Adaptive threshold protection method and system based on measurement switch
By dynamically adjusting the protection threshold of the measuring switch using a BP neural network model and combining it with a fault prediction model for graded response, the problem of maloperation or failure to operate of traditional measuring switches is solved, thereby improving the protection performance and reliability of the power system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-05
- Publication Date
- 2026-03-27
AI Technical Summary
Traditional measurement switches rely on fixed thresholds to prevent false triggering or failure to trigger, which cannot adapt to complex and ever-changing actual operating environments. Furthermore, cloud-based processing solutions are susceptible to network latency and lack the ability to collaboratively predict multiple faults.
An adaptive threshold protection method based on BP neural network is adopted, which dynamically adjusts the protection threshold by combining load rate and temperature, and achieves millisecond-level response by deploying a fault prediction model locally. A hierarchical response mechanism is used for protection.
It enables real-time dynamic adjustment of protection thresholds, reduces false alarm rate, improves the reliability and security of power systems, is suitable for complex operating conditions and weak network scenarios, and has the ability to intervene in early faults.
Smart Images
Figure CN121749078A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of fault prediction technology, and in particular relates to an adaptive threshold protection method and system based on a measurement switch. Background Technology
[0002] Traditional measuring switches primarily rely on preset fixed thresholds to achieve overvoltage, undervoltage, overload, and short-circuit protection functions. Their core logic is that when real-time monitored parameters such as current and voltage exceed the set threshold, the circuit breaker is triggered via electromagnetic or thermal tripping mechanisms. However, although these switches integrate sensors, microprocessors, and communication modules, possessing data acquisition and remote control capabilities, their protection logic still primarily relies on static threshold judgment, essentially operating in a passive response mode. When a system fault occurs, the switch can only execute protection actions based on preset conditions, lacking the ability to identify fault precursors, resulting in a protection response lagging behind fault development. For example, in scenarios of sudden load changes or temperature fluctuations, the fixed threshold may deviate from the actual operating conditions, easily leading to false tripping or failure to trip, thereby causing equipment damage or safety accidents. Furthermore, the threshold settings of traditional switches are usually based on laboratory nominal operating conditions, making it difficult to adapt to complex and changing actual operating environments. Their protection accuracy and adaptability can no longer meet the demands of modern power systems for high reliability and dynamic adjustment capabilities.
[0003] To improve protection performance, existing technologies attempt to increase the number of sensors and upload collected data to the cloud for complex analysis using cloud computing resources. However, some solutions use LSTM neural networks to perform time-series predictions for single fault types, training models with historical data to generate predicted values and comparing them with fixed thresholds to trigger protection actions. However, cloud processing relies on network communication, which may cause delays in protection actions due to network latency, or even completely lose predictive capabilities when communication is interrupted, degenerating into traditional threshold protection. Secondly, existing models are mostly designed for single fault types and lack the ability to predict multiple faults collaboratively, making it difficult to cope with complex coupled fault scenarios in power systems. Summary of the Invention
[0004] To address the problems existing in the prior art, the present invention provides an adaptive threshold protection method based on a measurement switch, comprising the following steps: Step S100: Collect real-time operating data from the load side of the power system. The real-time operating data includes voltage, current, power factor, temperature, and duration. Step S200: Calculate the load factor of the power system based on the voltage, current, and power factor; Step S300: Input the load rate and temperature into the threshold prediction model and output the protection thresholds, which include overvoltage protection threshold, undervoltage protection threshold, overload protection threshold and short circuit protection threshold; input the real-time operating data into the fault prediction model and output the fault probability. Step S400: Set a dynamic threshold based on the protection threshold, and perform protection actions according to the fault probability and the dynamic threshold.
[0005] Based on the above scheme, the load rate and temperature are used as input data for the threshold prediction model, and the overvoltage protection threshold, undervoltage protection threshold, overload protection threshold and short circuit protection threshold are used as output data to train the threshold prediction model.
[0006] Based on the above scheme, step S400 specifically includes: S401: Compare the fault probability with the preset probability; S402: If the fault probability is greater than the preset probability, a graded response mechanism is adopted for protection based on the real-time data and dynamic threshold.
[0007] Furthermore, the graded response in S402 specifically includes: (1) When the real-time data reaches 80% of the dynamic threshold, an alarm signal is triggered and the alarm signal is uploaded to the cloud platform; (2) When the real-time data reaches 95% of the dynamic threshold, disconnect the non-critical loads in the power system; (3) When the real-time data reaches the dynamic threshold, the circuit breaker is tripped and fault location information is sent to the operation and maintenance terminal simultaneously.
[0008] According to one embodiment, the fault prediction model is based on a BP neural network model, which includes an input layer, a hidden layer and an output layer. The input layer includes 10 nodes, corresponding to voltage, current, temperature, load rate and ambient temperature and humidity, respectively. The output layer includes 4 nodes, corresponding to overvoltage, undervoltage, overload and short circuit fault probabilities, respectively. The fault prediction model is obtained by training the BP neural network model based on historical operating data.
[0009] According to one embodiment, the threshold prediction model is established based on a BP neural network model, which includes an input layer, a hidden layer, and an output layer. The input layer includes two nodes, corresponding to the load rate and temperature, respectively, and the output layer includes four nodes, corresponding to overvoltage, undervoltage, overload, and short-circuit protection thresholds, respectively. The BP neural network model employs an adaptive learning mechanism to dynamically adjust parameters, specifically including: The system automatically adjusts the overvoltage, undervoltage, overload, and short-circuit protection thresholds based on temperature changes; and automatically adjusts the allowable duration of overload, overvoltage, and undervoltage protection based on load rate changes.
[0010] In this embodiment, the load factor is the ratio of the actual load power to the rated load power, and the actual load power is calculated based on real-time voltage, current, and power factor.
[0011] On the other hand, the present invention provides an adaptive threshold protection system based on a measurement switch for implementing the adaptive threshold protection method described above, the system comprising: The data acquisition module is used to collect real-time operating data from the load side of the power system. The real-time operating data includes voltage, current, power factor, temperature, and duration. A load factor calculation module is used to calculate the load factor based on the voltage, current, and power factor. The prediction module includes a threshold prediction submodule and a fault prediction submodule, wherein: The threshold prediction submodule is used to input the load rate and temperature into the threshold prediction model and output overvoltage, undervoltage, overload and short circuit protection thresholds. The fault prediction submodule is used to input the real-time operating data into the fault prediction model and output the probabilities of overvoltage, undervoltage, overload and short circuit faults. The protection response module is used to set a dynamic threshold based on the protection threshold output by the threshold prediction submodule, and to perform corresponding protection actions in combination with the fault probability output by the fault prediction submodule.
[0012] Preferably, the system includes a measurement switch with a built-in processor, the measurement switch with a built-in processor includes a prediction module and a protection response module, and the data acquisition module includes a voltage transformer, a current transformer and a temperature sensor.
[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. This application introduces a threshold prediction model based on BP neural network, which uses temperature and load rate as key input variables, so that the four types of protection thresholds of overvoltage, undervoltage, overload and short circuit can be dynamically adjusted in real time according to the actual operating status of the equipment and environmental conditions, thus overcoming the problem that traditional fixed thresholds are prone to false tripping or failure to trip under complex working conditions. 2. By combining the fault probability and dynamic threshold output by the fault prediction model, a dual judgment mechanism of "probability priority + threshold verification" is adopted to effectively distinguish between normal fluctuations and real faults, and significantly reduce the false alarm rate; at the same time, a graded early warning is triggered before the fault occurs, realizing early intervention in power system faults. 3. Deploying the model on the built-in processor of the measurement switch enables millisecond-level local response, without relying on cloud communication, effectively avoiding the risk of protection failure caused by network latency and interruption, suitable for weak network or no network scenarios, and improving the overall reliability and security of the system; 4. The BP neural network model, trained based on a large amount of historical operating data, can accurately capture the nonlinear coupling relationship between multi-dimensional variables such as voltage, current, temperature, and load rate. It is suitable for measurement switch products of different models and application scenarios. The model can update parameters through regular online learning to continuously adapt to long-term factors such as equipment performance degradation and changes in environmental temperature and humidity. Attached Figure Description
[0014] Figure 1 This is a flowchart illustrating the overall process of the method described in this application. Detailed Implementation
[0015] The invention will be further described below with reference to specific embodiments.
[0016] like Figure 1 As shown, this embodiment provides an adaptive threshold protection method based on a measurement switch, including the following steps: Step S100: Collect real-time operating data from the load side of the power system. The real-time operating data includes at least voltage, current, power factor, temperature, and duration. In this embodiment, step S100 specifically includes: S101 measures the voltage and current on the load side of the power system and the temperature of the switching equipment through voltage transformers, current transformers and temperature sensors. S102 measures ambient temperature and humidity using an ambient temperature and humidity sensor, and obtains power factor and duration using a measuring switch.
[0017] Step S200: Calculate the load factor of the power system based on the real-time acquired voltage, current and power factor; In this embodiment, step S200 includes: S201: Calculate the actual load power based on the collected real-time voltage, current and power factor of the load side; It should be noted that in S201, the voltage and current on the load side are acquired through a measuring switch, which can calculate the power factor and duration.
[0018] S202: The load factor is obtained based on the ratio of the actual load power to the rated load power. The rated power is the design power of the load equipment under standard operating conditions. S203 preprocesses real-time voltage, current, temperature, ambient temperature and humidity, and load rate, including: A wavelet threshold denoising algorithm is used to process real-time running data to eliminate high-frequency noise and retain effective signal features; The real-time running data is then normalized to eliminate the influence of the units in the real-time running data, providing a good data foundation for subsequent input into the model.
[0019] As those skilled in the art will know, in actual operation, measuring switches may experience overvoltage, undervoltage, overload, and short-circuit faults. The load side is the final point of failure, therefore, the direct manifestation of faults such as overvoltage, undervoltage, overload, and short circuit is the abnormal operating state of the load equipment. Based on the real-time operating data of the power system load side, it is possible to analyze whether the measuring switch equipment has malfunctioned.
[0020] According to this embodiment, after preprocessing the real-time operating data in step S200, the preprocessed real-time operating data is input into the prediction model in step S300. The prediction model includes a fault prediction model and a threshold prediction model. S300 specifically includes: S310: Input the preprocessed real-time operating data into the fault prediction model and output the fault probability, which includes overvoltage probability, undervoltage probability, overload probability and short circuit probability.
[0021] S310 specifically includes: S311: Construct a BP neural network model; In this embodiment, the fault prediction model is based on a BP neural network model, which includes an input layer, a hidden layer, and an output layer. The input layer includes 10 nodes, which correspond to data such as three-phase voltage, three-phase current, temperature, load rate, and ambient temperature and humidity. A dual-hidden-layer structure is adopted, in which the first hidden layer contains 32 nodes and the second hidden layer contains 16 nodes; The output layer includes four nodes, corresponding to the probabilities of overvoltage, undervoltage, overload, and short circuit, respectively.
[0022] The hidden layer in this embodiment ensures that the model has sufficient complexity to capture the nonlinear relationships in the data, while avoiding the risk of overfitting.
[0023] S312: After constructing the BP neural network model through the above network structure, the BP neural network model is trained based on historical operating data to obtain the fault prediction model.
[0024] The historical training data in this embodiment includes data such as voltage, current, temperature, load rate, ambient temperature and humidity, and probability. The probability and threshold are based on the collected real-time data and obtained through experimental detection. The training data obtained through experimental detection is used to train the fault prediction model to obtain a fault prediction model with the required prediction accuracy, which is then used for subsequent real-time prediction so that the power system can respond in a timely manner.
[0025] S313: Input the real-time acquired and preprocessed voltage, current, temperature, load rate and ambient temperature and humidity into the fault prediction model to predict the current overvoltage probability, undervoltage probability, overload probability and short circuit probability.
[0026] Since humidity-induced equipment failures can lead to abnormal fluctuations in voltage and current, and may exacerbate instability when combined with other factors, humidity is used as input data for the model. By training the model, the influence of humidity on the output results can be captured, thereby improving the reliability of the model.
[0027] S320: Input the load rate and temperature into the threshold prediction model and output protection thresholds, which include overvoltage protection threshold, undervoltage protection threshold, overload protection threshold and short circuit protection threshold.
[0028] In power systems, initial protection thresholds for overvoltage, undervoltage, overload, and short circuit are first set based on the rated voltage and current parameters of the equipment to ensure that the measuring switchgear does not falsely trigger protection under normal operating conditions. This embodiment considers that load rate and temperature affect the occurrence of faults; that is, protection actions are triggered by changing the thresholds based on changes in load rate and temperature. To reflect the influence of load rate and temperature on the thresholds, this principle is applied in the model's training data. The overvoltage, undervoltage, overload, and short circuit protection thresholds are automatically adjusted according to temperature changes; the allowable duration of overload, overvoltage, and undervoltage protection is automatically adjusted according to changes in load rate. The specific principle and operation are as follows: Increased temperature accelerates the aging of insulation materials and reduces the overvoltage tolerance of load equipment. Therefore, the overvoltage protection threshold needs to be lowered when the temperature rises and relaxed when the temperature falls. Under high load rates, equipment heats up faster, so the allowable duration of overvoltage or undervoltage needs to be shortened. When the temperature rises, the heat dissipation capacity of the equipment decreases, and the allowable overload current needs to be reduced; under high load rate, the equipment is close to full load, and the allowable overload duration needs to be shortened. As temperature rises, the equipment's ability to withstand short-circuit impacts decreases, necessitating a reduction in the short-circuit protection threshold.
[0029] According to one implementation of this embodiment, the threshold prediction model is established based on a BP neural network model, which includes an input layer, a hidden layer and an output layer, wherein: the input layer includes 2 nodes, corresponding to the load rate and temperature respectively, and the output layer includes 4 nodes, corresponding to the overvoltage, undervoltage, overload and short circuit protection thresholds respectively. The BP neural network model employs an adaptive learning mechanism to dynamically adjust parameters, specifically including: The system automatically adjusts the overvoltage, undervoltage, overload, and short-circuit protection thresholds based on temperature changes; and automatically adjusts the allowable duration of overload, overvoltage, and undervoltage protection based on load rate changes.
[0030] Based on historical data, the load rate and temperature are used as input data for the threshold prediction model, and the overvoltage protection threshold, undervoltage protection threshold, overload protection threshold, and short circuit protection threshold are used as output data. The thresholds are obtained through experimental detection based on the collected real-time data. The threshold prediction model is trained using historical data, and then the overvoltage, undervoltage, overload, and short circuit protection thresholds are predicted based on the load rate and temperature using this threshold prediction model.
[0031] Training data is obtained through experiments to train the fault prediction model, resulting in a fault prediction model with the required prediction accuracy. This model is then used for subsequent real-time predictions so that the power system can respond promptly.
[0032] In this embodiment, the performance of the fault prediction model is evaluated by the root mean square error (RMSE) and the coefficient of determination (CCD). The RMSE measures the average deviation between the predicted value and the actual value output by the prediction model, while the CCD reflects the prediction model's ability to capture data.
[0033] Furthermore, this embodiment also optimizes the prediction model through K-fold cross-validation. Based on the validation results, the hyperparameters such as the number of hidden layer nodes and the learning rate in the model structure are adjusted to optimize the performance of the prediction model.
[0034] Based on the above steps, the fault probability and protection threshold are predicted according to the current power system operation data. Based on the fault probability and protection threshold, a timely response is made, namely: step S400, setting a dynamic threshold based on the protection threshold, and performing protection actions according to the fault probability and dynamic threshold.
[0035] In this embodiment, a protection threshold is output through a threshold prediction model, and a dynamic threshold is set based on a 95% confidence interval of the protection threshold, allowing the threshold to be dynamically adjusted as the load changes. The confidence interval covers the probability of the true threshold, adapting to dynamic loads; the 95% confidence level balances risk and engineering requirements, avoiding excessive conservatism or aggressiveness.
[0036] Step S400 specifically includes: S401: Compare the fault probability with the preset probability; S402: If the fault probability reaches the preset probability, a graded response mechanism is adopted for protection based on the real-time data and dynamic threshold.
[0037] The preset probability in S402 is set by a technician based on experience in the specific application of the method of this embodiment.
[0038] Specifically, the graded response in S402 includes: (1) Level 1 warning: When the real-time data reaches 80% of the dynamic threshold, an alarm signal is triggered and the alarm signal is uploaded to the cloud platform, and the local recording function is started at the same time; (2) Level 2 warning: When the real-time data reaches 95% of the dynamic threshold, the non-critical loads in the power system are disconnected; (3) Level 3 protection: When the real-time data reaches the dynamic threshold, the millisecond-level tripping operation is executed in combination with the duration. After the measurement switch trips, the address of the device, the time of occurrence, the voltage and current at the time of tripping, the reason for tripping, and the fault phase are recorded, and the fault location information is pushed to the operation and maintenance personnel at the same time.
[0039] It should be noted that the fault prediction model and the threshold prediction model can make predictions based on the current state at the same time, and use the predicted data to determine whether a fault has occurred, and execute a graded response mechanism to perform corresponding protection actions.
[0040] Duration is a factor used to determine whether to trip. After reaching a preset probability and dynamic threshold, the corresponding level 3 protection is triggered when the duration also meets the duration threshold. For example, if the overvoltage protection threshold is 300V and the duration threshold is 10s, if the voltage exceeds 300V for 10s, tripping is performed; if the voltage returns to 220V at the 5th second, tripping is not performed.
[0041] Example 2 This embodiment illustrates the process of threshold prediction model from data processing, model building, and training, specifically including: Step A1: Collect historical operating data from the load side and the measuring switchgear, including load rate, temperature and corresponding overvoltage threshold, undervoltage threshold, overload threshold and short circuit threshold; Step A2: Preprocess the historical operation data; A21: In order to eliminate the difference in dimensions, the load rate and temperature are normalized respectively. Specifically, the load rate (0%~100%) is mapped to [0,1], and the temperature (-40℃~125℃) is mapped to [-1,1]. A22: Perform data cleaning on load rate and temperature to remove outliers, specifically: remove data with a load rate >120% or a temperature exceeding the temperature sensor's range; A23: To ensure the model's generalization ability, the data is divided into training set, validation set, and test set in a 7:2:1 ratio.
[0042] Step A3: Construct and initialize the BP neural network model; A31: Construct a three-layer BP neural network model, including an input layer, a hidden layer, and an output layer, wherein: the input layer includes 2 nodes, corresponding to the load rate and temperature respectively; the output layer includes 4 nodes, corresponding to the overvoltage, undervoltage, overload, and short-circuit protection thresholds respectively. A32: Assign initial values to all connection weights and neuron biases in the neural network to avoid gradient vanishing or exploding; set the learning rate to 0.01~0.1 to control the step size of parameter updates; use mean squared error (MSE) to measure the deviation between the predicted threshold and the true threshold.
[0043] Step A4: Train the BP neural network model to obtain the threshold prediction model; Each batch of normalized inputs (load rate, temperature) in the training set is fed into the network, and the four types of threshold prediction values of the output layer are calculated. Calculate the error between the predicted value and the true label based on the MSE loss function; Using gradient descent, the gradient of the loss with respect to the weights and biases of each layer is calculated, and the network parameters are updated. Use industry standards as constraints for model training, and repeat forward-backward propagation training until the loss function converges. If the validation set loss does not decrease for 10 consecutive epochs, training is terminated early to prevent overfitting. The trained model is evaluated using a test set. The threshold prediction model that meets the criteria is then deployed to the embedded processor of the measurement switch to generate dynamic protection thresholds.
[0044] On the other hand, based on the same technical concept, the present invention also provides an adaptive threshold protection system based on a measuring switch, for implementing the adaptive threshold protection method as described above, the system comprising: The data acquisition module is used to collect real-time operating data from the load side of the power system. The real-time operating data includes voltage, current, power factor, temperature, and duration. A load factor calculation module is used to calculate the load factor based on the voltage, current, and power factor. The prediction module includes a threshold prediction submodule and a fault prediction submodule, wherein: The threshold prediction submodule is used to input the load rate and temperature into the threshold prediction model and output overvoltage, undervoltage, overload and short circuit protection thresholds. The fault prediction submodule is used to input the real-time operating data into the fault prediction model and output the probabilities of overvoltage, undervoltage, overload and short circuit faults. The protection response module is used to set a dynamic threshold based on the protection threshold output by the threshold prediction submodule, and to perform corresponding protection actions in combination with the fault probability output by the fault prediction submodule.
[0045] Preferably, the system includes a measurement switch with a built-in processor, the measurement switch with a built-in processor includes a prediction module and a protection response module, and the data acquisition module includes a voltage transformer, a current transformer and a temperature sensor.
[0046] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
[0047] While the specific embodiments of the present invention have been described above, they are not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.
Claims
1. An adaptive threshold protection method based on a measuring switch, characterized in that, Includes the following steps: Step S100: Collect real-time operating data from the load side of the power system. The real-time operating data includes voltage, current, power factor, temperature, and duration. Step S200: Calculate the load factor of the power system based on the voltage, current, and power factor; Step S300: Input the load rate and temperature into the threshold prediction model and output the protection thresholds, which include overvoltage protection threshold, undervoltage protection threshold, overload protection threshold and short circuit protection threshold; input the real-time operating data into the fault prediction model and output the fault probability. Step S400: Set a dynamic threshold based on the protection threshold, and perform protection actions according to the fault probability and the dynamic threshold.
2. The adaptive threshold protection method based on a measuring switch according to claim 1, characterized in that, The load rate and temperature are used as input data for the threshold prediction model, and the overvoltage protection threshold, undervoltage protection threshold, overload protection threshold, and short circuit protection threshold are used as output data to train the threshold prediction model.
3. The adaptive threshold protection method based on a measuring switch according to claim 1, characterized in that, Step S400 specifically includes: S401: Compare the fault probability with the preset probability; S402: If the fault probability is greater than the preset probability, a graded response mechanism is adopted for protection based on the real-time data and dynamic threshold.
4. The adaptive threshold protection method based on a measuring switch according to claim 3, characterized in that, The graded response in S402 specifically includes: (1) When the real-time data reaches 80% of the dynamic threshold, an alarm signal is triggered and the alarm signal is uploaded to the cloud platform; (2) When the real-time data reaches 95% of the dynamic threshold, disconnect the non-critical loads in the power system; (3) When the real-time data reaches the dynamic threshold, the circuit breaker is tripped based on the duration, and fault location information is sent to the operation and maintenance terminal simultaneously.
5. The adaptive threshold protection method based on a measuring switch according to claim 2, characterized in that, The fault prediction model is based on a BP neural network model, which includes an input layer, a hidden layer, and an output layer. The input layer includes 10 nodes, corresponding to voltage, current, temperature, load rate, and ambient temperature and humidity, respectively. The output layer includes 4 nodes, corresponding to overvoltage, undervoltage, overload, and short-circuit fault probabilities, respectively. The fault prediction model is obtained by training the BP neural network model based on historical operating data.
6. The adaptive threshold protection method based on a measuring switch according to claim 1, characterized in that, The threshold prediction model is based on a BP neural network model, which includes an input layer, a hidden layer, and an output layer. The input layer has two nodes, corresponding to the load rate and temperature, respectively, and the output layer has four nodes, corresponding to the overvoltage, undervoltage, overload, and short-circuit protection thresholds, respectively. The BP neural network model employs an adaptive learning mechanism to dynamically adjust parameters, specifically including: The system automatically adjusts the overvoltage, undervoltage, overload, and short-circuit protection thresholds based on temperature changes; and automatically adjusts the allowable duration of overload, overvoltage, and undervoltage protection based on load rate changes.
7. The adaptive threshold protection method based on a measuring switch according to claim 2, characterized in that, The load factor is the ratio of the actual load power to the rated load power, and the actual load power is calculated based on real-time voltage, current, and power factor.
8. An adaptive threshold protection system based on a measuring switch, characterized in that, For implementing the adaptive threshold protection method as described in any one of claims 1-7, the system comprises: The data acquisition module is used to collect real-time operating data from the load side of the power system. The real-time operating data includes voltage, current, power factor, temperature, and duration. A load factor calculation module is used to calculate the load factor based on the voltage, current, and power factor. The prediction module includes a threshold prediction submodule and a fault prediction submodule, wherein: The threshold prediction submodule is used to input the load rate and temperature into the threshold prediction model and output overvoltage, undervoltage, overload and short circuit protection thresholds. The fault prediction submodule is used to input the real-time operating data into the fault prediction model and output the probabilities of overvoltage, undervoltage, overload and short circuit faults. The protection response module is used to set a dynamic threshold based on the protection threshold output by the threshold prediction submodule, and to perform corresponding protection actions in combination with the fault probability output by the fault prediction submodule.
9. The adaptive threshold protection system based on a measuring switch according to claim 8, characterized in that, The system includes a measurement switch with a built-in processor, which includes a prediction module and a protection response module. The data acquisition module includes a voltage transformer, a current transformer, and a temperature sensor.