type This formula selects the most likely fault type by maximizing the posterior probability of each fault type.

[0088] The fault feature extraction unit is used to extract and classify features from multi-dimensional sensor data such as voltage, electric field, leakage current, and vibration. By constructing multi-dimensional state feature vectors, it improves the accuracy of fault type identification.

[0089] Step 4.4 The formula for extracting frequency domain features using Fourier transform is as follows:

[0090] ,

[0091] Frequency domain representation of the signal Time-domain signal :frequency;

[0092] Step 4.5 Time-Frequency Analysis: Local features of the signal are processed using wavelet transform, as shown in the formula:

[0093] ,

[0094] The signal after wavelet transform. : Scaling factor, controls the resolution of time and frequency. Displacement factor, controlling the time and position of the signal. Basic wavelet functions;

[0095] Step 4.6 Feature selection and dimensionality reduction, the formula is as follows:

[0096] ,

[0097] z. The dimensionality-reduced feature vectors Transformation matrix : Original high-dimensional feature vector.

[0098] The precise positioning unit is used to accurately locate the fault location based on electric field distribution inversion, signal delay analysis, or spatial correlation calculation, and generate spatial fault coordinates corresponding to the insulator structure, so as to realize rapid location and visualization of the fault location.

[0099] Step 4.7 Electric field distribution inversion is used to estimate the fault location based on the measured electric field data. Through the inversion algorithm, the electric field data can be correlated with the fault source location. The formula is as follows:

[0100] ,

[0101] : The measured electric field value Spatial location coordinates Location of the fault source The propagation function describes the propagation of the electric field from the fault source to the measurement point. : Electric field intensity distribution of the fault source;

[0102] Step 4.8 Signal Delay Analysis: The fault location is estimated by analyzing the propagation time of the fault signal between different sensors. The signal propagation time difference is related to the location, and its formula is:

[0103] ,

[0104] : Propagation delay The distance between sensors Signal propagation speed;

[0105] Step 4.9 Spatial correlation analysis is used to calculate the correlation between measurements from different sensors to help locate the fault source. The formula is:

[0106] ,

[0107] :sensor and Spatial correlation between them :sensor and Covariance between :sensor and The standard deviation.

[0108] Preferably, step 5 includes:

[0109] The energy harvesting unit is used to acquire and convert weak energy from the high-voltage power grid operating environment using electromagnetic coupling, photovoltaic energy or piezoelectric energy harvesting technology, so as to provide a continuous and stable power supply for the system.

[0110] Step 5.1 Electromagnetic Energy Harvesting: Changes in the high-voltage power grid are converted into electrical energy through electromagnetic induction. This is achieved using Faraday's law of electromagnetic induction, the formula of which is:

[0111] ,

[0112] : , : , : , : ;

[0113] Step 5.2 The photovoltaic energy harvesting formula is:

[0114] ,

[0115] Photovoltaic output voltage, Open circuit voltage, Load current, : Series resistance of photovoltaic module;

[0116] Step 5.3 The formula for piezoelectric energy harvesting is:

[0117] ,

[0118] V: Output voltage, dQ: Charge change caused by applied pressure, C: Capacitance;

[0119] The energy management unit is used to store, regulate and distribute the collected energy. It achieves dynamic power control through a low-power power management chip, including energy storage, stable voltage output and system adaptive power consumption adjustment, to ensure reliable power supply to the system in extreme environments.

[0120] Step 5.4 Energy Storage and Regulation: For the battery charging process, the battery charging formula can be used:

[0121] ,

[0122] Charging current, Battery capacity, Battery voltage change rate;

[0123] Step 5.5 Voltage stabilization and output regulation uses a switching power supply for control. The relationship between the output voltage and the feedback signal is as follows:

[0124] ,

[0125] Output voltage, Reference voltage, Feedback resistor, : Set the resistor.

[0126] Preferably, step 5 further includes:

[0127] The low-power communication unit adopts a self-organizing network communication protocol to realize low-power, long-distance data transmission between multiple nodes. It also reduces the overall energy consumption of the system and improves the stability and robustness of communication through sleep strategies, data compression and adaptive communication strategies.

[0128] Step 5.6 Low-power communication modules typically use ad hoc networking protocols. The formula for estimating communication power consumption is as follows:

[0129] ,

[0130] Transmission power consumption Power consumption in idle state Power consumption during transmission. The amount of data transmitted. Data transmission rate;

[0131] Step 5.7 The energy consumption formula for hibernation mode is:

[0132] ,

[0133] Power consumption in sleep mode Idle power consumption : Duration of hibernation;

[0134] Step 5.8 The formula for the adaptive communication strategy is:

[0135] ,

[0136] Power consumption of adaptive communication Base power consumption The amount of data to be transmitted. Transmission time : Adaptability coefficient.

[0137] Preferably, step 6 includes:

[0138] The cloud-based data processing and storage unit is used to aggregate, clean, classify, and store multi-dimensional data such as voltage, electric field, leakage current, vibration, and environmental parameters from multiple monitoring terminals in the cloud, and to achieve high-concurrency data processing and intelligent analysis based on cloud computing resources.

[0139] Step 6.1 Data aggregation and cleaning, based on the Z-Score method formula:

[0140] ,

[0141] X: Data sample The mean of the data; : Standard deviation of the data; Z: Z score, used to determine whether the data is outlier;

[0142] Step 6.2 Data Classification and Storage: The formula for the K-means clustering algorithm based on machine learning is as follows:

[0143] ,

[0144] Clustering cost function Data points Clustering Membership degree between them Clustering The center point, Data points Number of data points Number of cluster centers;

[0145] Step 6.3 High-concurrency data processing uses the MapReduce model to process data in parallel. The formula is as follows:

[0146] , ,

[0147] Map: A mapping function that maps input... Convert to intermediate key-value pairs Reduce: A function that reduces intermediate key-value pairs. Aggregate into the final output ;

[0148] The real-time monitoring and early warning unit is used to monitor the operating status of insulators in real time through the cloud platform. It automatically triggers early warning signals based on historical big data models and risk threshold calculations, and pushes abnormal information to maintenance personnel to achieve remote rapid response.

[0149] Step 6.4 Real-time early warning model, based on the cumulative risk model formula:

[0150] ,

[0151] Cumulative risk value, representing the overall risk at the current time t. Real-time data of sensor i at time t. Risk weight of sensor i The number of sensors;

[0152] Step 6.5 The threshold detection and early warning triggering formula is as follows:

[0153] ,

[0154] Warning signal Current moment The monitoring value, : Preset threshold;

[0155] The feedback control and command issuance unit is used to generate corresponding control strategies in the cloud based on the fault information provided by the fault diagnosis and location module, and to issue commands to the field monitoring module through the wireless communication link to realize remote control actions such as sampling frequency adjustment, sensor calibration, node wake-up, and energy-saving mode switching, thus constructing a closed-loop feedback mechanism.

[0156] Step 6.6 Remote control strategy generation, the formula is as follows:

[0157] ,

[0158] u. Output control strategy, : Input fuzzy set;

[0159] Step 6.7 Instruction Issuance and Execution: The formula for calculating instruction transmission delay is as follows:

[0160] ,

[0161] L: Command transmission delay; B: Data size; C: Network bandwidth. Processing time;

[0162] Step 6.8 Remote control action optimization, based on the genetic algorithm formula:

[0163] Fitness = ∑ ,

[0164] Fitness: Optimizing the numerical value of the objective function, Error : No. Error values ​​of the control parameters.

[0165] This invention provides an intelligent insulator monitoring and fault detection system for high-voltage power grids. It has the following beneficial effects:

[0166] 1. This invention achieves an innovative breakthrough through multi-source sensor collaborative acquisition and data fusion processing. By acquiring multi-dimensional data such as voltage, electric field, vibration, and environment in real time, and combining filtering, fusion, and global state evaluation technologies, it effectively improves data reliability and the ability to sensitively identify minor faults and sudden fluctuations, realizing a leap from "single parameter judgment" to "multi-dimensional data collaborative decision-making".

[0167] 2. This invention introduces an adaptive nonlinear electric field model and intelligent algorithms, and through online model updates and machine learning diagnostic algorithms, achieves dynamic modeling of voltage-electric field distribution and intelligent fault identification. It transforms from experience-based judgment to intelligent modeling, from static discrimination to dynamic prediction, and from coarse-grained positioning to fine-grained positioning, significantly improving fault identification accuracy and response speed.

[0168] 3. This invention employs self-powered and low-power communication technologies, combined with cloud platform remote monitoring, to enable long-term independent operation of equipment, remote parameter management, and rapid alarm response. Through the integrated architecture of "self-powered + low-power + cloud monitoring", it improves system robustness, deployment flexibility, and operation and maintenance efficiency. Attached Figure Description

[0169] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0170] To enable those skilled in the art to understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0171] The present invention will now be described in detail with reference to the accompanying drawings:

[0172] Example:

[0173] Please see the appendix Figure 1 This invention provides an intelligent insulator monitoring and fault detection system for high-voltage power grids, comprising:

[0174] S1: Multi-source sensor measurement module, multi-source information acquisition, overcoming the limitations of single parameters;

[0175] S2: Multi-source information fusion processing module, multi-source data fusion processing, constructing a global status evaluation;

[0176] S3: Adaptive nonlinear electric field modeling module, which constructs adaptive nonlinear electric field models to improve the adaptability to complex electric fields;

[0177] S4: Intelligent Fault Diagnosis and Location Module, providing intelligent fault diagnosis and precise location to improve monitoring accuracy;

[0178] S5: Self-powered and low-power communication module, with self-powered and low-power communication design to enhance system robustness;

[0179] S6: Remote monitoring and feedback platform, cloud platform real-time monitoring, enabling rapid response.

[0180] Furthermore, the step 1 includes:

[0181] The multi-source sensor measurement module includes a voltage sensor, an electric field strength sensor, a leakage current sensor, a temperature and humidity sensor, and a vibration sensor. It is used to synchronously collect multi-dimensional parameters related to the insulator's operating status and form a multi-source raw dataset for subsequent data fusion processing.

[0182] Step 1.1 The multidimensional electric-field-humidity coupling index reflects the coupling characteristics between electric field strength, humidity, and leakage current. It can be used for early identification of insulation degradation. Its formula is:

[0183] ,

[0184] The electric field value measured by the electric field strength sensor. Leakage current, Relative humidity : Coupling weighting coefficients for electric field, current, and humidity;

[0185] Step 1.2 The combined characteristic quantity of insulator dynamic vibration and current is given by the following formula:

[0186] ,

[0187] Acceleration data points sampled by the vibration sensor. Mean acceleration Number of vibration data sampling points Leakage current;

[0188] Step 1.3 Normalized state energy for multiple source parameters, the formula is:

[0189] ,

[0190] The parameter measured by the k-th sensor. The long-run mean of the k-th parameter. The standard deviation of the k-th parameter. Weight;

[0191] Step 1.4 Nonlinear voltage-electric field response deviation, its formula is:

[0192] ,

[0193] The measured electric field value, The voltage measured by the voltage sensor. Voltage-electric field nonlinear coefficients, obtained through fitting;

[0194] Step 1.5 Multi-source raw data entropy, its formula is:

[0195] ,

[0196] The probability of multi-source data appearing after quantization. Number of data bins.

[0197] Specifically, by synchronously collecting multi-dimensional data such as electric field, leakage current, humidity, and vibration from multiple sources of sensors, and combining multiple mathematical models to achieve data fusion and accurate evaluation, this multi-dimensional analysis and fusion strategy can effectively identify insulator degradation in the early stages, improve the accuracy of fault prediction, and enhance the system's maintenance response capabilities.

[0198] Furthermore, step 2 includes:

[0199] The data fusion unit is used to fuse multidimensional data from multiple sensors. It employs data fusion algorithms such as Kalman filtering, particle filtering, and Bayesian inference to transform the raw data collected by each sensor into unified power grid status assessment information.

[0200] Step 2.1 The Kalman filter algorithm is used to extract and estimate the true state of the system from noise. In power grid monitoring, the Kalman filter can be applied to the fusion of sensor data such as voltage, electric field, and humidity. Its formula is:

[0201] ,

[0202] Current moment State estimates, The previous moment State estimates, Kalman gain represents the trade-off between measurement and prediction. Current moment Sensor measurements : Observation matrix, used to map state variables to measurement space;

[0203] Step 2.2 Particle filtering is applicable to nonlinear systems. In the fusion of multi-source sensor data, it can estimate the probability distribution of the system state. Its formula is:

[0204] ,

[0205] State estimation at the current time k. The state of the i-th particle in particle filtering. : The weight of particle i, adjusted according to the measured matching degree; N: The number of particles, used to approximate the distribution of the system state;

[0206] Step 2.3 Bayesian inference, based on probability theory, estimates the posterior probability by updating the prior probability, which is then used to fuse results from different sensors. The formula is as follows:

[0207] ,

[0208] State variables The posterior probability, Observational data The conditional probability, State variables The prior probability, Observational data The marginal likelihood.

[0209] Specifically, data fusion algorithms such as Kalman filtering, particle filtering, and Bayesian inference are used to effectively integrate data from multiple sensors, thereby improving the accuracy and reliability of power grid condition assessment and optimizing power grid monitoring and control strategies.

[0210] Furthermore, step 2 also includes:

[0211] The status assessment module is used to build a global status evaluation model of the power grid operation based on the fused data, and output the power grid health index and fault early warning information.

[0212] Step 2.4 After multi-source data fusion, the power grid health index is calculated using a global state evaluation model to provide a comprehensive assessment of the power grid's operating status. The formula is as follows:

[0213] ,

[0214] The power grid health index measures the overall health of the power grid. : No. The parameters measured by each sensor : No. Long-term average of each sensor, : No. The standard deviation of each sensor, : No. The weight of each sensor, Number of sensors;

[0215] Step 2.5 Based on the global state evaluation model, fault early warning information can be issued according to the changing trend of the health index, and the formula is as follows:

[0216] ,

[0217] : Fault warning information; if the value falls below a threshold, an alarm will be generated. A set health threshold is set; a value below this threshold indicates that a power grid failure may occur.

[0218] Step 2.6 uses time series analysis to assess the trend of the health status of multidimensional parameters, assisting in the real-time monitoring and fault prediction of the decision support system. The formula is as follows:

[0219] Trend Analysis = ∑i=1N (wi · d(xi(t)) / dt),

[0220] Trend Analysis: Multidimensional health trend analysis results, d(xi(t)) / dt: the rate of change of the i-th sensor over time, representing the trend of health change, xi(t): the measurement value of the i-th sensor over time, wi: the weight of the i-th sensor.

[0221] Specifically, a power grid health index is calculated by combining a global state evaluation model with multi-source data, and fault warnings are issued based on its changing trends. Time series analysis and trend assessment are used to assist in real-time monitoring and fault prediction, thereby improving the safety and reliability of power grid operation.

[0222] Furthermore, step 3 includes:

[0223] The electric field modeling unit is used to construct a nonlinear mathematical model of the electric field based on voltage, electric field and other sensor data. The model can automatically adjust its parameters to adapt to different electric field distributions according to changes in the power grid operating environment.

[0224] Step 3.1 To accurately describe the distribution of the electric field in a complex power grid environment, a nonlinear model is usually required, and its formula is as follows:

[0225] ,

[0226] Nonlinear electric field intensity :Voltage, Electric field intensity coefficient The nonlinear exponent of voltage with respect to electric field strength Leakage current, The coefficient representing the effect of leakage current on the electric field. : Nonlinear exponent of leakage current;

[0227] Step 3.2 To improve the adaptability of the electric field model, an adaptive algorithm is adopted, the formula of which is:

[0228] ,

[0229] The electric field strength after adaptive adjustment. : Initial electric field model output, Current moment The measured electric field data, Adaptive step size controls the model's update speed. The previous moment The estimated value of the electric field;

[0230] The model update mechanism is used to update the model online based on real-time monitoring data during power grid operation. It optimizes the electric field modeling results through adaptive algorithms to improve the model's adaptability to complex electric field environments.

[0231] Step 3.3 Formula for updating the electric field model using the adaptive filtering algorithm:

[0232] ,

[0233] Current moment Model parameters, The previous moment Model parameters, Adaptive step size, used to adjust the update amplitude. Current moment Measurement data, Observation matrix;

[0234] Step 3.4 Least squares model optimization is used to improve the prediction accuracy of the electric field model by reducing model error. The formula is as follows:

[0235] ,

[0236] : No. The actual electric field strength measurement values ​​of each data point Model-based and parameters Predicted electric field strength : No. Input features of each data point Total number of data points.

[0237] Specifically, the electric field intensity model is adjusted in real time by using a nonlinear electric field model and an adaptive algorithm. Combined with adaptive filtering and least squares optimization, the model can accurately predict electric field changes under the power grid operating environment, thereby improving the accuracy and adaptability of electric field modeling.

[0238] Furthermore, step 3 also includes:

[0239] The fault prediction module predicts and locates electric field anomalies and faults based on the real-time output of the nonlinear electric field model.

[0240] Step 3.5 Based on the nonlinear electric field model, statistical methods are used to detect abnormal fluctuations in the electric field. The formula is as follows:

[0241] Anomaly Score = ,

[0242] Anomaly Score: A higher value indicates a greater degree of electric field anomaly. The measured electric field strength, The electric field strength calculated based on the electric field model. The standard deviation of the electric field strength serves as a measure of electric field fluctuations.

[0243] Step 3.6 uses a prediction algorithm to predict potential faults based on electric field data. The formula is as follows:

[0244] ,

[0245] Fault prediction probability, representing the likelihood of a fault occurring. : No. The weights of each input feature, : No. Each input feature Bias term, used to adjust the model output. Activation function The total number of input features;

[0246] Step 3.7 Combines data from multiple sensors and uses multidimensional features for fault diagnosis, the formula of which is:

[0247] ,

[0248] Fault diagnosis results : No. The weight of each sensor data, : No. A function for processing sensor data.

[0249] Specifically, by using real-time electric field model output, anomaly detection, fault prediction, and multi-dimensional data diagnosis, it is possible to effectively identify electric field anomalies and predict potential faults, thereby improving the accuracy of early fault identification and location.

[0250] Furthermore, step 4 includes:

[0251] The intelligent diagnostic unit is used to analyze the operating status of insulators in real time based on the aforementioned multi-source information fusion results and the output of the adaptive nonlinear electric field model, and to identify different types of faults such as flashover, partial discharge, and mechanical loosening by using intelligent diagnostic methods such as pattern recognition algorithms, deep neural networks, or support vector machines.

[0252] Step 4.1 The Support Vector Machine (SVM) finds the optimal hyperplane for classification by maximizing the classification margin. The formula is as follows:

[0253] ,

[0254] Input data The classification results Lagrange multipliers in Support Vector Machines Support vectors' category labels. Kernel functions are used to calculate the similarity between input data and support vectors. Support vectors The bias term determines the position of the classification plane;

[0255] Step 4.2 Deep neural networks learn features through multiple layers of neurons, and the formula is as follows:

[0256] ,

[0257] y. Network output, : No. The weight matrix of the layer, : No. Layer bias terms, Activation function Input data;

[0258] Step 4.3 For different types of faults, a multi-class classification model is used for judgment, and the formula is as follows:

[0259] ,

[0260] Predicted fault type This formula selects the most likely fault type by maximizing the posterior probability of each fault type.

[0261] The fault feature extraction unit is used to extract and classify features from multi-dimensional sensor data such as voltage, electric field, leakage current, and vibration. By constructing multi-dimensional state feature vectors, it improves the accuracy of fault type identification.

[0262] Step 4.4 The formula for extracting frequency domain features using Fourier transform is as follows:

[0263] ,

[0264] Frequency domain representation of the signal Time-domain signal :frequency;

[0265] Step 4.5 Time-Frequency Analysis: Local features of the signal are processed using wavelet transform, as shown in the formula:

[0266] ,

[0267] The signal after wavelet transform. : Scaling factor, controls the resolution of time and frequency. Displacement factor, controlling the time and position of the signal. Basic wavelet functions;

[0268] Step 4.6 Feature selection and dimensionality reduction, the formula is as follows:

[0269] ,

[0270] z. The dimensionality-reduced feature vectors Transformation matrix : Original high-dimensional feature vector.

[0271] The precise positioning unit is used to accurately locate the fault location based on electric field distribution inversion, signal delay analysis, or spatial correlation calculation, and generate spatial fault coordinates corresponding to the insulator structure, so as to realize rapid location and visualization of the fault location.

[0272] Step 4.7 Electric field distribution inversion is used to estimate the fault location based on the measured electric field data. Through the inversion algorithm, the electric field data can be correlated with the fault source location. The formula is as follows:

[0273] ,

[0274] : The measured electric field value Spatial location coordinates Location of the fault source The propagation function describes the propagation of the electric field from the fault source to the measurement point. : Electric field intensity distribution of the fault source;

[0275] Step 4.8 Signal Delay Analysis: The fault location is estimated by analyzing the propagation time of the fault signal between different sensors. The signal propagation time difference is related to the location, and its formula is:

[0276] ,

[0277] : Propagation delay The distance between sensors Signal propagation speed;

[0278] Step 4.9 Spatial correlation analysis is used to calculate the correlation between measurements from different sensors to help locate the fault source. The formula is:

[0279] ,

[0280] :sensor and Spatial correlation between them :sensor and Covariance between :sensor and The standard deviation.

[0281] Specifically, through intelligent diagnosis, feature extraction and dimensionality reduction, and fault location technology, the system can accurately identify and locate insulator fault types, and combine multiple algorithms to improve the accuracy and efficiency of fault detection and location.

[0282] Furthermore, step 5 includes:

[0283] The energy harvesting unit is used to acquire and convert weak energy from the high-voltage power grid operating environment using electromagnetic coupling, photovoltaic energy or piezoelectric energy harvesting technology, so as to provide a continuous and stable power supply for the system.

[0284] Step 5.1 Electromagnetic Energy Harvesting: Changes in the high-voltage power grid are converted into electrical energy through electromagnetic induction. This is achieved using Faraday's law of electromagnetic induction, the formula of which is:

[0285] ,

[0286] : , : , : , : ;

[0287] Step 5.2 The photovoltaic energy harvesting formula is:

[0288] ,

[0289] Photovoltaic output voltage, Open circuit voltage, Load current, : Series resistance of photovoltaic module;

[0290] Step 5.3 The formula for piezoelectric energy harvesting is:

[0291] ,

[0292] V: Output voltage, dQ: Charge change caused by applied pressure, C: Capacitance;

[0293] The energy management unit is used to store, regulate and distribute the collected energy. It achieves dynamic power control through a low-power power management chip, including energy storage, stable voltage output and system adaptive power consumption adjustment, to ensure reliable power supply to the system in extreme environments.

[0294] Step 5.4 Energy Storage and Regulation: For the battery charging process, the battery charging formula can be used:

[0295] ,

[0296] Charging current, Battery capacity, Battery voltage change rate;

[0297] Step 5.5 Voltage stabilization and output regulation uses a switching power supply for control. The relationship between the output voltage and the feedback signal is as follows:

[0298] ,

[0299] Output voltage, Reference voltage, Feedback resistor, : Set the resistor.

[0300] Specifically, a highly reliable self-powered system is constructed by utilizing multiple energy harvesting technologies, including electromagnetic, photovoltaic, and piezoelectric technologies, to extract meager energy from the high-voltage power grid environment. This energy is then stored, stabilized, and dynamically controlled through an energy management unit. Multi-source energy harvesting enhances power supply stability, while intelligent energy management ensures continuous and reliable operation even in extreme environments.

[0301] Furthermore, step 5 also includes:

[0302] The low-power communication unit adopts a self-organizing network communication protocol to realize low-power, long-distance data transmission between multiple nodes. It also reduces the overall energy consumption of the system and improves the stability and robustness of communication through sleep strategies, data compression and adaptive communication strategies.

[0303] Step 5.6 Low-power communication modules typically use ad hoc networking protocols. The formula for estimating communication power consumption is as follows:

[0304] ,

[0305] Transmission power consumption Power consumption in idle state Power consumption during transmission. The amount of data transmitted. Data transmission rate;

[0306] Step 5.7 The energy consumption formula for hibernation mode is:

[0307] ,

[0308] Power consumption in sleep mode Idle power consumption : Duration of hibernation;

[0309] Step 5.8 The formula for the adaptive communication strategy is:

[0310] ,

[0311] Power consumption of adaptive communication Base power consumption The amount of data to be transmitted. Transmission time : Adaptability coefficient.

[0312] Specifically, it achieves low-power, long-distance communication through a self-organizing network protocol, and reduces system energy consumption by combining sleep and adaptive communication strategies; its power consumption model makes energy use quantifiable and optimized, thereby improving the stability and robustness of communication, and is suitable for long-term online monitoring systems.

[0313] Furthermore, step 6 includes:

[0314] The cloud-based data processing and storage unit is used to aggregate, clean, classify, and store multi-dimensional data such as voltage, electric field, leakage current, vibration, and environmental parameters from multiple monitoring terminals in the cloud, and to achieve high-concurrency data processing and intelligent analysis based on cloud computing resources.

[0315] Step 6.1 Data aggregation and cleaning, based on the Z-Score method formula:

[0316] ,

[0317] X: Data sample The mean of the data; : Standard deviation of the data; Z: Z score, used to determine whether the data is outlier;

[0318] Step 6.2 Data Classification and Storage: The formula for the K-means clustering algorithm based on machine learning is as follows:

[0319] ,

[0320] Clustering cost function Data points Clustering Membership degree between them Clustering The center point, Data points Number of data points Number of cluster centers;

[0321] Step 6.3 High-concurrency data processing uses the MapReduce model to process data in parallel. The formula is as follows:

[0322] , ,

[0323] Map: A mapping function that maps input... Convert to intermediate key-value pairs Reduce: A function that reduces intermediate key-value pairs. Aggregate into the final output ;

[0324] The real-time monitoring and early warning unit is used to monitor the operating status of insulators in real time through the cloud platform. It automatically triggers early warning signals based on historical big data models and risk threshold calculations, and pushes abnormal information to maintenance personnel to achieve remote rapid response.

[0325] Step 6.4 Real-time early warning model, based on the cumulative risk model formula:

[0326] ,

[0327] Cumulative risk value, representing the overall risk at the current time t. Real-time data of sensor i at time t. Risk weight of sensor i The number of sensors;

[0328] Step 6.5 The threshold detection and early warning triggering formula is as follows:

[0329] ,

[0330] Warning signal Current moment The monitoring value, : Preset threshold;

[0331] The feedback control and command issuance unit is used to generate corresponding control strategies in the cloud based on the fault information provided by the fault diagnosis and location module, and to issue commands to the field monitoring module through the wireless communication link to realize remote control actions such as sampling frequency adjustment, sensor calibration, node wake-up, and energy-saving mode switching, thus constructing a closed-loop feedback mechanism.

[0332] Step 6.6 Remote control strategy generation, the formula is as follows:

[0333] ,

[0334] u. Output control strategy, : Input fuzzy set;

[0335] Step 6.7 Instruction Issuance and Execution: The formula for calculating instruction transmission delay is as follows:

[0336] ,

[0337] L: Command transmission delay; B: Data size; C: Network bandwidth. Processing time;

[0338] Step 6.8 Remote control action optimization, based on the genetic algorithm formula:

[0339] Fitness = ∑ ,

[0340] Fitness: Optimizing the numerical value of the objective function, Error : No. Error values ​​of the control parameters.

[0341] Specifically, the system aggregates, cleans, and processes multi-dimensional monitoring data in the cloud, provides real-time early warnings based on risk models, and combines fuzzy control and genetic algorithm optimization to enable command issuance and remote control, thus constructing an efficient and intelligent cloud monitoring and feedback mechanism that significantly improves the system's reliability and response efficiency.

[0342] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A high-voltage power grid intelligent insulator monitoring and fault detection system, characterized in that, include: S1: Multi-source sensor measurement module, multi-source information acquisition, overcoming the limitations of single parameters; S2: Multi-source information fusion processing module, multi-source data fusion processing, constructing a global status evaluation; S3: Adaptive nonlinear electric field modeling module, which constructs adaptive nonlinear electric field models to improve the adaptability to complex electric fields; S4: Intelligent Fault Diagnosis and Location Module, providing intelligent fault diagnosis and precise location to improve monitoring accuracy; S5: Self-powered and low-power communication module, with self-powered and low-power communication design to enhance system robustness; S6: Remote monitoring and feedback platform, real-time cloud platform monitoring, enabling rapid response.

2. The intelligent insulator monitoring and fault detection system for high-voltage power grids according to claim 1, characterized in that, The step 1 includes: The multi-source sensor measurement module includes a voltage sensor, an electric field strength sensor, a leakage current sensor, a temperature and humidity sensor, and a vibration sensor. It is used to synchronously collect multi-dimensional parameters related to the insulator's operating status and form a multi-source raw dataset for subsequent data fusion processing. Step 1.1 Multidimensional electric-field-humidity coupling index, used to reflect the coupling characteristics between electric field strength, humidity and leakage current, can be used for early identification of insulation degradation. Its formula is: , The electric field value measured by the electric field strength sensor. Leakage current, Relative humidity : Coupling weighting coefficients for electric field, current, and humidity; Step 1.2 The combined characteristic quantity of insulator dynamic vibration and current is given by the following formula: , Acceleration data points sampled by the vibration sensor. Mean acceleration Number of vibration data sampling points Leakage current; Step 1.3 Normalized state energy for multiple source parameters, the formula is: , The parameter measured by the k-th sensor. The long-run mean of the k-th parameter. The standard deviation of the k-th parameter. Weight; Step 1.4 Nonlinear voltage-electric field response deviation, its formula is: , The measured electric field value, The voltage measured by the voltage sensor. Voltage-electric field nonlinear coefficients, obtained through fitting; Step 1.5 Multi-source raw data entropy, its formula is: , The probability of multi-source data appearing after quantization. Number of data bins.

3. The intelligent insulator monitoring and fault detection system for high-voltage power grids according to claim 1, characterized in that, The step 2 includes: The data fusion unit is used to fuse multidimensional data from multiple sensors. It employs data fusion algorithms such as Kalman filtering, particle filtering, and Bayesian inference to transform the raw data collected by each sensor into unified power grid status assessment information. Step 2.1 The Kalman filter algorithm is used to extract and estimate the true state of the system from noise. In power grid monitoring, the Kalman filter can be applied to the fusion of sensor data such as voltage, electric field, and humidity. Its formula is: , Current moment State estimates, The previous moment State estimates, Kalman gain represents the trade-off between measurement and prediction. Current moment Sensor measurements : Observation matrix, used to map state variables to measurement space; Step 2.2 Particle filtering is applicable to nonlinear systems. In the fusion of multi-source sensor data, it can estimate the probability distribution of the system state. Its formula is: , State estimation at time k. The state of the i-th particle in particle filtering. : The weight of particle i, adjusted according to the measured matching degree; N: The number of particles, used to approximate the distribution of the system state; Step 2.3 Bayesian inference, based on probability theory, estimates the posterior probability by updating the prior probability, which is then used to fuse results from different sensors. The formula is as follows: ， State variables The posterior probability, Observational data The conditional probability, State variables The prior probability, Observational data The marginal likelihood.

4. The intelligent insulator monitoring and fault detection system for high-voltage power grids according to claim 1, characterized in that, The step 2 also includes: The status assessment module is used to build a global status evaluation model of the power grid operation based on the fused data, and output the power grid health index and fault early warning information. Step 2.4 After multi-source data fusion, the power grid health index is calculated using a global state evaluation model to provide a comprehensive assessment of the power grid's operating status. The formula is as follows: , The power grid health index measures the overall health of the power grid. : No. The parameters measured by each sensor : No. Long-term average of each sensor, : No. The standard deviation of each sensor, : No. The weight of each sensor, Number of sensors; Step 2.5 Based on the global state evaluation model, fault early warning information can be issued according to the changing trend of the health index, and the formula is as follows: , : Fault warning information; if the value falls below a threshold, an alarm will be generated. A set health threshold is set; a value below this threshold indicates that a power grid failure may occur. Step 2.6 uses time series analysis to assess the trend of the health status of multidimensional parameters, assisting in the real-time monitoring and fault prediction of the decision support system. The formula is as follows: Trend Analysis = ∑i=1N (wi · d(xi(t)) / dt), Trend Analysis: Multidimensional health trend analysis results, d(xi(t)) / dt: the rate of change of the i-th sensor over time, representing the trend of health change, xi(t): the measurement value of the i-th sensor over time, wi: the weight of the i-th sensor.

5. The intelligent insulator monitoring and fault detection system for high-voltage power grids according to claim 1, characterized in that, The step 3 includes: The electric field modeling unit is used to construct a nonlinear mathematical model of the electric field based on voltage, electric field and other sensor data. The model can automatically adjust its parameters to adapt to different electric field distributions according to changes in the power grid operating environment. Step 3.1 To accurately describe the distribution of the electric field in a complex power grid environment, a nonlinear model is usually required, and its formula is as follows: , Nonlinear electric field intensity :Voltage, Electric field intensity coefficient The nonlinear exponent of voltage with respect to electric field strength Leakage current, The coefficient representing the effect of leakage current on the electric field. : Nonlinear exponent of leakage current; Step 3.2 To improve the adaptability of the electric field model, an adaptive algorithm is adopted, the formula of which is: , The electric field strength after adaptive adjustment. : Initial electric field model output, Current moment The measured electric field data, Adaptive step size controls the model's update speed. The previous moment The estimated value of the electric field; The model update mechanism is used to update the model online based on real-time monitoring data during power grid operation. It optimizes the electric field modeling results through adaptive algorithms to improve the model's adaptability to complex electric field environments. Step 3.3 Formula for updating the electric field model using the adaptive filtering algorithm: , Current moment Model parameters, The previous moment Model parameters, Adaptive step size, used to adjust the update amplitude. Current moment Measurement data, Observation matrix; Step 3.4 Least squares model optimization is used to improve the prediction accuracy of the electric field model by reducing model error. The formula is as follows: , : No. The actual electric field strength measurement values ​​of each data point Model-based and parameters Predicted electric field strength : No. Input features of each data point Total number of data points.

6. The intelligent insulator monitoring and fault detection system for high-voltage power grids according to claim 1, characterized in that, The step 3 also includes: The fault prediction module predicts and locates electric field anomalies and faults based on the real-time output of the nonlinear electric field model. Step 3.5 Based on the nonlinear electric field model, statistical methods are used to detect abnormal fluctuations in the electric field. The formula is as follows: Anomaly Score = , Anomaly Score: A higher value indicates a greater degree of electric field anomaly. The measured electric field strength, The electric field strength calculated based on the electric field model. The standard deviation of the electric field strength serves as a measure of electric field fluctuations. Step 3.6 uses a prediction algorithm to predict potential faults based on electric field data. The formula is as follows: , Fault prediction probability, representing the likelihood of a fault occurring. : No. The weights of each input feature, : No. Each input feature Bias term, used to adjust the model output. Activation function The total number of input features; Step 3.7 Combines data from multiple sensors and uses multidimensional features for fault diagnosis, the formula of which is: , Fault diagnosis results : No. The weight of each sensor data, : No. A function for processing sensor data.

7. The intelligent insulator monitoring and fault detection system for high-voltage power grids according to claim 1, characterized in that, The step 4 includes: The intelligent diagnostic unit is used to analyze the operating status of insulators in real time based on the aforementioned multi-source information fusion results and the output of the adaptive nonlinear electric field model, and to identify different types of faults such as flashover, partial discharge, and mechanical loosening by using intelligent diagnostic methods such as pattern recognition algorithms, deep neural networks, or support vector machines. Step 4.1 The Support Vector Machine (SVM) finds the optimal hyperplane for classification by maximizing the classification margin. The formula is as follows: , Input data The classification results Lagrange multipliers in Support Vector Machines Support vectors' category labels. Kernel functions are used to calculate the similarity between input data and support vectors. Support vectors The bias term determines the position of the classification plane; Step 4.2 Deep neural networks learn features through multiple layers of neurons, and the formula is as follows: , Network output, : No. The weight matrix of the layer, : No. Layer bias terms, Activation function Input data; Step 4.3 For different types of faults, a multi-class classification model is used for judgment, and the formula is as follows: , Predicted fault type This formula selects the most likely fault type by maximizing the posterior probability of each fault type. The fault feature extraction unit is used to extract and classify features from multi-dimensional sensor data such as voltage, electric field, leakage current, and vibration. By constructing multi-dimensional state feature vectors, it improves the accuracy of fault type identification. Step 4.4 The formula for extracting frequency domain features using Fourier transform is as follows: , Frequency domain representation of the signal Time-domain signal :frequency; Step 4.5 Time-Frequency Analysis: Local features of the signal are processed using wavelet transform, with the following formula: , The signal after wavelet transform. : Scaling factor, controls the resolution of time and frequency. Displacement factor, controlling the time and position of the signal. Basic wavelet functions; Step 4.6 Feature selection and dimensionality reduction, the formula is as follows: ， : The feature vector after dimensionality reduction Transformation matrix : Original high-dimensional feature vector; The precise positioning unit is used to accurately locate the fault location based on electric field distribution inversion, signal delay analysis, or spatial correlation calculation, and generate spatial fault coordinates corresponding to the insulator structure, so as to realize rapid location and visualization of the fault location. Step 4.7 Electric field distribution inversion is used to estimate the fault location based on the measured electric field data. Through the inversion algorithm, the electric field data can be correlated with the fault source location. The formula is as follows: , : The measured electric field value Spatial location coordinates Location of the fault source The propagation function describes the propagation of the electric field from the fault source to the measurement point. : Electric field intensity distribution of the fault source; Step 4.8 Signal Delay Analysis: The fault location is estimated by analyzing the propagation time of the fault signal between different sensors. The signal propagation time difference is related to the location, and its formula is: ， : Propagation delay The distance between sensors Signal propagation speed; Step 4.9 Spatial correlation analysis is used to calculate the correlation between measurements from different sensors to help locate the fault source. The formula is: , :sensor and Spatial correlation between them :sensor and Covariance between :sensor and The standard deviation.

8. The intelligent insulator monitoring and fault detection system for high-voltage power grids according to claim 1, characterized in that, The step based on step 5 includes: The energy harvesting unit is used to acquire and convert weak energy from the high-voltage power grid operating environment using electromagnetic coupling, photovoltaic energy or piezoelectric energy harvesting technology, so as to provide a continuous and stable power supply for the system. Step 5.1 Electromagnetic Energy Harvesting: Changes in the high-voltage power grid are converted into electrical energy through electromagnetic induction. This is achieved using Faraday's law of electromagnetic induction, the formula of which is: ， ： ， ： ， ： ， ： ； Step 5.2 The photovoltaic energy harvesting formula is: ， Photovoltaic output voltage, Open circuit voltage, Load current, : Series resistance of photovoltaic module; Step 5.3 The formula for piezoelectric energy harvesting is: ， V: Output voltage, dQ: Charge change caused by applied pressure, C: Capacitance; The energy management unit is used to store, regulate and distribute the collected energy. It achieves dynamic power control through a low-power power management chip, including energy storage, stable voltage output and system adaptive power consumption adjustment, to ensure reliable power supply to the system in extreme environments. Step 5.4 Energy Storage and Regulation: For the battery charging process, the battery charging formula can be used: , Charging current, Battery capacity, Battery voltage change rate; Step 5.5 Voltage stabilization and output regulation uses a switching power supply for control. The relationship between the output voltage and the feedback signal is as follows: , Output voltage, Reference voltage, Feedback resistor, : Set the resistor.

9. The intelligent insulator monitoring and fault detection system for high-voltage power grids according to claim 1, characterized in that, The step based on step 5 also includes: The low-power communication unit adopts a self-organizing network communication protocol to realize low-power, long-distance data transmission between multiple nodes. It also reduces the overall energy consumption of the system and improves the stability and robustness of communication through sleep strategies, data compression and adaptive communication strategies. Step 5.6 Low-power communication modules typically use ad hoc networking protocols. The formula for estimating communication power consumption is as follows: , Transmission power consumption Power consumption in idle state Power consumption during transmission. The amount of data transmitted. Data transmission rate; Step 5.7 The energy consumption formula for hibernation mode is: ， Power consumption in sleep mode Idle power consumption : Duration of hibernation; Step 5.8 The formula for the adaptive communication strategy is: ， Power consumption of adaptive communication Base power consumption The amount of data to be transmitted. Transmission time : Adaptability coefficient.

10. The intelligent insulator monitoring and fault detection system for high-voltage power grids according to claim 1, characterized in that, The step based on step 6 includes: The cloud-based data processing and storage unit is used to aggregate, clean, classify, and store multi-dimensional data such as voltage, electric field, leakage current, vibration, and environmental parameters from multiple monitoring terminals in the cloud, and to achieve high-concurrency data processing and intelligent analysis based on cloud computing resources. Step 6.1 Data aggregation and cleaning, based on the Z-Score method formula: ， X: Data sample : The mean of the data; : Standard deviation of the data; Z: Z score, used to determine whether the data is outlier; Step 6.2 Data Classification and Storage: The formula for the K-means clustering algorithm based on machine learning is as follows: , Clustering cost function Data points Clustering Membership degree between them Clustering The center point, Data points Number of data points Number of cluster centers; Step 6.3 High-concurrency data processing uses the MapReduce model to process data in parallel. The formula is as follows: , , Map: A mapping function that maps input... Convert to intermediate key-value pairs Reduce: A function that reduces intermediate key-value pairs. Aggregate into the final output ; The real-time monitoring and early warning unit is used to monitor the operating status of insulators in real time through the cloud platform. It automatically triggers early warning signals based on historical big data models and risk threshold calculations, and pushes abnormal information to maintenance personnel to achieve remote rapid response. Step 6.4 Real-time early warning model, based on the cumulative risk model formula: , Cumulative risk value, representing the overall risk at the current time t. Real-time data of sensor i at time t. Risk weight of sensor i The number of sensors; Step 6.5 The threshold detection and early warning triggering formula is as follows: , Warning signal Current moment The monitoring value, : Preset threshold; The feedback control and command issuance unit is used to generate corresponding control strategies in the cloud based on the fault information provided by the fault diagnosis and location module, and to issue commands to the field monitoring module through the wireless communication link to realize remote control actions such as sampling frequency adjustment, sensor calibration, node wake-up, and energy-saving mode switching, thus constructing a closed-loop feedback mechanism. Step 6.6 Remote control strategy generation, the formula is as follows: , Output control strategy : Input fuzzy set; Step 6.7 Instruction Issuance and Execution: The formula for calculating instruction transmission delay is as follows: ， L: Command transmission delay; B: Data size; C: Network bandwidth. Processing time; Step 6.8 Remote control action optimization, based on the genetic algorithm formula: Fitness = ∑ , Fitness: Optimizing the numerical value of the objective function, Error : No. Error values ​​of the control parameters.

Images