Soft starter control method and system based on dynamic threshold correction and current change prediction

Abstract

The invention belongs to the technical field of soft starter control, and particularly provides a soft starter control method and system based on dynamic threshold correction and current change prediction, and the method comprises the steps: data collection: collecting the collected data, including the environment temperature, the input current collection and the operation duration of a soft starter; data cleaning and preprocessing; establishing a temperature-time-current threshold correction relation; a soft starter state classification model is constructed based on the long-short-term memory network and used for perfecting the temperature-time-current threshold correction relation constructed in the third step and outputting the soft starter state; soft starter state classification model training optimization and evaluation; and using the trained soft starter state classification model to carry out soft starter operation state discrimination. The method and the system are combined with multi-dimensional data such as temperature and time, a current protection threshold is dynamically adjusted, and the problem that a traditional static constant value is poor in adaptability is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of soft starter control, and in particular, relates to a soft starter control method and system based on dynamic threshold correction and current change prediction. Background Art

[0002] A soft starter is a novel motor control device that integrates soft starting, soft stopping, light-load energy saving, and various protection functions. It is often used for current-limited starting in production. During operation, the output voltage of the thyristor voltage regulator gradually increases, causing the motor to accelerate until the thyristor fully conducts and the motor operates at its rated voltage. This limits the motor's starting current, achieving a smooth start, thereby reducing current surges and preventing overcurrent tripping.

[0003] In actual production, soft starters are often used to control the startup of pumps. Depending on the input voltage and rated voltage, the current passing through the soft starter will also vary accordingly during the different stages of the pump startup. Currently, a fixed value is used to monitor the current during the soft starter's startup state. That is, if the current exceeds the fixed value for a period of time during the startup process, it is considered an overload, and the soft starter automatically powers off and stops the startup process. The rated current of a soft starter varies with environmental factors such as temperature, and the fixed value is manually set based on historical experience. Therefore, a uniform fixed value is not universally applicable in different environments and conditions. As a result, the actual current value of the soft starter often slightly exceeds the fixed value, which does not affect actual operation, but the soft starter settings will trigger an overload alarm and stop the operation.

[0004] As the soft starter gradually increases voltage, it incorporates internal overload and underload detection devices, responding to current fluctuations exceeding the set upper and lower limits. However, existing soft starters are limited to threshold detection. They lack monitoring for sudden, high-amplitude current fluctuations that do not exceed the detection threshold, making it impossible to identify extreme jumps. While some soft starter models include an acceleration ramp time for the applied torque, this is only specified in the program's operating logic, without corresponding monitoring logic, resulting in no fault detection. Furthermore, at the production site, the PLC lacks the logic to properly assess current fluctuations. Therefore, when an internal soft starter anomaly causes a sudden increase in voltage, but the current does not reach the detection threshold, the monitoring system cannot respond promptly and correctly, potentially damaging the motor.

[0005] The existing soft starter control method has the following technical problems: 1. The artificially set value relies entirely on experience and ignores the impact of external factors such as temperature on the threshold. It lacks accuracy, adaptability, and rigor. 2. Once the thyristor inside the soft starter fails, the voltage acting on the pump motor may jump, causing damage to the pump; 3. For gradually increasing current values, existing technologies do not effectively monitor and limit the overall curve of current value changes over time. For extreme jump conditions that do not reach the monitoring threshold, the soft starter may not be able to effectively identify them. Similarly, due to the low precision limitations of manually set thresholds, the soft starter may stop running due to slight fluctuations. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a soft starter control method and system based on dynamic threshold correction and current change prediction, combining multi-dimensional data such as temperature and time to dynamically adjust the current protection threshold and solve the problem of poor adaptability of traditional static set values.

[0007] To solve the above technical problems, the technical solution adopted by the present invention is: a soft starter control method based on dynamic threshold correction and current change prediction, comprising the following steps: Step 1: Data collection: The collected data includes ambient temperature, input current, and soft starter operation time; Step 2: Data cleaning and preprocessing: including denoising and normalization, as well as sequence filling and truncation; Step 3: Establishing a temperature-time-current threshold correction relationship; Step 4: Build a soft starter state classification model based on the long short-term memory network to improve the temperature-time-current threshold correction relationship built in step 3 and output the soft starter state; Step 5: Training, optimization and evaluation of the soft starter state classification model; Step 6: Use the trained soft starter status classification model to judge the soft starter operating status.

[0008] In a preferred solution, in step 2, a median filtering algorithm is used to perform denoising on the collected current data to remove abnormal values ​​and noise interference.

[0009] In a preferred solution, in step 2, the normalization process is to normalize the ambient temperature and current data to the range of [0, 1], and the expression is: (1); Among them, x is the original data, is the minimum value of the data, is the maximum value of the data, The data are normalized.

[0010] In a preferred solution, in step 2, sequence padding and truncation are performed by using the pad_sequences() function to unify sequences of different step lengths into the same step length by using zero padding and data truncation.

[0011] In a preferred solution, in step 3, a temperature-time-current threshold correction relationship is established, and the expression is: (2); in, is the corrected threshold, is the basic threshold, is the temperature compensation coefficient, T is the current ambient temperature, is the standard operating temperature, is the time decay factor, is the rate of change compensation weight.

[0012] In a preferred solution, the long short-term memory network of the soft starter state classification model includes an input layer, an LSTM layer, and a fully connected layer, which are as follows: Input layer: The LSTM input data is constructed as a three-dimensional NumPy array with a shape of (N, P, 3(I, T, X)), where N is the number of valid samples, P is the time step, 3(I, T, X) represents the current characteristic value within a time step, I is the current value, T is the temperature value, and X is the operating status of the soft starter. LSTM layer: The input NumPy array is fed into the LSTM layer. The LSTM layer consists of two layers. In the first layer, samples where X's operating status is in the normal startup phase and normal operation phase are marked as True, and samples where X's operating status is in the soft starter overload phase and soft starter underload phase are marked as False. The output of the second LSTM layer is the temperature-time-current prediction curve for the next 20 seconds. This curve is a three-dimensional NumPy array with a shape of (N, P, 3(I, T, X)). Fully connected layer: The vector group obtained by the LSTM layer and the NumPy array are input into the fully connected layer as the output of the linear regression task; Use the linear activation function to retain the features of the input vector group and map it to the output end to obtain the output label prediction result. The output form for each group of samples is A 2D array of vectors and (P, 3(I, T, X)) where is the temperature compensation coefficient, is the time decay factor, is the rate of change compensation weight, P is the time step, I is the current value, T is the temperature value, and X is the operating status of the soft starter.

[0013] In a preferred solution, the output of the soft starter state classification model is 、 、 Input the temperature-time-current threshold correction relationship constructed in step 3 to improve the temperature-time-current threshold correction relationship.

[0014] In a preferred solution, in step five, an Adam optimizer is used, and real historical operating data is used as a training set to optimize the soft starter state classification model parameters through a back propagation algorithm.

[0015] In a preferred solution, in step 5, the mean square error of the model current prediction value and the soft starter state discrimination accuracy are used to evaluate the model prediction accuracy. The mean square error of the model current prediction value is calculated as follows: (3); Use the following formula to calculate the soft starter status judgment accuracy: (4); Where P is the time step, is the predicted current value, is the true value of the current, The number of steps for the soft starter to predict the state and the actual state.

[0016] The present invention also provides a soft starter intelligent protection system for executing the soft starter control method based on dynamic threshold correction and current change prediction, comprising: Data acquisition module: including temperature sensor and current detection module, used to collect ambient temperature, input current and running time data in real time; Data preprocessing module: cleaning and preprocessing of collected data; Dynamic threshold calculation module: establishes the temperature-time-current threshold correction relationship and outputs the dynamic current threshold; State classification and prediction module: uses the soft starter state classification model to classify and predict the current change trend and soft starter operating status; Intelligent decision-making module: Combines dynamic thresholds with prediction results to issue a stop command to the soft starter if it is judged to be overloaded / underloaded, and sends an alarm to the monitoring system.

[0017] The present invention provides a soft starter control method and system based on dynamic threshold correction and current change prediction, which has the following beneficial effects: 1. By combining multi-dimensional data such as temperature and time with an LSTM model, the current protection threshold is dynamically adjusted, addressing the poor adaptability of traditional static settings. The introduction of temperature compensation and time decay terms dynamically adjusts the current threshold based on changes in ambient temperature and operating time, improving protection accuracy and reliability.

[0018] 2. This system collects multi-dimensional data, including ambient temperature, current, and operating time, and uses an LSTM model to intelligently analyze current trends, enabling classification and prediction of the soft starter's operating status. Data preprocessing ensures the quality of the model input data, improving model training and prediction accuracy. The LSTM model captures the long-term dependencies and dynamic trends of current signals, addressing the inability of traditional technologies to detect sudden current changes and progressive faults.

[0019] 3. Through the synergistic effect of multi-dimensional perception, dynamic modeling and intelligent decision-making, the present invention systematically solves the three core problems of threshold rigidity, mutation omission and trend out of control in traditional technologies, significantly improving the safety and reliability of the soft start system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present invention will be further described below with reference to the accompanying drawings and examples: Figure 1 Flowchart of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0022] Example 1: like Figure 1 As shown, a soft starter control method based on dynamic threshold correction and current change prediction includes the following steps: Step 1: Data collection: The collected data includes ambient temperature, input current, and soft starter operation time.

[0023] High-precision temperature sensors and current detection modules are deployed on the soft starter body to collect real-time data on ambient temperature (T), input current (I), and operating time (t). The current sensor is used to obtain real-time current waveforms at a sampling rate of 50Hz ( ).

[0024] Step 2: Data cleaning and preprocessing: including denoising and normalization as well as sequence filling and truncation.

[0025] Denoising: Use the median filter algorithm to denoise the collected current data and remove outliers and noise interference.

[0026] Value filtering is an established technique, a signal processing method based on sorted statistics. Its core idea is to replace the central value with the median value within a data window, effectively eliminating noise. This method is particularly suitable for processing signals containing impulsive noise (such as salt and pepper noise). The steps to use the median filter algorithm to denoise the current data are as follows: 1) Determine an appropriate data window size (5 is acceptable).

[0027] 2) The sliding window moves point by point on the current data sequence.

[0028] 3) Sort the data in the window and take the median as the new data point.

[0029] 4) Replace the original value of the window center point with the median value to complete the denoising process.

[0030] Normalization processing: Normalization processing is to normalize the ambient temperature and current data to the range of [0, 1]. The expression is: (1); Among them, x is the original data, is the minimum value of the data, is the maximum value of the data, The data are normalized.

[0031] Sequence padding and truncation: Sequence padding and truncation uses the pad_sequences() function to unify sequences of different step lengths into the same step length by zero padding and data truncation to facilitate subsequent model training.

[0032] Step 3: Establish a temperature-time-current threshold correction relationship.

[0033] The temperature-time-current threshold correction relationship is expressed as: (2); in, is the corrected threshold, is the basic threshold, is the temperature compensation coefficient, T is the current ambient temperature, is the standard operating temperature, is the time decay factor, is the rate of change compensation weight.

[0034] is the temperature compensation term. When the ambient temperature When the threshold Increases, allowing greater current to offset the increase in conductor resistance caused by high temperature; when When the voltage is too low, the threshold is appropriately lowered to avoid the risk of overcurrent caused by embrittlement of the insulation material in a low temperature environment.

[0035] is the time decay term, at the initial startup t ≈0, ≈1, at this time the threshold is mainly affected by temperature; and after long-term operation, the threshold gradually decreases to prevent cumulative overload caused by continuous heating.

[0036] Step 4: Build a soft starter status classification model based on the long short-term memory network (LSTM) to improve the temperature-time-current threshold correction relationship built in step 3 and output the soft starter status.

[0037] The long short-term memory network of the soft starter state classification model includes an input layer, an LSTM layer, and a fully connected layer, which are as follows: Input layer: LSTM processes sequential data, so the input data is in the form of time steps. The LSTM input data is constructed as a three-dimensional NumPy array with the shape: (N, P, 3(I, T, X)). N is the number of valid samples; P represents the time step, which is 1000 here. Since the current acquisition frequency is 50 Hz, a time step of 1000 represents the current trend over 20 seconds, that is, 1000 consecutive current eigenvalues. 3(I, T, X) represents the current eigenvalue within a time step, where I is the current value and T is the temperature value. Both I and T are filtered and normalized data. X represents the operating status of the soft starter, in the form of a one-hot-encoded binary column. This encoding method effectively represents the mutual exclusivity between categories and converts each sample category into an independent binary vector, avoiding issues of priority or order between categories. The specific correspondence is shown in Table 1.

[0038]

[0039] LSTM layer: Input the input layer NumPy array into the LSTM layer. There are two LSTM layers. In the first layer, samples with X (1, ​​0, 0, 0) and (0, 1, 0, 0) will be marked as True, and samples with X (0, 0, 1, 0) and (0, 0, 0, 1) will be marked as False, so as to find the correct temperature-time-current threshold correction relationship during training. Combined with formula (1.2), the output form is A vector group of . is the temperature compensation coefficient, is the time decay factor, The output of the second LSTM layer is the temperature-time-current prediction curve for the next 20 seconds. This curve is a three-dimensional NumPy array with the shape (N, P, 3(I, T, X)), where N is the number of valid samples; P is the time step, which is 1000 here; I is the current value; T is the temperature value; and X is the operating status of the soft starter.

[0040] Fully connected layer: The vector group obtained by the LSTM layer and the NumPy array are input into the fully connected layer as the output of the linear regression task; during training, the Dropout function is used to randomly select some neurons and set their outputs to 0, so that these neurons are ignored in this round of forward and backward propagation. This prevents overfitting of the neural network and improves the generalization ability. The linear activation function of formula (3) is used to retain the features of the input vector group and map it to the output end to obtain the output label prediction result. For each group of samples, the output format is A two-dimensional array of vectors + (P, 3(I, T, X)). is the temperature compensation coefficient, is the time decay factor, is the rate of change compensation weight, P is the time step, which is 1000 here, I is the current value, T is the temperature value, and X is the operating status of the soft starter.

[0041] (3).

[0042] The soft starter state classification model outputs 、 、 Input the temperature-time-current threshold correction relationship constructed in step 3 to improve the temperature-time-current threshold correction relationship.

[0043] Step 5: Training, optimization and evaluation of the soft starter state classification model.

[0044] We use real historical operating data as the training set and optimize model parameters through supervised learning. Using the Adam optimizer, we update the neural network parameters through the back-propagation algorithm during training to gradually improve the model's prediction accuracy.

[0045] The backpropagation algorithm uses the optim.Adam() function in Python to calculate the gradient of the loss function (MSE) with respect to the output layer. This algorithm propagates the error from the output layer back to the input layer layer by layer, thus guiding parameter updates. The Adam optimizer combines momentum with an adaptive learning rate to automatically adjust the learning rate, accelerating convergence and improving training stability.

[0046] The current waveform 20 seconds after the corresponding sample of the training set is used as the test set to verify the model performance. The mean square error of the model current prediction value and the accuracy of soft starter state judgment are used to evaluate the model prediction accuracy to ensure the reliability of the dynamic threshold.

[0047] The formula for calculating the mean square error of the model current prediction value is as follows: (4); Use the following formula to calculate the soft starter status judgment accuracy: (5); Where P is the time step, is the predicted current value, is the true value of the current, The number of steps for the soft starter to predict the state and the actual state.

[0048] Step 6: Use the trained soft starter status classification model to judge the soft starter operating status.

[0049] The trained temperature-time-current threshold correction relationship formula is connected to the running soft starter PLC. A sliding window is used to predict the current change curve within 20 seconds in real time. The current threshold is dynamically updated according to formula (2) to determine the soft starter's operating status. If the soft starter's operating status is judged to be overloaded or underloaded in the future, a stop command is issued and an alarm is sent to the monitoring system.

[0050] Example 2: In order to verify the effect of the present invention, this embodiment uses a real historical operation data set, including the ambient temperature, input current and operation time data of the soft starter.

[0051] 1. The dataset contains the following: Ambient temperature (T): The ambient temperature collected during different time periods, in degrees Celsius (°C).

[0052] Input current (I): Real-time current waveform acquired at a 50 Hz sampling rate, in amperes (A).

[0053] Running time (t): The running time of the soft starter, in seconds (s).

[0054] The dataset is divided into a training set and a test set. The training set contains 10,000 samples and the test set contains 2,000 samples. The time step of each sample is 1,000, corresponding to a 20-second current change trend.

[0055] 2. Model parameter setting During model training, we used the following parameter settings: Optimizer: Adam optimizer with a learning rate of 0.001, momentum parameter beta1=0.9, and adaptive learning rate parameter beta2=0.999.

[0056] LSTM layer: Two layers of LSTM, the number of neurons in the first layer is 128, and the number of neurons in the second layer is 64.

[0057] Fully connected layer: output vector group and predicted current waveforms.

[0058] Number of training epochs: 100.

[0059] Batch size (batch_size): 32.

[0060] 3. Model training process The model training process is divided into the following steps: Data preprocessing: The median filter algorithm is used to denoise the current data.

[0061] The ambient temperature and current data are normalized to a range of [0, 1].

[0062] Use the pad_sequences() function to pad and truncate sequences of different step lengths to make them the same step length.

[0063] Model training: The preprocessed data is fed into the LSTM model.

[0064] The Adam optimizer is used to update the model parameters through the back-propagation algorithm to minimize the loss function.

[0065] During training, the Dropout function (dropout rate = 0.2) is used to prevent overfitting.

[0066] The mean squared error (MSE) is used as the loss function for the regression task.

[0067] The cross entropy loss function is used as the loss function for the classification task.

[0068] 4. Model Performance Evaluation The performance of the model is evaluated on the test set, and the results are as follows: The MSE of the test set is 0.005, indicating that the model has a high prediction accuracy for the current change trend.

[0069] The classification accuracy of the test set is 98.5%, indicating that the model can accurately identify the operating status of the soft starter.

[0070] The effects of dynamic threshold correction are as follows: Through the dynamic threshold correction coefficient (0.02, 0.007, 0.005) output by the model, the threshold of the soft starter under different ambient temperatures and operating times can be adaptively adjusted.

[0071] At high temperatures, the threshold increases, allowing for greater current to offset the increased conductor resistance caused by the high temperature.

[0072] In low temperature environments, the threshold is appropriately lowered to avoid overcurrent risks.

[0073] Through the above implementation cases, we can draw the following conclusions: High current prediction accuracy: The model has a high prediction accuracy for current change trends, with an MSE of only 0.005, which can provide a reliable basis for dynamic threshold correction.

[0074] High classification accuracy: The model has an accuracy rate of 98.5% in classifying the soft starter's operating status, and can effectively identify abnormal conditions such as overload and underload.

[0075] Dynamic threshold correction is effective: Through the dynamic threshold correction relationship output by the model, the soft starter can adaptively adjust the threshold according to the ambient temperature and operating time, improving the safety and reliability of operation.

[0076] Example 3: This embodiment provides a soft starter intelligent protection system for executing the soft starter control method based on dynamic threshold correction and current change prediction described in Embodiment 1, including: Data acquisition module: including temperature sensor and current detection module, used to collect ambient temperature, input current and running time data in real time; Data preprocessing module: cleaning and preprocessing of collected data; Dynamic threshold calculation module: establishes the temperature-time-current threshold correction relationship and outputs the dynamic current threshold; State classification and prediction module: uses the soft starter state classification model to classify and predict the current change trend and soft starter operating status; Intelligent decision-making module: Combines dynamic thresholds with prediction results to issue a stop command to the soft starter if it is judged to be overloaded / underloaded, and sends an alarm to the monitoring system.

[0077] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A soft starter control method based on dynamic threshold correction and current change prediction, characterized in that: The following steps are involved: Step 1: Data collection: The collected data includes ambient temperature, input current, and soft starter operation time; Step 2: Data cleaning and preprocessing: including denoising and normalization, as well as sequence filling and truncation; Step 3: Establishing a temperature-time-current threshold correction relationship; Step 4: Build a soft starter state classification model based on the long short-term memory network to improve the temperature-time-current threshold correction relationship built in step 3 and output the soft starter state; Step 5: Training, optimization and evaluation of the soft starter state classification model; Step 6: Use the trained soft starter status classification model to judge the soft starter operating status.

2. A soft starter control method based on dynamic threshold correction and current change prediction according to claim 1, characterized in that: In the second step, a median filter algorithm is used to perform denoising on the collected current data to remove abnormal values ​​and noise interference.

3. The soft starter control method based on dynamic threshold correction and current change prediction according to claim 1, characterized in that: In step 2, the normalization process is to normalize the ambient temperature and current data to the range of [0, 1], and the expression is: （1）； Among them, x is the original data, is the minimum value of the data, is the maximum value of the data, The data are normalized.

4. The soft starter control method based on dynamic threshold correction and current change prediction according to claim 1, characterized in that: In the step 2, sequence padding and truncation are performed by using the pad_sequences() function to unify sequences of different step lengths into the same step length by using zero padding and data truncation.

5. The soft starter control method based on dynamic threshold correction and current change prediction according to claim 1, characterized in that: In step 3, a temperature-time-current threshold correction relationship is established, and the expression is: （2）； in, is the corrected threshold, is the basic threshold, is the temperature compensation coefficient, T is the current ambient temperature, is the standard operating temperature, is the time decay factor, is the rate of change compensation weight.

6. The soft starter control method based on dynamic threshold correction and current change prediction according to claim 1, characterized in that: The long short-term memory network of the soft starter state classification model includes an input layer, an LSTM layer, and a fully connected layer, which are as follows: Input layer: The LSTM input data is constructed as a three-dimensional NumPy array with a shape of (N, P, 3(I, T, X)), where N is the number of valid samples, P is the time step, 3(I, T, X) represents the current characteristic value within a time step, I is the current value, T is the temperature value, and X is the operating status of the soft starter. LSTM layer: The input NumPy array is fed into the LSTM layer. The LSTM layer consists of two layers. In the first layer, samples where X's operating status is in the normal startup phase and normal operation phase are marked as True, and samples where X's operating status is in the soft starter overload phase and soft starter underload phase are marked as False. The output of the second LSTM layer is the temperature-time-current prediction curve for the next 20 seconds. This curve is a three-dimensional NumPy array with a shape of (N, P, 3(I, T, X)). Fully connected layer: The vector group obtained by the LSTM layer and the NumPy array are input into the fully connected layer as the output of the linear regression task; Use the linear activation function to retain the features of the input vector group and map it to the output end to obtain the output label prediction result. The output form for each group of samples is A 2D array of vectors and (P, 3(I, T, X)) where is the temperature compensation coefficient, is the time decay factor, is the rate of change compensation weight, P is the time step, I is the current value, T is the temperature value, and X is the operating status of the soft starter.

7. A soft starter control method based on dynamic threshold correction and current change prediction according to claim 6, characterized in that: The soft starter state classification model outputs 、 、 Input the temperature-time-current threshold correction relationship constructed in step 3 to improve the temperature-time-current threshold correction relationship.

8. The soft starter control method based on dynamic threshold correction and current change prediction according to claim 1, characterized in that: In step 5, the Adam optimizer is used, and the real historical operation data is used as a training set to optimize the soft starter state classification model parameters through the back propagation algorithm.

9. The soft starter control method based on dynamic threshold correction and current change prediction according to claim 1, characterized in that: In step 5, the mean square error of the model current prediction value and the soft starter state judgment accuracy are used to evaluate the model prediction accuracy. The mean square error of the model current prediction value is calculated as follows: （3）； Use the following formula to calculate the soft starter status judgment accuracy: （4）； Where P is the time step, is the predicted current value, is the true value of the current, The number of steps for the soft starter to predict the state and the actual state.

10. A soft starter intelligent protection system, characterized in that: The soft starter control method based on dynamic threshold correction and current change prediction for executing any one of claims 1 to 9 comprises: Data acquisition module: including temperature sensor and current detection module, used to collect ambient temperature, input current and running time data in real time; Data preprocessing module: cleaning and preprocessing of collected data; Dynamic threshold calculation module: establishes the temperature-time-current threshold correction relationship and outputs the dynamic current threshold; State classification and prediction module: uses the soft starter state classification model to classify and predict the current change trend and soft starter operating status; Intelligent decision-making module: Combines dynamic thresholds with prediction results to issue a stop command to the soft starter if it is judged to be overloaded / underloaded, and sends an alarm to the monitoring system.

Images