High-voltage fuse thermal instability early warning method

Abstract

The invention discloses a thermal instability early warning method for a high-voltage fuse, and relates to the technical field of fault early warning of high-voltage power equipment, and the method comprises the steps: carrying out the real-time sensing of the high-voltage fuse, and constructing a thermal state feature vector through feature analysis; building two channels containing thermal instability prediction and early warning sub-branches; synchronizing the feature vector to a prediction sub-branch to calculate and generate a thermal instability risk index, and transmitting the index to an early warning sub-branch to generate a thermal instability initial early warning signal; and finally, the initial signal is sent to a central control end for inversion tracking verification, a thermal instability initial early warning signal is updated according to a signal verification result, a final thermal instability early warning signal is formed, linkage response is executed, and accurate early warning control is realized. The technical problems that in a traditional early warning method, the monitoring data dimension is single, and evaluation is lack of accuracy and dynamic adaptability are solved, and the technical effects that reliable early warning and efficient management and control of the equipment thermal instability risk are achieved, and operation safety of key equipment of an intelligent power grid is guaranteed are achieved.

Description

Technical Field

[0001] This invention relates to the field of fault early warning technology for high-voltage power equipment, and in particular to a method for early warning of thermal instability of high-voltage fuses. Background Technology

[0002] The stable operation of a smart grid relies on the safety of critical equipment. High-voltage fuses, as core protection components, are crucial for fault prediction and health management, directly impacting the reliability of power supply. Current technologies for high-voltage fuse thermal status detection largely depend on single-sensor acquisition or fixed-threshold alarms, playing a role under stable operating conditions. However, due to the complex operating environment of the power grid and the dynamic load on equipment, traditional methods cannot comprehensively capture the multi-dimensional characteristics of thermal status and lack dynamic prediction and closed-loop verification mechanisms. This results in inaccurate thermal instability risk assessments and delayed early warnings, failing to meet the precise equipment status control requirements of smart grid fault prediction and health management. Summary of the Invention

[0003] This application provides a method for early warning of thermal instability of high-voltage fuses, which solves the technical problems of single monitoring data dimension, lack of accuracy and dynamic adaptability in traditional early warning methods.

[0004] The first aspect of this application provides a method for early warning of thermal instability in high-voltage fuses. The method includes: real-time operation sensing of the high-voltage fuse to obtain a real-time monitoring dataset for feature analysis, and constructing a thermal state feature vector; constructing a dual-channel thermal instability system, wherein the dual-channel includes a thermal instability prediction sub-branch and a thermal instability early warning sub-branch; synchronizing the thermal state feature vector to the thermal instability prediction sub-branch for risk calculation, generating a thermal instability risk index; synchronizing the thermal instability risk index to the thermal instability early warning sub-branch for thermal instability analysis, and generating an initial thermal instability early warning signal; sending the initial thermal instability early warning signal to a central control terminal for inversion tracking verification of the high-voltage fuse; updating the initial thermal instability early warning signal based on the signal verification result; and constructing a thermal instability early warning signal for linkage response.

[0005] In the second aspect of the present application, a thermal instability warning system for a high-voltage fuse is provided. The system includes: a thermal state feature vector construction module for performing real-time operation sensing on the high-voltage fuse, obtaining a real-time monitoring data set for feature analysis, and constructing a thermal state feature vector; a thermal instability dual-channel construction module for constructing a thermal instability dual-channel, where the thermal instability dual-channel includes a thermal instability prediction sub-branch and a thermal instability warning sub-branch; a thermal instability initial warning signal acquisition module for synchronizing the thermal state feature vector to the thermal instability prediction sub-branch for risk calculation, generating a thermal instability risk index, and synchronizing the thermal instability risk index to the thermal instability warning sub-branch for thermal instability analysis, generating a thermal instability initial warning signal; a thermal instability warning signal construction module for sending the thermal instability initial warning signal to the central control terminal for inversion tracking verification of the high-voltage fuse, updating the thermal instability initial warning signal according to the signal verification result, and constructing a thermal instability warning signal for linkage response.

[0006] One or more technical solutions provided in the present application have at least the following technical effects or advantages: In the present application, data is collected through multi-dimensional real-time sensing of the high-voltage fuse, and thermal state-related information is obtained through processing such as filtering, noise reduction, and feature integration. After constructing the feature vector, the risk level is calculated through collaborative analysis of the two branches, and the warning signal is optimized by combining historical data backtracking verification and confidence adjustment, so as to accurately predict the thermal-related risks of the equipment and trigger an appropriate response, achieving the technical effects of reliable warning and efficient control of the thermal instability risks of the equipment and ensuring the safe operation of the key equipment of the smart grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0008] Figure 1 It is a schematic flowchart of a method for warning thermal instability of a high-voltage fuse provided by an embodiment of the present application.

[0009] Figure 2 It is a schematic structural diagram of a thermal instability warning system for a high-voltage fuse provided by an embodiment of the present application.

[0010] Description of the reference numerals: Thermal state feature vector construction module 1, Thermal instability dual-channel construction module 2, Thermal instability initial warning signal acquisition module 3, Thermal instability warning signal construction module 4. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] This application provides a method for early warning of thermal instability of high-voltage fuses, which solves the technical problems of single monitoring data dimension, lack of accuracy and dynamic adaptability in traditional early warning methods.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0013] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0014] Example 1, as Figure 1 As shown, a method for early warning of thermal instability of a high-voltage fuse is disclosed, wherein the method includes: Real-time operation sensing is performed on high-voltage fuses to obtain real-time monitoring datasets for feature analysis, and thermal state feature vectors are constructed.

[0015] Specifically, by deploying multiple sensing units in the high-voltage fuse to synchronously collect raw data of current, voltage, and temperature, the load level, load characteristics, energy input, absolute thermal state, and thermal dynamic trend characteristics are extracted through filtering and noise reduction, effective data analysis, and fluctuation calculation. After multi-dimensional integration, a thermal state feature vector is constructed.

[0016] A dual-channel thermal instability mechanism is constructed, which includes a thermal instability prediction sub-branch and a thermal instability early warning sub-branch.

[0017] Optionally, firstly, long-term operating data of high-voltage fuses are obtained by constructing multiple virtual environment parameters. After stage division and labeling, training and test sets are divided. The training model identifies and evaluates thermal instability modes. Once the evaluation value meets the preset index, the model structure parameters are determined and loaded into the embedded data processing unit to complete the construction of the thermal instability prediction sub-branch.

[0018] Next, based on the multiple virtual environmental parameters used to construct the thermal instability prediction sub-branch, the operating conditions of the high-voltage fuse are analyzed and the operating condition combination parameters are set. Accordingly, a threshold query library containing the initial risk warning threshold and the initial critical alarm threshold is established. The real-time environmental parameters are retrieved to define dynamic adjustment rules to fine-tune the initial threshold and obtain the target threshold. The warning triggering conditions are set and integrated in combination with the target threshold to complete the construction of the thermal instability warning sub-branch.

[0019] The thermal state feature vector is synchronized to the thermal instability prediction sub-branch for risk calculation, generating a thermal instability risk index. The thermal instability risk index is then synchronized to the thermal instability early warning sub-branch for thermal instability analysis, generating an initial thermal instability early warning signal.

[0020] In one embodiment of this application, the thermal state feature vector is first sorted according to the sampling time to construct a time series data block. After defining the risk preset range, it is synchronized to the thermal instability prediction sub-branch. Deep-level time series pattern features are extracted and three types of thermal risk indices, namely thermal accumulation, heat dissipation lag and fault precursor, are calculated respectively. After integration, a thermal instability risk index is generated.

[0021] Next, the thermal instability risk index is synchronized to the thermal instability early warning sub-branch, and compared with the risk warning threshold and critical alarm threshold to determine the low risk, warning risk or alarm risk level. Combined with the auxiliary diagnostic information generated by the thermal state feature vector, a comprehensive analysis is performed to generate the initial thermal instability early warning signal.

[0022] The initial thermal instability warning signal is sent to the central control terminal to perform inversion tracking and verification of the high-voltage fuse. The initial thermal instability warning signal is updated according to the signal verification result, and a thermal instability warning signal is constructed for linkage response.

[0023] In this embodiment, the central control terminal is a dedicated monitoring host or server system that has the functions of receiving, analyzing, storing, and issuing linkage commands for high-voltage fuse operation data.

[0024] Specifically, firstly, after receiving the thermal instability early warning signal, the central control terminal extracts the equipment identifier and retrieves its historical operation records for trend backtracking analysis. It then conducts real-time event correlation analysis in conjunction with real-time data monitoring streams. By combining the results of the two types of analysis, it completes the inversion tracking and verification of the early warning signal and generates the signal verification result.

[0025] Next, confidence analysis is carried out based on the signal verification results. The thermal instability risk index is adjusted accordingly with high and low confidence levels. The root causes of the risk are traced and the risk level is redefined based on the verification results. The initial warning signal for thermal instability is updated accordingly to obtain the thermal instability warning signal.

[0026] Finally, a linkage response strategy matrix is ​​constructed with signal risk level and risk root cause as input keys. The target strategy cell is queried and located, and the response operation instructions are parsed and extracted. These instructions are converted into control commands and sent to the execution terminal. The execution status parameters are then fed back to the central control terminal to complete the linkage response.

[0027] Furthermore, the method provided in this application embodiment includes: Multiple sensing units are deployed across the high-voltage fuse, and these units are integrated for synchronous data acquisition to obtain a raw synchronous monitoring data stream, which includes current, voltage, and temperature data. Based on the current, voltage, and temperature data, filtering and noise reduction are performed to construct current, voltage, and temperature sequences. Real-time data analysis is then performed according to the current and voltage sequences to obtain multiple valid datasets, which include real-time RMS current values, real-time RMS voltage values, real-time active power, and power factor. Fluctuation calculations are then performed on key components of the high-voltage fuse according to the temperature sequence to determine the instantaneous temperature. The instantaneous temperature value includes the rate of temperature change; based on the real-time effective current value and the real-time effective voltage value, load level characteristics are generated; based on the power factor, load analysis is performed to generate load nature characteristics; based on the real-time active power, energy analysis is performed to generate energy input characteristics; based on the instantaneous temperature value, key components are analyzed to generate absolute thermal state characteristics; based on the rate of temperature change, trend analysis is performed to generate thermal dynamic trend characteristics; the load level characteristics, load nature characteristics, energy input characteristics, absolute thermal state characteristics, and thermal dynamic trend characteristics are integrated in multiple dimensions to construct the thermal state feature vector.

[0028] Specifically, firstly, current transformers, voltage transformers, and thermistor sensors are installed at key locations such as the contacts, fuse element, and base of the high-voltage fuse. A network time protocol is used to synchronize the timestamps of multiple sensing units, ensuring that each unit collects data at the same time point, with a data acquisition frequency of 1 minute per acquisition. By integrating the acquisition modules of each sensing unit, a raw synchronous monitoring data stream containing current, voltage, and temperature data is synchronously acquired. Next, a Kalman filter is applied to the current and voltage data in the raw synchronous monitoring data stream. By establishing state and observation equations, the optimal estimate for the current moment is calculated using the estimated value from the previous moment and the observed value from the current moment, eliminating interference factors such as grid harmonics. For the temperature data, a moving average filter is used, selecting the temperature data from five consecutive sampling points to calculate the arithmetic mean, eliminating interference from ambient temperature fluctuations, and constructing smooth current, voltage, and temperature sequences respectively.

[0029] Then, on the one hand, real-time data analysis is performed on the current and voltage sequences. The real-time RMS values ​​of current and voltage are calculated using the arithmetic mean method after full-wave rectification. Real-time active power is calculated by multiplying the real-time RMS voltage, real-time RMS current, and power factor. The power factor is calculated by dividing the active power by the apparent power, which is obtained by multiplying the real-time RMS voltage and real-time RMS current. Finally, multiple effective datasets containing real-time RMS current, real-time RMS voltage, real-time active power, and power factor are obtained. On the other hand, for the temperature sequence, real-time data is directly extracted as the instantaneous temperature values ​​of key components of the high-voltage fuse. The temperature change rate is calculated using the difference method. The instantaneous temperature values ​​at two adjacent sampling times are selected, and the temperature value at the later time is subtracted from the temperature value at the previous time, then divided by the time interval between the two sampling times to obtain the temperature change rate within that time period. The temperature change rate is incorporated into the relevant data of the instantaneous temperature values.

[0030] Subsequently, based on the real-time RMS current and voltage values, the ratio of the real-time RMS current to the rated current of the high-voltage fuse is calculated, and this ratio is used as a load level characteristic. Based on the power factor, the load type is determined: a power factor greater than 0.9 indicates a resistive load, and a power factor less than 0.9 indicates an inductive load; this load type is used as a load characteristic. Based on real-time active power, the cumulative active power over a certain period is calculated, and this cumulative value is used as an energy input characteristic. Based on instantaneous temperature values, the difference between the temperature of key components and the ambient temperature is calculated, and this difference is used as an absolute thermal state characteristic. Based on the rate of temperature change, the sign of the rate of change indicates a rising or falling temperature trend, and the absolute value of the rate of change indicates the speed of temperature change; the combination of trend and speed of change is used as a thermal dynamic trend characteristic.

[0031] Finally, the load level characteristics, load nature characteristics, energy input characteristics, absolute thermal state characteristics, and thermal dynamic trend characteristics are arranged in a fixed order to form a five-dimensional thermal state feature vector. Each feature corresponds to one dimension of the vector, ensuring that the feature vector can comprehensively reflect the thermal state of the high-voltage fuse.

[0032] Through multi-sensor synchronous acquisition, filtering and noise reduction, effective data analysis, feature extraction and multi-dimensional integration, a comprehensive perception and accurate characterization of the thermal state of high-voltage fuses is achieved, providing a reliable feature data foundation for subsequent thermal instability early warning.

[0033] Furthermore, the method provided in this application embodiment includes: Multiple virtual environment parameters are constructed, and the high-voltage fuse is continuously monitored according to these parameters to generate long-term operational data. The long-term operational data is then traversed and divided into stages to generate multiple stage data. These stage data are then labeled to generate multiple state labels. Based on these state labels, the long-term operational data is divided into a training set and a test set. The training set is used to train the model and identify multiple thermal instability modes. The test set is used to accurately evaluate these multiple thermal instability modes, generating data evaluation values. These data evaluation values ​​are compared with preset evaluation indicators. When the data evaluation values ​​meet the preset evaluation indicators, the model structure parameters are determined. The model structure parameters are then loaded into an embedded data processing unit to construct the thermal instability prediction sub-branch.

[0034] In this embodiment, the virtual environment parameters are a set of simulated parameters for various actual operating conditions that affect the operation of the high-voltage fuse, such as ambient temperature, humidity, wind speed, and power grid load fluctuation range. The thermal instability mode is the specific thermal instability development state exhibited by the high-voltage fuse from normal operation to the occurrence of a thermal instability fault, including heat accumulation, delayed heat dissipation, and fault precursors.

[0035] Optionally, multiple virtual environmental parameters are constructed, covering key parameters related to the operation of high-voltage fuses, such as ambient temperature, humidity, wind speed, and power grid load fluctuation range. A simulation model of high-voltage fuse operation is built using Matlab / Simulink. The virtual environmental parameters are input into the model, and the high-voltage fuse is continuously simulated and monitored for 3 months. The sampling frequency is set to 1 minute / time to generate long-term series operation data. This data includes time-series data of current, voltage, and temperature under different environmental and load conditions, as well as data related to the equipment's operating status.

[0036] Next, the sliding window method is used to traverse the long-term series data to divide it into stages. The window size is set to 60 sampling points corresponding to 1 hour, and the step size is 30 sampling points. First, the current fluctuation amplitude in each window is calculated, which is the difference between the maximum and minimum current values ​​in the window, and the temperature change trend is calculated, which is the slope of the temperature data in the window obtained by linear fitting. Then, the difference in current fluctuation amplitude and the difference in temperature change trend slope between two adjacent windows are calculated. Two fixed judgment thresholds are preset, corresponding to the current fluctuation amplitude difference threshold and the temperature change trend slope difference threshold, respectively. When the difference in current fluctuation amplitude between adjacent windows exceeds the corresponding threshold, or the difference in temperature change trend slope exceeds the corresponding threshold, it is determined that a sudden change feature has occurred. The starting sampling point of the next window is used as the stage division node. Based on this, the long-term series data is divided into multiple stages such as normal operation stage, heat accumulation stage, heat dissipation lag stage, and fault precursor stage.

[0037] Then, the data from multiple stages are labeled according to their attributes. Using a combination of manual labeling and threshold judgment, the labels for the normal operation stage are set as 0, the heat accumulation stage as 1, the heat dissipation lag stage as 2, and the fault precursor stage as 3. The corresponding status labels are labeled for each time series sub-data in each stage, generating multiple status labels.

[0038] Based on multiple state labels, a stratified sampling method is used to divide the long-term series data into training and test sets. The training set accounts for 70% and the test set accounts for 30% to ensure that the label distribution ratio of the data in each stage of the training set and the test set is consistent, so as to avoid data bias affecting the model training effect. During the division process, a random seed of 42 is set to ensure that the results are repeatable.

[0039] A Long Short-Term Memory (LSTM) network model was constructed for the core of the thermal instability prediction sub-branch. The model consists of three layers: an input layer, a hidden layer, and an output layer. The input layer has 5 dimensions, corresponding to the 5 dimensions of the thermal state feature vector. There are 2 hidden layers, each with 64 neurons, using the tanh activation function. The forget gate, input gate, and output gate all use the sigmoid activation function. The output layer has 4 neurons, corresponding to 4 operating state labels, and uses the softmax activation function. The training set data was input into the model for training. The optimizer was set to Adam, the learning rate to 0.001, the cross-entropy loss function to be used, the number of iterations to be set to 100, and the batch size to be 32. During training, backpropagation was used to adjust the model weights and biases, identifying multiple thermal instability modes, including normal operation, heat accumulation, heat dissipation lag, and fault precursors. The test set data was input into the trained model, and the recognition performance of multiple thermal instability modes was evaluated using accuracy, precision, and recall, generating data evaluation values.

[0040] Next, the preset evaluation metrics are accuracy ≥ 95%, precision ≥ 94%, and recall ≥ 94%. The generated data evaluation values ​​are compared with the preset evaluation metrics one by one. When all data evaluation values ​​meet the preset evaluation metrics, the final model structure parameters are determined, including key parameters such as input layer dimension, number of hidden layers and number of neurons, optimizer type, learning rate, and number of iterations.

[0041] Finally, the STM32H743 microcontroller was selected as the embedded data processing unit. The determined model structure parameters were compiled into binary files using the KeilMDK development environment. The binary files were then burned into the flash memory of the embedded data processing unit using JTAG download, thus completing the solidification of the model parameters. At the same time, the unit's timer and communication interface were configured to ensure that it could receive thermal state feature vectors in real time and output calculation results, thus completing the construction of the thermal instability prediction sub-branch.

[0042] By generating multi-condition data through simulation modeling, processing data through hierarchical division and labeling, constructing and training a long short-term memory network model, and solidifying and deploying the model, the feasible construction of the thermal instability prediction sub-branch was realized, providing reliable model support for the accurate calculation of the thermal instability risk index.

[0043] Furthermore, the method provided in this application embodiment includes: The high-voltage fuse is analyzed for operating conditions based on multiple virtual environmental parameters, and multiple operating condition combination parameters are set. A threshold query library is set based on the multiple operating condition combination parameters, which includes an initial risk warning threshold and an initial critical alarm threshold. Real-time environmental parameters are retrieved to define dynamic adjustment rules, and the initial risk warning threshold and the initial critical alarm threshold are dynamically fine-tuned according to the dynamic adjustment rules to generate a risk warning threshold and a critical alarm threshold. Warning triggering conditions are set based on the risk warning threshold and the critical alarm threshold, and the warning triggering conditions are integrated with the risk warning threshold and the critical alarm threshold to construct the thermal instability warning sub-branch.

[0044] Specifically, based on the multiple virtual environmental parameters used in constructing the thermal instability prediction sub-branch in the aforementioned steps, including ambient temperature, humidity, wind speed, and power grid load fluctuation range, an enumeration method is used to perform operating condition combination analysis on these parameters. First, each virtual environmental parameter is divided into multiple discrete intervals. For example, ambient temperature is divided into intervals such as -10℃~10℃, 10℃~30℃, and 30℃~50℃, and power grid load fluctuation range is divided into intervals such as 0~20%, 20%~40%, and 40%~60%. Then, the intervals of different parameters are combined in pairs using the enumeration method to generate multiple operating condition combination parameters. Each operating condition combination parameter corresponds to a unique set of environmental and load operating conditions. For example, "ambient temperature 10℃~30℃ + power grid load fluctuation 20%~40% + humidity 40%~60% + wind speed 1~3m / s" is a set of operating condition combination parameters.

[0045] Then, a threshold query library is set based on the generated multiple operating condition combination parameters. This query library is constructed using a hash table. The key of the hash table is a unique identifier of the operating condition combination parameter, which is generated by concatenating the parameter range codes into strings. For example, if the code for "ambient temperature 10℃~30℃" is set as T2 and the code for "grid load fluctuation 20%~40%" is set as L2, then the corresponding operating condition combination key is T2-L2. The values ​​of the hash table are the initial risk warning threshold and the initial critical alarm threshold under the corresponding operating condition. The initial threshold is determined through statistical analysis of previous experimental data. The average risk index before the occurrence of thermal instability fault of high-voltage fuse under each operating condition is selected as the initial critical alarm threshold, and 70% of this average is selected as the initial risk warning threshold. The key-value pairs corresponding to each operating condition are entered into the hash table one by one to complete the construction of the threshold query library.

[0046] Next, real-time environmental parameters from the high-voltage fuse operating site are retrieved, including real-time ambient temperature, humidity, and wind speed. A piecewise linear adjustment algorithm is used to define dynamic adjustment rules. These rules are determined based on the degree of influence of each environmental parameter on the thermal instability of the high-voltage fuse. Increased ambient temperature reduces the heat dissipation efficiency of the high-voltage fuse, increasing the risk of thermal instability; therefore, a higher temperature corresponds to a lower adjustment ratio that lowers the threshold. Conversely, decreased temperature improves the heat dissipation efficiency and reduces the risk of thermal instability; therefore, a lower temperature corresponds to a higher adjustment ratio that raises the threshold. Humidity above a certain level affects the insulation performance and heat dissipation of the high-voltage fuse, increasing the risk of thermal instability; therefore, a corresponding adjustment ratio is set for humidity levels above a certain threshold. The adjustment ratio lowers the threshold, while increasing wind speed improves the heat dissipation efficiency of the high-voltage fuse and reduces the risk of thermal instability. Therefore, an adjustment ratio is set to adjust the threshold upward when the wind speed exceeds a certain value. The segmented linear adjustment interval is divided according to the critical interval of the impact of each environmental parameter on the thermal instability of the high-voltage fuse. Each interval corresponds to a linear adjustment trend. According to the above dynamic adjustment rules, the initial threshold corresponding to the current operating condition is retrieved from the threshold query library and substituted into the adjustment formula to calculate the dynamically fine-tuned risk warning threshold and critical alarm threshold. The adjustment formula is: adjusted threshold = initial threshold × (1 + sum of adjustment ratios corresponding to each environmental parameter), where the adjustment ratio corresponding to each environmental parameter is determined according to its interval and the degree of influence on thermal instability.

[0047] Finally, based on the dynamically adjusted risk warning threshold and critical alarm threshold, three core judgment conditions are set: when the thermal instability risk index is less than or equal to the risk warning threshold, it is judged as a low-risk level; when the thermal instability risk index is greater than the risk warning threshold but less than the critical alarm threshold, it is judged as a warning risk level; when the thermal instability risk index is greater than or equal to the critical alarm threshold, it is judged as an alarm risk level. Subsequently, the warning triggering conditions are integrated with the risk warning threshold, critical alarm threshold, hash table threshold query logic, and piecewise linear adjustment logic. First, the initial threshold is obtained by matching the hash table with real-time operating parameters, then the target threshold is obtained through dynamic adjustment, and finally the risk level judgment is executed to complete the construction of the thermal instability warning sub-branch.

[0048] By employing a series of steps including enumeration of operating conditions, hash table threshold storage, piecewise linear dynamic adjustment, and multi-condition branch determination, a thermal instability early warning sub-branch adapted to complex operating conditions was constructed. This enabled dynamic adaptation of the early warning threshold and accurate determination of the risk level, providing reliable judgment logic and threshold basis for the subsequent generation of initial thermal instability early warning signals.

[0049] Furthermore, the method provided in this application embodiment includes: The thermal state feature vectors are arranged sequentially according to multiple sampling times to construct a time series data block; a risk preset range is defined, and the time series data block is synchronized to the thermal instability prediction sub-branch for analysis and processing according to the risk preset range to extract deep-level time series pattern features; the probability of thermal accumulation risk is calculated based on the deep-level time series pattern features to generate a first thermal risk index; the probability of heat dissipation lag risk is calculated based on the deep-level time series pattern features to generate a second thermal risk index; the probability of fault precursor risk is calculated based on the deep-level time series pattern features to generate a third thermal risk index; the first thermal risk index, the second thermal risk index, and the third thermal risk index are integrated to construct the thermal instability risk index.

[0050] Specifically, firstly, the thermal state feature vectors are arranged in ascending order according to the timestamps of the sampling times to construct time series data blocks. Each data block is set to contain thermal state feature vectors from 60 consecutive sampling times, with the sampling time interval consistent with the previous data acquisition of 1 minute / time. The five-dimensional feature vectors of each time moment are concatenated in chronological order using array indexing to form a two-dimensional time series data block with a dimension of 60×5, ensuring that the data block can completely reflect the continuous change trend of the thermal state of the high-voltage fuse over a period of time.

[0051] Next, the risk preset range was defined as 0 to 1, corresponding to the complete interval from zero to the existence of thermal instability risk. The constructed time-series data blocks were synchronously input into the Long Short-Term Memory (LSTM) network model of the thermal instability prediction sub-branch. Before input, a min-max normalization method was used to map each feature value in the data block to the 0-1 interval to avoid the influence of differences in feature dimensions on the model's analysis performance. The model performs nonlinear transformation on the time-series data through the tanh activation function in the hidden layer, and utilizes the synergistic effect of the forget gate, input gate, and output gate to filter key time-series information, automatically extracting deep-level time-series pattern features reflecting the changing patterns of thermal state. These features are output in vector form, with a dimension consistent with the number of neurons in the model's hidden layer (64 dimensions).

[0052] Then, during the model training phase of the thermal instability prediction sub-branch in the aforementioned steps, after completing the stage division and labeling of the long-term series running data, all sample data with state labels indicating the thermal accumulation stage are selected. The same deep-level temporal pattern feature extraction operation as subsequent risk calculations is performed on these thermal accumulation sample data to obtain a set of feature vectors for the thermal accumulation samples. For each feature component related to thermal accumulation, its mean and standard deviation in the set of feature vectors for the thermal accumulation samples are calculated. The range of mean ± 2 times standard deviation is used to determine the thermal accumulation sample feature distribution interval corresponding to that feature component. The distribution intervals of all related feature components together constitute the thermal accumulation sample feature distribution interval. Then, based on the extracted deep-level temporal pattern features, the maximum likelihood estimation method is used to calculate the thermal accumulation risk probability. First, the feature components related to thermal accumulation in the feature vectors are statistically analyzed, such as the slope of the temperature change trend and the feature dimension corresponding to the cumulative energy input value. The probability that these feature components fall within the aforementioned thermal accumulation sample feature distribution interval is calculated. This probability is used as the first thermal risk index, with a value ranging from 0 to 1; a larger value indicates a higher thermal accumulation risk.

[0053] Furthermore, the acquisition process for the second and third thermal risk indices is similar to that of the first thermal risk index. In the model training phase of the thermal instability prediction sub-branch, sample data corresponding to the state labels are selected, and the same deep-level time-series pattern feature extraction operation is performed to obtain a set of sample feature vectors. For each feature component related to the corresponding risk, its mean and standard deviation in the set of sample feature vectors are calculated. The feature distribution interval is determined by the range of mean ± 2 times standard deviation. Based on the extracted deep-level time-series pattern features, the probability that the relevant feature component falls within the corresponding sample feature distribution interval is calculated using the maximum likelihood estimation method. This probability is used as the second and third thermal risk indices, respectively, with values ​​ranging from 0 to 1. The larger the value, the higher the corresponding risk.

[0054] Finally, a weighted summation method is used to integrate the first, second, and third thermal risk indices. For example, weighting coefficients are set to 0.4, 0.3, and 0.3, respectively. These weights can be determined based on experimental statistical results of the impact of each risk factor on thermal instability failure. The integration formula is: Thermal Instability Risk Index = First Thermal Risk Index × 0.4 + Second Thermal Risk Index × 0.3 + Third Thermal Risk Index × 0.3. The calculated thermal instability risk index ranges from 0 to 1, achieving a comprehensive quantitative characterization of the three types of thermal instability risks.

[0055] Through a series of steps including time-series data construction, model feature extraction, multi-dimensional risk probability calculation, and weighted integration, the thermal instability risk was accurately quantified. The generated thermal instability risk index provides a reliable quantitative basis for the subsequent level determination of the early warning sub-branch.

[0056] Furthermore, the method provided in this application embodiment includes: The thermal instability risk index is synchronized to the thermal instability early warning sub-branch and used for judgment based on the risk warning threshold and critical alarm threshold. When the thermal instability risk index is less than or equal to the risk warning threshold, it is determined to be a low-risk level. When the thermal instability risk index is greater than the risk warning threshold and less than the critical alarm threshold, it is determined to be a warning risk level. When the thermal instability risk index is greater than or equal to the critical alarm threshold, it is determined to be an alarm risk level. The thermal state feature vector is used as auxiliary judgment data in combination with the thermal instability risk index to diagnose the high-voltage fuse and generate auxiliary diagnostic information. Based on the low-risk level, the warning risk level, or the alarm risk level, combined with the auxiliary diagnostic information, a comprehensive analysis is performed to generate the initial thermal instability warning signal.

[0057] Specifically, the thermal instability risk index is synchronized to the thermal instability early warning sub-branch via a preset data communication interface. The early warning sub-branch calls the risk warning threshold and critical alarm threshold obtained after dynamic fine-tuning in the aforementioned steps, and uses a multi-condition branch judgment method to determine the relationship between the thermal instability risk index and the two thresholds one by one. The specific steps are as follows: When the thermal instability risk index is less than or equal to the risk warning threshold, the current high-voltage fuse is directly determined to be in a low-risk level, which corresponds to an extremely low risk of thermal instability and stable equipment operation. When the thermal instability risk index is greater than the risk warning threshold but less than the critical alarm threshold, the current level is determined to be in the warning risk level, which corresponds to a situation where thermal instability risk has emerged and the equipment operation needs to be monitored. When the thermal instability risk index is greater than or equal to the critical alarm threshold, the current level is determined to be in the alarm risk level, which corresponds to an extremely high risk of thermal instability and a situation where the equipment may be about to fail.

[0058] Then, the thermal state feature vector constructed in the aforementioned steps is retrieved as auxiliary judgment data. The high-voltage fuse is diagnosed using a feature threshold comparison method combined with a thermal instability risk index. First, based on sample data from the normal operation phase of the high-voltage fuse, the normal value ranges of each component in the thermal state feature vector, namely load level characteristics, load nature characteristics, energy input characteristics, absolute thermal state characteristics, and thermal dynamic trend characteristics, are determined. Then, each component of the current thermal state feature vector is compared with its corresponding normal value range, and the feature components that exceed the normal range and the degree of exceedance are recorded to generate auxiliary diagnostic information. The auxiliary diagnostic information must clearly define the type and severity of the current abnormal characteristics.

[0059] Finally, the determined risk level is comprehensively analyzed with the auxiliary diagnostic information. When the risk level is low and the auxiliary diagnostic information shows no abnormalities, an initial thermal instability warning signal is generated without prior warning. When the risk level is at a warning level, a warning signal containing the risk level and the cause of the abnormality is generated based on the type of abnormality in the auxiliary diagnostic information. When the risk level is at an alarm level, a warning signal containing an emergency alarm and details of the abnormality is generated based on the severity of the abnormality in the auxiliary diagnostic information. The initial thermal instability warning signal is output in a standardized digital code format, with different types of signals corresponding to unique codes, facilitating identification by subsequent signal transmission and processing modules.

[0060] Furthermore, the method provided in this application embodiment includes: When the central control terminal receives an initial thermal instability warning signal, it extracts the device identification information of the initial thermal instability warning signal, maps the device identification to the central database according to the timestamp, retrieves the data, and extracts historical operation records. Based on the historical operation records, it performs historical trend retrospective analysis on the initial thermal instability warning signal to obtain trend analysis results. It introduces a real-time data monitoring stream, and performs real-time event correlation analysis on the initial thermal instability warning signal based on the real-time data monitoring stream to obtain correlation analysis results. Based on the trend analysis results and the correlation analysis results, it verifies the initial thermal instability warning signal and generates a signal verification result.

[0061] In one embodiment, the initial thermal instability warning signal is first sent to the central control terminal via the industrial Ethernet communication protocol. Upon receiving the signal, the central control terminal uses a string parsing method to extract the device identification information carried in the signal. This device identification information is a unique code for the high-voltage fuse. Based on this unique code, query conditions are constructed according to the timestamp in the signal. The device identification is mapped to a central relational database using a database index matching method. The historical operating records of the high-voltage fuse for the past six months are then retrieved. These historical operating records include time-series data such as historical thermal state feature vectors, historical thermal instability risk indices, historical environmental parameters, and historical warning records.

[0062] Next, a time series trend analysis method was employed to conduct a historical trend retrospective analysis of the initial thermal instability warning signal based on retrieved historical operational data. A risk index change curve was plotted with time on the horizontal axis and the historical thermal instability risk index on the vertical axis. The curve was smoothed using a moving average method, and features such as the curve's slope and peak frequency were extracted. The risk level corresponding to the current initial thermal instability warning signal was compared with the development trend of similar risk levels in the past to determine whether the current warning signal is consistent with historical trends. Trend analysis results were generated, including two types: consistent trends and abnormal trends.

[0063] Then, a real-time data monitoring stream based on a Kafka message queue is introduced. This monitoring stream collects data such as the current thermal state feature vector of the high-voltage fuse, real-time environmental parameters, and power grid operating parameters in real time. A feature matching method is used to perform real-time event correlation analysis on the initial thermal instability warning signal based on the real-time data monitoring stream. The real-time collected data is matched one by one with the abnormal features corresponding to the initial thermal instability warning signal. Specifically, for the five features in the thermal state feature vector—load level feature, load nature feature, energy input feature, absolute thermal state feature, and thermal dynamic trend feature—the corresponding feature component in the real-time data is compared with the normal value range of the feature, marking abnormal features that exceed the normal range. Simultaneously, the identified abnormal feature types in the initial thermal instability warning signal are recorded. The number of successfully matched abnormal features is counted, with a threshold of 3. When the number of successfully matched abnormal features is greater than or equal to 3, the real-time data is deemed to sufficiently support the rationality of the current warning signal, and the correlation analysis result is "correlation established." When the number of successfully matched abnormal features is less than 3, the real-time data is deemed to insufficiently support the current warning signal, and the correlation analysis result is "correlation not established." This method is used to determine whether the real-time data can support the rationality of the current warning signal and obtain the correlation analysis result, which includes two types: "correlation established" and "correlation not established."

[0064] Finally, the initial warning signal for thermal instability is verified. If the trend analysis shows a consistent trend and the correlation analysis shows a valid correlation, the warning signal is deemed valid, and a verification result is generated. If the trend analysis shows an abnormal trend or the correlation analysis shows a invalid correlation, the warning signal is deemed abnormal, and a result indicating the signal needs further review is generated. If the trend analysis shows an abnormal trend and the correlation analysis shows a invalid correlation, the warning signal is deemed invalid, and a verification result indicating the signal fails is generated. All signal verification results are accompanied by key evidence from the trend analysis and correlation analysis, facilitating subsequent verification by those skilled in the art.

[0065] Through the above-mentioned sequential steps, the initial warning signal of thermal instability was accurately verified, ensuring the reliability of the warning signal and providing an accurate decision-making basis for the subsequent operation and control of high-voltage fuses.

[0066] Furthermore, the method provided in this application embodiment includes: Based on the signal verification results, confidence analysis is performed to construct confidence levels. The thermal instability risk index is then weighted and corrected according to these confidence levels: S1: When the confidence level is greater than a preset confidence threshold, a high confidence level is identified, and the thermal instability risk index is maintained and weighted to generate a first risk weight coefficient; S2: When the confidence level is less than a preset confidence threshold, a low confidence level is identified, and the thermal instability risk index is adjusted downwards to generate a second risk weight coefficient; Based on the first and second risk weight coefficients combined with the signal verification results, risk tracing is performed to determine the root cause information of the risk; The risk level is redefined according to the root cause information to determine the signal risk level; The initial thermal instability warning signal is updated based on the signal risk level and the root cause information to obtain a thermal instability warning signal.

[0067] Optionally, confidence analysis is performed based on the signal verification results, using a weighted comprehensive scoring method to calculate the confidence level. Trend analysis and correlation analysis results are used as the two core indicators for confidence analysis. A value of 0.5 is assigned for consistent trends, 0.2 for abnormal trends, 0.5 for established correlations, and 0.2 for invalid correlations. The confidence level is calculated using the weighted summation formula: Confidence Level = Trend Analysis Indicator Score + Correlation Analysis Indicator Score, with a value ranging from 0 to 1. Confidence levels are constructed, classifying confidence levels into high and low confidence levels. A preset confidence threshold of 0.7 is used; a confidence level greater than or equal to 0.7 is considered high confidence, and a confidence level less than 0.7 is considered low confidence. The thermal instability risk index is then weighted and adjusted according to the confidence level.

[0068] When the confidence level is high, indicating high reliability of the current warning signal, a maintenance weighting strategy is adopted to generate a first risk weight coefficient. The first risk weight coefficient is set to 1.0. The corrected thermal instability risk index = original thermal instability risk index × first risk weight coefficient. This coefficient maintains the quantitative result of the original thermal instability risk index, ensuring that the risk index is not erroneously adjusted in high-confidence scenarios. When the confidence level is low, indicating insufficient reliability of the current warning signal, a downward adjustment weighting strategy is adopted to generate a second risk weight coefficient. The second risk weight coefficient is set to 0.8. The corrected thermal instability risk index = original thermal instability risk index × second risk weight coefficient. This coefficient reduces the quantitative result of the risk index, avoiding over-warning in low-confidence scenarios.

[0069] Subsequently, based on the first or second risk weighting coefficient combined with the signal verification results, fault tree analysis was used to trace the source of risk. The abnormal risk corresponding to the corrected thermal instability risk index was taken as the top event of the fault tree, and the trend analysis details and correlation analysis details in the signal verification results were taken as intermediate events, decomposing downwards to the bottom events. Bottom events included abnormal environmental factors, power grid load fluctuations, equipment aging, and monitoring data errors. By comparing the correlation between the corrected risk index and each bottom event, the bottom event with the highest correlation was selected as the root cause information of the risk. The root cause information needed to clearly define the specific type and occurrence scenario of the risk.

[0070] Next, based on the root cause information, the severity of the risk root causes is divided into three levels: high, medium, and low. The modified thermal instability risk index is also divided into three intervals: high, medium, and low. A risk matrix is ​​constructed, where each intersection point corresponds to a signal risk level, including low risk, early warning risk, and alarm risk levels. By matching the severity of the risk root causes with the intervals of the modified risk index, the corresponding results are extracted from the risk matrix to determine the final signal risk level.

[0071] Finally, based on the determined signal risk level and risk root cause information, the risk level field is extracted from the initial thermal instability warning signal and replaced with the newly determined signal risk level. A risk root cause information field is added to the initial thermal instability warning signal, and the determined risk root cause information is filled in. The updated signal is then standardized and coded to ensure that the coding format is consistent with the recognition requirements of the central control terminal and the linkage response module. After the update is completed, the thermal instability warning signal is obtained.

[0072] Through the above-mentioned sequential steps, the initial early warning signal for thermal instability was precisely optimized, ensuring the reliability and relevance of the early warning signal and providing accurate signal support for the linkage response of high-voltage fuses.

[0073] Furthermore, the method provided in this application embodiment includes: A linkage response strategy matrix is ​​constructed, using the signal risk level and the risk root cause information as input keys. Based on the input keys, the linkage response strategy matrix is ​​queried and located to determine the target strategy cell. The target strategy cell is parsed to extract response operation instructions, which are then converted into tasks to generate control commands. The control commands are then sent to the execution terminal, generating execution status parameters that are fed back to the central control terminal for linkage response.

[0074] In one embodiment, a linkage response strategy matrix is constructed using a two-dimensional matrix structure. The row dimension is set as the signal risk level, including three classifications: low risk level, warning risk level, and alarm risk level. The column dimension is set as the risk source information, including four classifications: abnormal environmental factors, grid load fluctuations, equipment aging, and monitoring data errors. Each intersection point of the matrix corresponds to a target strategy cell. During the construction process, in combination with historical fault handling cases of high-voltage fuses and the experience of domain experts, response operation strategies for the combination of corresponding signal risk levels and risk source information are pre-stored in each target strategy cell. The strategy content clearly defines core elements such as the execution entity, operation actions, and execution parameters, ensuring the executability of the strategy. Using the signal risk level and risk source information as input keys, the double-keyword exact matching method is used to query and locate the linkage response strategy matrix. The signal risk level in the input key is compared one by one with the row dimension labels of the matrix, and at the same time, the risk source information in the input key is compared one by one with the column dimension labels of the matrix. When the row dimension label is exactly the same as the signal risk level and the column dimension label is exactly the same as the risk source information, the cell corresponding to this intersection point is determined as the target strategy cell.

[0075] Next, the target strategy cell is parsed, and response operation instructions are extracted from the response operation strategies pre-stored in the cell according to the preset instruction structure. The response operation instructions include core contents such as the execution terminal type, operation actions, and execution parameters. The instruction standardization conversion method is used to perform task conversion on the response operation instructions. According to the communication protocol and control requirements of the execution terminal, the extracted response operation instructions are converted into control commands recognizable by the execution terminal. The control commands include fields such as the execution terminal address, operation code, and parameter value. Among them, the operation code corresponds one by one to the actions of the execution terminal, and the parameter value is determined according to the execution parameters in the response operation instructions, ensuring that the control commands can be accurately recognized and executed by the execution terminal.

[0076] Finally, the industrial bus communication protocol is used to encapsulate the control commands into standard data packets, and the data packets are sent to the corresponding execution terminals through the preset communication interface. The execution terminals include environmental regulation devices, operation and maintenance notification terminals, power supply control devices, etc. Different types of control commands correspond to different execution terminals. After receiving the control commands, the execution terminals execute the corresponding operations according to the operation codes and parameter values in the commands, and at the same time collect the status parameters during the execution process, including operation execution time, execution result, current operating status of the equipment, etc., to form execution status parameters. The execution terminals adopt an active reporting mechanism to encapsulate the execution status parameters into feedback data packets and upload them to the central control end through the same communication interface. After receiving the feedback data packets, the central control end analyzes and stores the execution status parameters to complete the closed-loop management of the linkage response.

[0077] By constructing a two-dimensional linkage response strategy matrix, accurately matching and locating target strategy cells with dual keywords, extracting and standardizing structured instructions, and implementing a coherent process of industrial bus communication for issuing and status feedback, precise linkage response to thermal instability early warning signals is achieved. This ensures that differentiated response measures are taken for different signal risk levels and risk root cause information, thereby improving the safety and reliability of high-voltage fuse operation.

[0078] In summary, the high-voltage fuse thermal instability early warning method provided in this application has the following technical effects: This application constructs a feature vector by collecting thermal state and environmental parameters of high-voltage fuses, calculates the risk index through the thermal instability prediction sub-branch, determines the risk level by combining the dynamic threshold of the early warning sub-branch, updates the early warning signal through inversion verification at the central control terminal, and generates control commands for execution by matching the linkage response strategy. This achieves accurate early warning and response to thermal instability, improves the safety and reliability of equipment operation, and achieves the technical effect of reliable early warning and efficient management of equipment thermal instability risk, ensuring the safe operation of key equipment in the smart grid.

[0079] Example 2, as Figure 2 As shown, based on the same inventive concept as the aforementioned Embodiment 1, this application provides a high-voltage fuse thermal instability early warning system, the system comprising: Thermal state feature vector construction module 1 is used to perform real-time operation sensing on the high-voltage fuse, obtain real-time monitoring dataset, perform feature analysis, and construct thermal state feature vector.

[0080] Thermal instability dual-channel construction module 2 is used to construct a thermal instability dual channel, which includes a thermal instability prediction sub-branch and a thermal instability early warning sub-branch.

[0081] The thermal instability initial warning signal acquisition module 3 is used to synchronize the thermal state feature vector to the thermal instability prediction sub-branch for risk calculation, generate a thermal instability risk index, synchronize the thermal instability risk index to the thermal instability warning sub-branch for thermal instability analysis, and generate a thermal instability initial warning signal.

[0082] Thermal instability early warning signal construction module 4 is used to send the initial thermal instability early warning signal to the central control terminal to perform inversion tracking and verification of the high-voltage fuse, update the initial thermal instability early warning signal according to the signal verification result, and construct a thermal instability early warning signal for linkage response.

[0083] Furthermore, the thermal state feature vector construction module 1 is used to perform the following steps: Multiple sensing units are deployed across the high-voltage fuse, and these units are integrated for synchronous data acquisition to obtain a raw synchronous monitoring data stream, which includes current, voltage, and temperature data. Based on the current, voltage, and temperature data, filtering and noise reduction are performed to construct current, voltage, and temperature sequences. Real-time data analysis is then performed according to the current and voltage sequences to obtain multiple valid datasets, which include real-time RMS current values, real-time RMS voltage values, real-time active power, and power factor. Fluctuation calculations are then performed on key components of the high-voltage fuse according to the temperature sequence to determine the instantaneous temperature. The instantaneous temperature value includes the rate of temperature change; based on the real-time effective current value and the real-time effective voltage value, load level characteristics are generated; based on the power factor, load analysis is performed to generate load nature characteristics; based on the real-time active power, energy analysis is performed to generate energy input characteristics; based on the instantaneous temperature value, key components are analyzed to generate absolute thermal state characteristics; based on the rate of temperature change, trend analysis is performed to generate thermal dynamic trend characteristics; the load level characteristics, load nature characteristics, energy input characteristics, absolute thermal state characteristics, and thermal dynamic trend characteristics are integrated in multiple dimensions to construct the thermal state feature vector.

[0084] Furthermore, the thermal instability dual-channel construction module 2 is used to perform the following steps: Multiple virtual environment parameters are constructed, and the high-voltage fuse is continuously monitored according to these parameters to generate long-term operational data. The long-term operational data is then traversed and divided into stages to generate multiple stage data. These stage data are then labeled to generate multiple state labels. Based on these state labels, the long-term operational data is divided into a training set and a test set. The training set is used to train the model and identify multiple thermal instability modes. The test set is used to accurately evaluate these multiple thermal instability modes, generating data evaluation values. These data evaluation values ​​are compared with preset evaluation indicators. When the data evaluation values ​​meet the preset evaluation indicators, the model structure parameters are determined. The model structure parameters are then loaded into an embedded data processing unit to construct the thermal instability prediction sub-branch.

[0085] Furthermore, the thermal instability dual-channel construction module 2 is used to perform the following steps: The high-voltage fuse is analyzed for operating conditions based on multiple virtual environmental parameters, and multiple operating condition combination parameters are set. A threshold query library is set based on the multiple operating condition combination parameters, which includes an initial risk warning threshold and an initial critical alarm threshold. Real-time environmental parameters are retrieved to define dynamic adjustment rules, and the initial risk warning threshold and the initial critical alarm threshold are dynamically fine-tuned according to the dynamic adjustment rules to generate a risk warning threshold and a critical alarm threshold. Warning triggering conditions are set based on the risk warning threshold and the critical alarm threshold, and the warning triggering conditions are integrated with the risk warning threshold and the critical alarm threshold to construct the thermal instability warning sub-branch.

[0086] Furthermore, the thermal instability initial early warning signal acquisition module 3 is used to perform the following steps: The thermal state feature vectors are arranged sequentially according to multiple sampling times to construct a time series data block; a risk preset range is defined, and the time series data block is synchronized to the thermal instability prediction sub-branch for analysis and processing according to the risk preset range to extract deep-level time series pattern features; the probability of thermal accumulation risk is calculated based on the deep-level time series pattern features to generate a first thermal risk index; the probability of heat dissipation lag risk is calculated based on the deep-level time series pattern features to generate a second thermal risk index; the probability of fault precursor risk is calculated based on the deep-level time series pattern features to generate a third thermal risk index; the first thermal risk index, the second thermal risk index, and the third thermal risk index are integrated to construct the thermal instability risk index.

[0087] Furthermore, the thermal instability initial early warning signal acquisition module 3 is used to perform the following steps: The thermal instability risk index is synchronized to the thermal instability early warning sub-branch and used for judgment based on the risk warning threshold and critical alarm threshold. When the thermal instability risk index is less than or equal to the risk warning threshold, it is determined to be a low-risk level. When the thermal instability risk index is greater than the risk warning threshold and less than the critical alarm threshold, it is determined to be a warning risk level. When the thermal instability risk index is greater than or equal to the critical alarm threshold, it is determined to be an alarm risk level. The thermal state feature vector is used as auxiliary judgment data in combination with the thermal instability risk index to diagnose the high-voltage fuse and generate auxiliary diagnostic information. Based on the low-risk level, the warning risk level, or the alarm risk level, combined with the auxiliary diagnostic information, a comprehensive analysis is performed to generate the initial thermal instability warning signal.

[0088] The high-voltage fuse thermal instability early warning system provided in this embodiment of the invention can execute the high-voltage fuse thermal instability early warning method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0089] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0090] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for early warning of thermal instability in high-voltage fuses, characterized in that, The methods include: Real-time operation sensing of high-voltage fuses is performed to obtain real-time monitoring datasets for feature analysis and to construct thermal state feature vectors. A dual-channel thermal instability mechanism is constructed, which includes a thermal instability prediction sub-branch and a thermal instability early warning sub-branch. The thermal state feature vector is synchronized to the thermal instability prediction sub-branch for risk calculation, generating a thermal instability risk index. The thermal instability risk index is then synchronized to the thermal instability early warning sub-branch for thermal instability analysis, generating an initial thermal instability early warning signal. The initial thermal instability warning signal is sent to the central control terminal to perform inversion tracking and verification of the high-voltage fuse. The initial thermal instability warning signal is updated according to the signal verification result, and a thermal instability warning signal is constructed for linkage response.

2. The method for early warning of thermal instability of a high-voltage fuse as described in claim 1, characterized in that, Real-time operation sensing of high-voltage fuses is performed to obtain real-time monitoring datasets for feature analysis, and thermal state feature vectors are constructed. The methods include: Multiple sensing units are deployed across the high-voltage fuse, and the multiple sensing units are integrated for synchronous acquisition to obtain the original synchronous monitoring data stream, which includes current data, voltage data, and temperature data. Based on the current data, voltage data, and temperature data, filtering and noise reduction are performed to construct current sequences, voltage sequences, and temperature sequences. Real-time data analysis is performed on the current sequence and the voltage sequence to obtain multiple valid datasets, which include real-time effective current values, real-time effective voltage values, real-time active power, and power factor. Fluctuation calculations are performed on key components of the high-voltage fuse according to the temperature sequence to determine instantaneous temperature values, which include the rate of temperature change. Based on the analysis of the real-time current RMS value and the real-time voltage RMS value, load level characteristics are generated; based on the power factor, load analysis is performed to generate load characteristic characteristics. Energy analysis is performed based on the real-time active power to generate energy input characteristics; Based on the instantaneous temperature value, the key parts are analyzed to generate absolute thermal state characteristics, and based on the temperature change rate, trend analysis is performed to generate thermal dynamic trend characteristics. The load level characteristics, load property characteristics, energy input characteristics, absolute thermal state characteristics, and thermal dynamic trend characteristics are integrated in multiple dimensions to construct the thermal state feature vector.

3. The method for early warning of thermal instability of a high-voltage fuse as described in claim 1, characterized in that, The construction process and methods for the thermal instability prediction sub-branch include: Multiple virtual environment parameters are constructed, and the high-voltage fuse is continuously monitored according to the multiple virtual environment parameters to generate long-term series operation data; The long-term motion data is traversed and divided into stages to generate multiple stage data. The data from the multiple stages are labeled to generate multiple status labels; Based on the multiple state labels, the long-term series running data is divided into a training set and a test set; The training set is used to train the model, identify multiple thermal instability modes, and the test set is used to accurately evaluate the multiple thermal instability modes to generate data evaluation values. The data evaluation value is compared with the preset evaluation index. When the data evaluation value meets the preset evaluation index, the model structure parameters are determined. The model structure parameters are loaded into the embedded data processing unit to construct the thermal instability prediction sub-branch.

4. The method for early warning of thermal instability of a high-voltage fuse as described in claim 3, characterized in that, The construction process of the thermal instability early warning sub-branch includes the following methods: Based on the aforementioned multiple virtual environment parameters, the operating conditions of the high-voltage fuse are analyzed, and multiple operating condition combination parameters are set. A threshold query library is set based on the multiple operating condition combination parameters, and the threshold query library includes an initial risk warning threshold and an initial critical alarm threshold. Real-time environmental parameters are retrieved to define dynamic adjustment rules. The initial risk warning threshold and the initial critical alarm threshold are dynamically fine-tuned according to the dynamic adjustment rules to generate the risk warning threshold and the critical alarm threshold. Based on the risk warning threshold and the critical alarm threshold, a warning triggering condition is set, and the warning triggering condition is integrated with the risk warning threshold and the critical alarm threshold to construct the thermal instability warning sub-branch.

5. The method for early warning of thermal instability of a high-voltage fuse as described in claim 1, characterized in that, The thermal state feature vector is synchronized to the thermal instability prediction sub-branch for risk calculation to generate a thermal instability risk index. The method includes: The thermal state feature vectors are arranged sequentially according to multiple sampling times to construct a time series data block; Define the risk preset range, and synchronize the time series data block to the thermal instability prediction sub-branch according to the risk preset range for analysis and processing, and extract deep-level time series pattern features; Based on the deep temporal pattern characteristics, the probability of thermal accumulation risk is calculated to generate a first thermal risk index. Based on the deep-seated time-series pattern characteristics, the probability of heat dissipation lag risk is calculated to generate a second thermal risk index. Based on the aforementioned deep-seated temporal pattern characteristics, the probability of fault precursor risk is calculated to generate a third thermal risk index. The first thermal risk index, the second thermal risk index, and the third thermal risk index are integrated to construct the thermal instability risk index.

6. The method for early warning of thermal instability of a high-voltage fuse as described in claim 4, characterized in that, The thermal instability risk index is synchronized to the thermal instability early warning sub-branch for thermal instability analysis to generate an initial thermal instability early warning signal. The method includes: The thermal instability risk index is synchronized to the thermal instability early warning sub-branch and used to determine the risk early warning threshold and critical alarm threshold. When the thermal instability risk index is less than or equal to the risk warning threshold, it is determined to be a low-risk level; When the thermal instability risk index is greater than the risk warning threshold and less than the critical alarm threshold, it is determined to be a warning risk level. When the thermal instability risk index is greater than or equal to the critical alarm threshold, it is determined to be an alarm risk level; The thermal state feature vector is used as auxiliary judgment data and combined with the thermal instability risk index to diagnose the high-voltage fuse and generate auxiliary diagnostic information. Based on the low risk level, the early warning risk level, or the alarm risk level, combined with the auxiliary diagnostic information, a comprehensive analysis is performed to generate the initial early warning signal for thermal instability.

7. The method for early warning of thermal instability of a high-voltage fuse as described in claim 1, characterized in that, The method of sending the initial early warning signal of thermal instability to the central control terminal for inversion and tracing verification of the high-voltage fuse includes: When the central control terminal receives the initial warning signal of thermal instability, it extracts the equipment identification information of the initial warning signal of thermal instability, maps the equipment identification to the central database according to the timestamp, retrieves the data, and extracts the historical operation files; Based on the historical operation records, a historical trend retrospective analysis was performed on the initial early warning signal of thermal instability to obtain the trend analysis results. A real-time data monitoring stream is introduced, and a real-time event correlation analysis is performed on the initial early warning signal of thermal instability based on the real-time data monitoring stream to obtain the correlation analysis results. The initial warning signal for thermal instability is verified based on the trend analysis results and the correlation analysis results, and a signal verification result is generated.

8. The method for early warning of thermal instability of a high-voltage fuse as described in claim 1, characterized in that, The method includes updating the initial thermal instability warning signal based on the signal verification results and constructing a thermal instability warning signal for coordinated response: Based on the signal verification results, a confidence analysis is performed to construct a confidence level, and the thermal instability risk index is then weighted and corrected according to the confidence level. S1: When the confidence level is greater than the preset confidence threshold, a high confidence level is identified, and the thermal instability risk index is maintained for weighting to generate the first risk weight coefficient; S2: When the confidence level is less than the preset confidence threshold, a low confidence level is identified, and the thermal instability risk index is adjusted downward to generate a second risk weight coefficient. Based on the first risk weight coefficient, the second risk weight coefficient, and the signal verification results, risk source tracing is performed to determine the root cause information of the risk. The risk level is redefined based on the aforementioned risk source information to determine the signal risk level. The initial thermal instability warning signal is updated based on the signal risk level and the risk root cause information to obtain the thermal instability warning signal.

9. A method for early warning of thermal instability of a high-voltage fuse as described in claim 8, characterized in that, Methods for constructing thermal instability early warning signals and implementing coordinated responses include: Construct a linkage response strategy matrix, using the signal risk level and the risk root cause information as input keys, and query and locate the linkage response strategy matrix based on the input keys to determine the target strategy cell; The target strategy cell is parsed to extract response operation instructions, and the response operation instructions are converted into tasks to generate control commands; The control command is sent to the execution terminal, and the execution status parameters are generated and fed back to the central control terminal for linkage response.

10. A high-voltage fuse thermal instability early warning system, characterized in that, For implementing the high-voltage fuse thermal instability early warning method according to any one of claims 1-9, the system comprises: The thermal state feature vector construction module is used to perform real-time operation sensing on high-voltage fuses, obtain real-time monitoring datasets for feature analysis, and construct thermal state feature vectors. A thermal instability dual-channel construction module is used to construct a thermal instability dual-channel, which includes a thermal instability prediction sub-branch and a thermal instability early warning sub-branch; The thermal instability initial warning signal acquisition module is used to synchronize the thermal state feature vector to the thermal instability prediction sub-branch for risk calculation, generate a thermal instability risk index, synchronize the thermal instability risk index to the thermal instability warning sub-branch for thermal instability analysis, and generate a thermal instability initial warning signal. The thermal instability early warning signal construction module is used to send the initial thermal instability early warning signal to the central control terminal to perform inversion tracking and verification of the high-voltage fuse, update the initial thermal instability early warning signal according to the signal verification result, and construct the thermal instability early warning signal for linkage response.

11. A high-voltage fuse thermal instability early warning system as described in claim 10, characterized in that, The thermal state feature vector construction module includes: Multiple sensing units are deployed across the high-voltage fuse, and the multiple sensing units are integrated for synchronous acquisition to obtain the original synchronous monitoring data stream, which includes current data, voltage data, and temperature data. Based on the current data, voltage data, and temperature data, filtering and noise reduction are performed to construct current sequences, voltage sequences, and temperature sequences. Real-time data analysis is performed on the current sequence and the voltage sequence to obtain multiple valid datasets, which include real-time effective current values, real-time effective voltage values, real-time active power, and power factor. Fluctuation calculations are performed on key components of the high-voltage fuse according to the temperature sequence to determine instantaneous temperature values, which include the rate of temperature change. Based on the analysis of the real-time current RMS value and the real-time voltage RMS value, load level characteristics are generated; based on the power factor, load analysis is performed to generate load characteristic characteristics. Energy analysis is performed based on the real-time active power to generate energy input characteristics; Based on the instantaneous temperature value, the key parts are analyzed to generate absolute thermal state characteristics, and based on the temperature change rate, trend analysis is performed to generate thermal dynamic trend characteristics. The load level characteristics, load property characteristics, energy input characteristics, absolute thermal state characteristics, and thermal dynamic trend characteristics are integrated in multiple dimensions to construct the thermal state feature vector.

12. The high-voltage fuse thermal instability early warning system as described in claim 10, characterized in that, The thermal instability dual-channel construction module includes: Multiple virtual environment parameters are constructed, and the high-voltage fuse is continuously monitored according to the multiple virtual environment parameters to generate long-term series operation data; The long-term motion data is traversed and divided into stages to generate multiple stage data. The data from the multiple stages are labeled to generate multiple status labels; Based on the multiple state labels, the long-term series running data is divided into a training set and a test set; The training set is used to train the model, identify multiple thermal instability modes, and the test set is used to accurately evaluate the multiple thermal instability modes to generate data evaluation values. The data evaluation value is compared with the preset evaluation index. When the data evaluation value meets the preset evaluation index, the model structure parameters are determined. The model structure parameters are loaded into the embedded data processing unit to construct the thermal instability prediction sub-branch.

13. A high-voltage fuse thermal instability early warning system as described in claim 12, characterized in that, The thermal instability dual-channel construction module includes: Based on the aforementioned multiple virtual environment parameters, the operating conditions of the high-voltage fuse are analyzed, and multiple operating condition combination parameters are set. A threshold query library is set based on the multiple operating condition combination parameters, and the threshold query library includes an initial risk warning threshold and an initial critical alarm threshold. Real-time environmental parameters are retrieved to define dynamic adjustment rules. The initial risk warning threshold and the initial critical alarm threshold are dynamically fine-tuned according to the dynamic adjustment rules to generate the risk warning threshold and the critical alarm threshold. Based on the risk warning threshold and the critical alarm threshold, a warning triggering condition is set, and the warning triggering condition is integrated with the risk warning threshold and the critical alarm threshold to construct the thermal instability warning sub-branch.

14. The high-voltage fuse thermal instability early warning system as described in claim 10, characterized in that, The thermal instability initial early warning signal acquisition module includes: The thermal state feature vectors are arranged sequentially according to multiple sampling times to construct a time series data block; Define the risk preset range, and synchronize the time series data block to the thermal instability prediction sub-branch according to the risk preset range for analysis and processing, and extract deep-level time series pattern features; Based on the deep temporal pattern characteristics, the probability of thermal accumulation risk is calculated to generate a first thermal risk index. Based on the deep-seated time-series pattern characteristics, the probability of heat dissipation lag risk is calculated to generate a second thermal risk index. Based on the aforementioned deep-seated temporal pattern characteristics, the probability of fault precursor risk is calculated to generate a third thermal risk index. The first thermal risk index, the second thermal risk index, and the third thermal risk index are integrated to construct the thermal instability risk index.

15. A high-voltage fuse thermal instability early warning system as described in claim 13, characterized in that, The thermal instability initial early warning signal acquisition module includes: The thermal instability risk index is synchronized to the thermal instability early warning sub-branch and used to determine the risk early warning threshold and critical alarm threshold. When the thermal instability risk index is less than or equal to the risk warning threshold, it is determined to be a low-risk level; When the thermal instability risk index is greater than the risk warning threshold and less than the critical alarm threshold, it is determined to be a warning risk level. When the thermal instability risk index is greater than or equal to the critical alarm threshold, it is determined to be an alarm risk level; The thermal state feature vector is used as auxiliary judgment data and combined with the thermal instability risk index to diagnose the high-voltage fuse and generate auxiliary diagnostic information; based on the low risk level, the early warning risk level, or the alarm risk level, combined with the auxiliary diagnostic information, a comprehensive analysis is performed to generate the initial early warning signal for thermal instability.

Images