Speed regulation control method and system of high-voltage frequency conversion device

Abstract

The invention discloses a speed regulation control method and system for a high-voltage frequency conversion device, and relates to the field of motor control, and the method comprises the steps: obtaining historical operation data, and carrying out the preprocessing, and obtaining a corresponding historical time sequence data set; extracting fault feature factors, and calculating a historical abnormal feature value and an importance weight of each fault feature factor; constructing and training a risk prediction model to obtain a model parameter corresponding to each fault feature factor; obtaining current abnormal feature values corresponding to the fault feature factors at the current sudden stop moment of the motor, obtaining fault risk probabilities of the fault feature factors based on the model, and calculating an overall risk value of the motor at the current moment; and the restart power of the motor is calculated based on the overall risk value, and a corresponding restart control strategy is executed. According to the invention, functions of intelligent fault detection and adaptive speed regulation control on the motor can be realized, and safe and smooth restarting of the motor can be effectively guaranteed.

Description

Technical Field

[0001] This invention relates to the field of electric motor control, and in particular to a speed control method and system for a high-voltage frequency converter. Background Technology

[0002] In modern industrial power systems, high-voltage, high-power electric motors are the core driving force for many critical pieces of equipment, such as pumps, fans, compressors, and conveyor belts. High-voltage frequency converters, as the speed control hub for these motors, are widely used for speed control of high-power motors. Their working principle is to achieve smooth speed regulation by changing the motor's power supply frequency and voltage. Their core value lies in significant energy saving and consumption reduction, improved process control accuracy, and soft-start protection for both the motor and the power grid. In practical applications, electric motors and their drive systems operating in complex and harsh industrial environments may experience unplanned emergency stops due to various reasons such as power grid fluctuations, sudden load changes, internal insulation aging, mechanical jamming, and overheating. Emergency stops are not only a direct cause of production interruptions but can also be a symptom or trigger of a potentially serious malfunction.

[0003] To protect the motor and the power grid, high-voltage frequency converters need to promptly detect motor faults and ensure a smooth restart of the motor to its original operating state after troubleshooting. However, traditional high-voltage frequency converter control systems mostly rely on simple threshold comparisons for fault detection. Once a motor parameter exceeds the limit, the high-voltage frequency converter immediately triggers an alarm and controls the motor to perform a protective shutdown. The restart process often follows a fixed pattern, using a fixed restart power and speed. This traditional control method for high-voltage frequency converters is not intelligent enough and has significant limitations.

[0004] First, existing control methods for high-voltage frequency converters lack comprehensive judgment, focusing only on the instantaneous exceedance of single parameters, such as current, voltage, and winding temperature. They fail to comprehensively utilize multi-sensor information for correlation analysis and overall assessment of faults, easily leading to false alarms or missed alarms. Simultaneously, the restart process of existing control methods lacks intelligence. After an emergency stop of the motor, traditional systems typically only provide fault type alarms, such as overcurrent, overvoltage, overheating, and short circuit, but cannot quantify the fault risk level. Restart decisions heavily rely on human experience; operators must conduct on-site inspections and resets, often attempting restarts using fixed modes. If the underlying fault is not eliminated, it may lead to secondary impacts, fault expansion, or even permanent damage to the equipment. Furthermore, existing control methods lack sufficient diagnostic depth for motor faults; system alarm information is relatively general, making it difficult to accurately locate the root cause of the fault, resulting in time-consuming and labor-intensive troubleshooting, and prolonging downtime. Summary of the Invention

[0005] To achieve adaptive selection of motor restart strategies and address the shortcomings of traditional high-voltage frequency converter speed control methods, which lack intelligent fault detection and adaptive speed regulation capabilities, this invention provides a speed control method and system for high-voltage frequency converters, the technical solution of which is as follows: In a first aspect, the present invention provides a speed control method for a high-voltage variable frequency drive (VFD), comprising the following steps: sampling and acquiring historical operating data within a preset time period prior to the historical emergency stop of the motor and preprocessing it to obtain a corresponding historical time series dataset; extracting various fault characteristic factors related to motor faults, and calculating historical abnormal characteristic values ​​for each fault characteristic factor based on the historical time series dataset; calculating the importance weight of each fault characteristic factor based on the historical abnormal characteristic values; constructing a risk prediction model and training it based on the historical abnormal characteristic values ​​to obtain model parameters corresponding to each fault characteristic factor; acquiring the current abnormal characteristic values ​​corresponding to each fault characteristic factor at the current emergency stop of the motor, and obtaining the fault risk probability of each fault characteristic factor based on the risk prediction model and the corresponding model parameters; calculating the overall risk value of the motor at the current moment based on the fault risk probability and importance weight of each fault characteristic factor; calculating the restart power of the motor based on the overall risk value, and executing a corresponding restart control strategy.

[0006] Preferably, voltage sensors, current sensors, temperature sensors, and speed sensors installed on the motor are used to collect operating data sequences corresponding to each sensor within a preset time period before each emergency stop of the motor at the same sampling frequency. These data sequences include voltage value sequences, current value sequences, winding temperature sequences, and speed value sequences. All emergency stop times before the current emergency stop of the motor are taken as historical emergency stop times, and the motor operating data sequences corresponding to the historical emergency stop times are taken as historical data sequences. The historical data sequences corresponding to multiple historical emergency stop times, as well as the motor normal or motor fault judgment labels corresponding to each historical emergency stop time, are obtained. The mean interpolation method is used to fill in the missing values ​​in each historical data sequence. The timestamps of all sampling points in each historical data sequence are aligned to obtain the time series data corresponding to each historical data sequence. The collection of time series data corresponding to all historical emergency stop times is taken as the historical time series dataset.

[0007] Preferably, the rated operating parameters of the motor are obtained, including the rated voltage value, the rated current value, and the standard temperature range during normal operation of the motor. Based on the actual application scenario and historical fault types of the motor, various fault characteristic factors related to motor faults are extracted, including winding temperature characteristics, voltage deviation characteristics, current deviation characteristics, and load change characteristics. Data corresponding to each fault characteristic factor in the historical time series dataset is extracted in sequence. The historical time series dataset contains time series data of multiple historical emergency stop moments. The time series data corresponding to each historical emergency stop moment includes voltage value time series data, current value time series data, winding temperature time series data, and speed value time series data. Power value time series data is obtained based on the voltage value time series data and the current value time series data. Torque value time series data is obtained based on the power value time series data and the speed value time series data.

[0008] Preferably, based on the voltage value time series data corresponding to the voltage deviation characteristics at each historical emergency stop, the sum of the squares of the differences between the data values ​​of each sampling point and the rated voltage value within a certain voltage value time series is calculated, and the mean value is obtained based on the number of sampling points within the voltage value time series. The square root of the mean value is used as the historical abnormal feature value of the voltage deviation characteristics at the corresponding historical emergency stop, and then the historical abnormal feature value of the voltage deviation characteristics at each historical emergency stop is calculated sequentially. Similarly, based on the current value time series data corresponding to the current deviation characteristics at each historical emergency stop, the historical abnormal feature value of the current deviation characteristics at each historical emergency stop is obtained. Based on the winding temperature time series data corresponding to the winding temperature characteristics at each historical emergency stop, sampling points in the winding temperature time series where the data value exceeds the upper limit of the standard temperature range are extracted as winding temperature characteristic sample points at the corresponding historical emergency stop. The difference between the data value of each winding temperature characteristic sample point and the upper limit of the standard temperature range is calculated in turn. The difference between the upper and lower limits of the standard temperature range is used as the benchmark value. Based on the number of winding temperature characteristic sample points, the cumulative value of the ratio between the difference and the benchmark value is used as the historical abnormal characteristic value of the winding temperature characteristics at the corresponding historical emergency stop. Then, the historical abnormal characteristic value of the winding temperature characteristics at each historical emergency stop is calculated in turn. Based on the time series data of torque values ​​corresponding to the load mutation characteristics at each historical emergency stop, the absolute value of the difference between the current sampling point and the previous sampling point in a certain torque value time series is calculated sequentially to obtain the torque value absolute difference sequence. The overall standard deviation of all data in the torque value absolute difference sequence is used as the historical abnormal feature value of the load mutation characteristics at the corresponding historical emergency stop. Then, the historical abnormal feature value of the load mutation characteristics at each historical emergency stop is calculated sequentially.

[0009] Preferably, the judgment labels for whether the motor is normal or faulty are extracted at each historical emergency stop time. The historical abnormal feature values ​​of a certain fault feature factor at all historical emergency stop times are used as samples, and the samples are divided into normal sample set and fault sample set based on the judgment labels. The historical abnormal feature values ​​in the normal sample set and the fault sample set are sorted in ascending order. The overlapping interval between the normal sample set and the fault sample set is used as the transition interval of the fault feature factor. The maximum value of the historical abnormal feature value in the normal sample set is used as the upper limit of the transition interval, and the minimum value of the historical abnormal feature value in the fault sample set is used as the lower limit of the transition interval. The number of samples in the transition interval is used as the interval width. The total number of all samples of the fault feature factor is used as the total sample width. The ratio between the interval width and the total sample width is used as the fuzzy weight of the fault feature factor. Similarly, the transition interval and fuzzy weight of each fault feature factor are obtained. For a transition interval of a certain fault characteristic factor, samples with different data values ​​within the transition interval are sequentially used as judgment thresholds. Judgment conditions are set: when the historical abnormal characteristic value of the fault characteristic factor is greater than or equal to the judgment threshold, it is marked as a fault; when it is less than the judgment threshold, it is marked as normal. For a certain judgment threshold, based on the judgment conditions and the judgment labels corresponding to each historical emergency stop time, the ratio between the number of samples that meet the judgment conditions and the interval width is used as the accuracy rate of the judgment threshold. The accuracy rate of each judgment threshold is calculated sequentially, and the average of the judgment threshold accuracy rates is used as the reliability weight of the fault characteristic factor. Similarly, the reliability weight of each fault characteristic factor is obtained. The importance coefficients of fuzzy weight and reliability weight are set according to the actual application scenario and usage time of the motor. Based on the values ​​of the fuzzy weight and reliability weight of each fault characteristic factor, the importance weight of each fault characteristic factor is calculated sequentially.

[0010] Preferably, a risk prediction model is constructed using a BP neural network classification model and a sigmoid function is used as the activation function. The historical abnormal feature values ​​of each fault characteristic factor at all historical emergency stop times are sequentially input into the risk prediction model for training to obtain the model parameters corresponding to each fault characteristic factor.

[0011] Preferably, the operating data sequence of the motor at the current emergency stop time within a preset time period is acquired and preprocessed to obtain the current time series data. The data acquisition method and preprocessing process for the current emergency stop time are the same as those for historical operating data. Based on the current time series data, the same calculation process as for historical abnormal feature values ​​is used to obtain the current abnormal feature values ​​corresponding to each fault feature factor. Based on the risk prediction model and the model parameters corresponding to each fault feature factor, the current abnormal feature values ​​corresponding to each fault feature factor are input into the risk prediction model to obtain the fault risk probability of each fault feature factor.

[0012] Preferably, the sum of the importance weights of all fault characteristic factors is taken as the total weight value, and the ratio between the importance weight of a certain fault characteristic factor and the total weight value is taken as the normalized importance weight of that fault characteristic factor. The normalized importance weight of each fault characteristic factor is calculated in sequence. The sum of the products between the normalized importance weight of each fault characteristic factor and the corresponding fault risk probability is taken as the overall risk value of the motor at the current moment.

[0013] Preferably, the rated power of the motor is obtained, the overall risk value range is divided into three risk levels: low, medium, and high, and a piecewise calculation function for the motor restart power is set; for the low risk level, the rated power of the motor is... The restart power of the motor is used as the restart power; for medium-risk levels, an exponential function is used to reverse-map the overall risk value, and the product of the reverse-mapped value and the rated power of the motor is used as the restart power of the motor; for high-risk levels, the rated power of the motor is used as the restart power. The restart power of the motor is used as the starting power. Based on the overall risk value of the motor at the current moment, the restart power of the motor at the current moment is calculated and converted into the corresponding speed control command to control the motor to execute.

[0014] Secondly, the present invention provides a speed control system for a high-voltage frequency converter, used to implement the speed control method of the high-voltage frequency converter described above, comprising: a processor, a memory, a communication interface, and a data acquisition device. The processor stores computer program instructions for implementing the speed control method of the high-voltage frequency converter described above. The data acquisition device is a sensor installed on a motor. The communication interface is communicatively connected to the data acquisition device and the motor.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: Compared with traditional speed control methods for high-voltage frequency converters, the technical solution provided by this invention transcends the limitations of simple threshold protection. Through multi-sensor data fusion and intelligent algorithms, it possesses intelligent diagnostic and decision-making capabilities, enabling comprehensive detection and precise fault location guidance after a fault occurs. This strengthens proactive intervention in the motor, effectively shortening the fault investigation and repair time for on-site personnel. Simultaneously, it comprehensively evaluates multi-dimensional operating data during motor emergency stops, quantifies the probability of fault risk, and adaptively selects the safest and most efficient restart speed control strategy based on the real-time assessed fault risk level. This ensures a safe and smooth restart of the motor, reducing energy and equipment losses during motor startup. Consequently, it effectively improves the reliability and efficiency of the equipment's power system, maximizing production continuity, equipment safety, and operational economy. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the implementation of a speed control method for a high-voltage frequency converter according to an embodiment of the present invention. Figure 2 This is a structural block diagram of a speed control system for a high-voltage frequency converter according to an embodiment of the present invention. Detailed Implementation

[0017] The technical features of the present invention will be further described in detail below with reference to the accompanying drawings so that those skilled in the art can understand them.

[0018] A speed control method for a high-voltage frequency converter, the implementation process is as follows: Figure 1 As shown, the specific implementation steps are as follows: Step S1: Sample and obtain historical operating data within a preset time period before the historical emergency stop of the motor and preprocess it to obtain the corresponding historical time series dataset.

[0019] Specifically, by using voltage sensors, current sensors, temperature sensors, and speed sensors installed on the motor, the operating data sequence corresponding to each sensor within a preset time period before each emergency stop of the motor is collected at the same sampling frequency. This includes voltage value sequence, current value sequence, winding temperature sequence, and speed value sequence. All emergency stop times before the current emergency stop of the motor are taken as historical emergency stop times, and the motor operating data sequence corresponding to the historical emergency stop times is taken as historical data sequence. The process involves acquiring historical data sequences corresponding to multiple historical emergency stop moments, along with labels indicating whether the motor is functioning normally or malfunctioning for each historical emergency stop moment. Since raw data collected from industrial sites may contain missing values ​​or misaligned timestamps, preprocessing the collected data ensures data quality and improves the accuracy of subsequent calculations. The missing values ​​in each historical data sequence are filled using mean interpolation, and the timestamps of all sampling points in each historical data sequence are aligned to obtain the time series data corresponding to each historical data sequence. The collection of time series data corresponding to all historical emergency stop moments is then used as the historical time series dataset. In a preferred embodiment of the present invention, the preset time period is set to a range of 10 to 30 seconds, the sampling frequency is 10 Hz, the voltage sensor is used to collect the voltage value of the motor in real time, the current sensor is used to collect the current value of the motor in real time, the temperature sensor is used to collect the winding temperature of the motor in real time, and the speed sensor is used to collect the rotor speed of the motor in real time.

[0020] Step S2: Extract various fault characteristic factors related to motor faults, and calculate the historical abnormal characteristic value of each fault characteristic factor based on the historical time series dataset.

[0021] Since raw sensor data cannot directly reflect certain fault modes, such as load mutations, it is necessary to transform the collected raw data into historical abnormal feature values ​​of each fault characteristic factor in order to construct effective features and unify the feature scale. Moreover, the numerical range and dimensions of different features vary greatly. The process of calculating the historical abnormal feature values ​​of each fault characteristic factor itself is to extract the raw time series of different features into a dimensionless, comparable scalar value, which can greatly improve the convergence speed and performance of the BP neural network during subsequent training.

[0022] Specifically, the rated operating parameters of the motor are obtained, including the rated voltage, rated current, and standard temperature range during normal operation. Based on the actual application scenario and historical fault types of the motor, various fault characteristic factors related to motor faults are extracted, including winding temperature characteristics, voltage deviation characteristics, current deviation characteristics, and load change characteristics. Data corresponding to each fault characteristic factor in the historical time series dataset is extracted in turn. The historical time series dataset contains time series data of multiple historical emergency stop moments. The time series data corresponding to each historical emergency stop moment includes voltage value time series data, current value time series data, winding temperature time series data, and speed value time series data. Power value time series data is obtained based on voltage value time series data and current value time series data. Torque value time series data is obtained based on power value time series data and speed value time series data. Among them, the winding temperature characteristics correspond to overheating faults in actual application scenarios, the voltage deviation characteristics correspond to overvoltage or undervoltage faults in actual application scenarios, the current deviation characteristics correspond to overcurrent or undercurrent faults in actual application scenarios, and the load mutation characteristics correspond to mechanical transmission system faults in actual application scenarios. Since power is equal to the product of current and voltage values, power time series data can be obtained from voltage and current time series data. Similarly, since the torque of a motor is equal to the ratio of power to speed, torque time series data can be obtained from power and speed time series data. Since changes in motor torque can map changes in resistance encountered during motor rotation, the load mutation characteristics mapped from torque time series data can reflect mechanical transmission system faults in actual application scenarios.

[0023] Furthermore, based on the time series data of voltage values ​​corresponding to the voltage deviation characteristics at each historical emergency stop, the sum of the squares of the differences between the data values ​​of each sampling point and the rated voltage value within a certain voltage value time series is calculated, and the mean value is obtained based on the number of sampling points within the voltage value time series. The square root of the mean value is used as the historical abnormal feature value of the voltage deviation characteristics at the corresponding historical emergency stop. Then, the historical abnormal feature value of the voltage deviation characteristics at each historical emergency stop is calculated sequentially. Similarly, based on the time series data of current values ​​corresponding to the current deviation characteristics at each historical emergency stop, the historical abnormal feature value of the current deviation characteristics at each historical emergency stop is obtained. Among them, the historical anomaly characteristic value of the voltage deviation at the j-th historical emergency stop time is The calculation formula is as follows: In the formula, This indicates the number of sampling points in the sequence, that is, the number of elements in the time series data. This represents the data value of the i-th sampling point in the time series data of the voltage value at the j-th historical emergency stop. This indicates the rated voltage of the motor; Similarly, the historical anomaly characteristic value of the current deviation at the j-th historical emergency stop is... The calculation formula is as follows: In the formula, Indicates the number of sampling points in the sequence. This represents the data value of the i-th sampling point in the time series data of the current value at the j-th historical emergency stop. This indicates the rated current value of the motor.

[0024] Furthermore, based on the winding temperature time series data corresponding to the winding temperature characteristics at each historical emergency stop, sampling points in the winding temperature time series where the data value exceeds the upper limit of the standard temperature range are extracted as winding temperature characteristic sample points at the corresponding historical emergency stop. The difference between the data value of each winding temperature characteristic sample point and the upper limit of the standard temperature range is calculated sequentially. The difference between the upper and lower limits of the standard temperature range is used as the benchmark value. Based on the number of winding temperature characteristic sample points, the cumulative value of the ratio between the difference and the benchmark value is used as the historical abnormal characteristic value of the winding temperature characteristics at the corresponding historical emergency stop. Thus, the historical abnormal characteristic value of the winding temperature characteristics at each historical emergency stop is calculated sequentially. Among them, the historical abnormal characteristic value of the winding temperature at the j-th historical emergency stop time is The calculation formula is as follows: In the formula, This indicates the number of sampling points in the winding temperature time series data where the data value exceeds the upper limit of the standard temperature range. This indicates the sampling point in the winding temperature time series data at the j-th historical emergency stop time where the i-th data value exceeds the upper limit of the standard temperature range. This indicates the upper limit of the standard temperature range for normal operation of the electric motor. This indicates the lower limit of the standard temperature range during normal operation of the electric motor.

[0025] In addition, based on the time series data of torque values ​​corresponding to the load mutation characteristics at each historical emergency stop, the absolute value of the difference between the current sampling point and the previous sampling point in a certain torque value time series is calculated sequentially to obtain the torque value absolute difference sequence. The overall standard deviation of all data in the torque value absolute difference sequence is used as the historical abnormal feature value of the load mutation characteristics at the corresponding historical emergency stop. Then, the historical abnormal feature value of the load mutation characteristics at each historical emergency stop is calculated sequentially. Among them, the historical abnormal feature value of the load mutation feature at the j-th historical emergency stop time is The calculation formula is as follows: In the formula, Indicates the number of sampling points in the sequence. This represents the value of the i-th element in the sequence of absolute differences in torque values ​​at the j-th historical emergency stop. This represents the data value of the i-th sampling point in the time series data of torque value at the j-th historical emergency stop. This represents the data value of the (i-1)th sampling point in the time series data of the torque value at the j-th historical emergency stop, which is the data value of the sampling point preceding the i-th sampling point in the sequence. This represents the mean of all elements in the sequence of absolute differences in torque values ​​at the j-th historical emergency stop.

[0026] Step S3: Calculate the importance weight of each fault characteristic factor based on historical abnormal characteristic values.

[0027] During motor operation, fault characteristics such as normal and fault states of temperature and current are not always clearly defined, but rather there is a fuzzy transition range. For example, when the winding temperature fluctuates near the upper limit of the standard, the risk is uncertain, and directly using a simple threshold to judge is very likely to lead to false alarms or missed alarms. Therefore, it is necessary to define fuzziness weights to quantify the uncertainty of this state. At the same time, the diagnostic reliability of different fault characteristics in the transition range is also different, and it is necessary to define reliability weights to evaluate the accuracy of the judgment, so as to map the reliability of each characteristic factor as a motor fault criterion. Moreover, different fault characteristic factors have varying degrees of impact on the overall risk of the motor. For example, in older motors, temperature characteristics may be a better predictor of risk than current characteristics. It is necessary to define importance weights to balance the contributions of each characteristic factor, rather than simply treating them as equal. Furthermore, a single fuzzy weight or reliability weight cannot fully reflect the comprehensive impact of fault characteristic factors. Therefore, it is necessary to consider both the fuzziness and reliability of the characteristics, and to calculate the importance weight by integrating the fuzzy weight and reliability weight, in order to avoid the one-sidedness of a single indicator and thus more accurately represent the proportion of different fault characteristic factors in causing motor failures.

[0028] Specifically, the judgment labels for whether the motor is normal or faulty are extracted for each historical emergency stop. The historical abnormal feature values ​​of a certain fault characteristic factor at all historical emergency stop times are used as samples, and the samples are divided into normal sample set and fault sample set based on the judgment labels. The historical abnormal feature values ​​in the normal sample set and the fault sample set are sorted in ascending order. The overlapping interval between the normal sample set and the fault sample set is used as the transition interval of the fault characteristic factor. The maximum value of the historical abnormal feature value in the normal sample set is used as the upper limit of the transition interval, and the minimum value of the historical abnormal feature value in the fault sample set is used as the lower limit of the transition interval. The number of samples in the transition interval is used as the interval width. The total number of samples of the fault characteristic factor is used as the total sample width. The ratio between the interval width and the total sample width is used as the fuzzy weight of the fault characteristic factor. Similarly, the transition interval and fuzzy weight of each fault characteristic factor are obtained. The fuzziness weight is used to directly map the probability that the corresponding fault characteristic factor falls within the transition interval. The more samples in the interval, the wider the interval, and the larger the value of the fuzziness weight. This indicates that there is more overlap between the normal and fault states of the corresponding fault characteristic factor. The higher the fuzziness and the worse the reliability of the corresponding fault characteristic factor when used as a basis for judgment, the more cautious consideration should be given in subsequent comprehensive evaluation.

[0029] The fuzzy weight of the Fth fault characteristic factor is: The calculation formula is as follows: In the formula, This represents the width of the transition interval for the Fth type of fault characteristic factor, i.e., the number of samples within the transition interval for the Fth type of fault characteristic factor. This represents the total sample width of the Fth type of fault characteristic factor, which is the total number of all samples of the Fth type of fault characteristic factor.

[0030] Furthermore, for a transition interval of a certain fault characteristic factor, samples with different data values ​​within the transition interval are sequentially used as judgment thresholds, and judgment conditions are set. When the historical abnormal characteristic value of the fault characteristic factor is greater than or equal to the judgment threshold, it is marked as a fault; when it is less than the judgment threshold, it is marked as normal. For a certain judgment threshold, based on the judgment conditions and the judgment labels corresponding to each historical emergency stop time, the ratio between the number of samples that meet the judgment conditions and the interval width within the transition interval is used as the accuracy rate of the judgment threshold. The accuracy rate of each judgment threshold is calculated sequentially, and the average of the accuracy rates of the judgment thresholds is used as the reliability weight of the fault characteristic factor. Similarly, the reliability weight of each fault characteristic factor is obtained. The essence of reliability weight is the conditional probability that the corresponding fault characteristic factor is identified as the real fault source within the transition interval. It is used to map the historical performance and predictive ability of the corresponding fault characteristic factor. The higher the value of reliability weight, the higher the classification accuracy of the corresponding fault characteristic factor can be even within the fuzzy interval, and the more reliable the diagnostic results are when the corresponding fault characteristic factor is used as the basis for judgment.

[0031] The reliability weight of the Fth type of fault characteristic factor is: The calculation formula is as follows: In the formula, K represents the number of judgment thresholds, that is, the number of samples with different data values ​​within the transition interval of the Fth fault characteristic factor; This represents the number of samples that meet the judgment condition within the transition interval for the Fth type of fault characteristic factor, when the judgment threshold is selected to include samples with different data values ​​at the pth time. This represents the width of the transition interval for the Fth type of fault characteristic factor.

[0032] In addition, importance coefficients for fuzzy weights and reliability weights are set according to the actual application scenarios and usage time of the motor. Based on a certain fault characteristic factor, the sum of the products of its fuzzy weight and reliability weight with the corresponding importance coefficients is taken as the importance weight of the fault characteristic factor. Similarly, the importance weight of each fault characteristic factor is obtained. Importance weight is a combination of fuzzy weight and reliability weight, which can more reasonably reflect the true contribution of each fault characteristic factor. The importance coefficients of fuzzy weight and reliability weight can be set according to the actual application scenario and usage time of the motor. For example, the importance coefficient can be adjusted according to the specific equipment conditions, such as the age of the equipment and the severity of the operating conditions. For older equipment, the proportion of reliability weight can be increased.

[0033] Among them, the importance weight of the Fth fault characteristic factor is: The calculation formula is as follows: In the formula, This represents the fuzzy weight of the Fth type of fault characteristic factor. This represents the reliability weight of the Fth type of failure characteristic factor. The importance coefficient represents the weight of fuzziness. The importance coefficient represents the reliability weight.

[0034] Step S4: Construct a risk prediction model and train it based on historical anomaly feature values ​​to obtain the model parameters corresponding to each fault feature factor.

[0035] Directly inputting raw data collected from industrial sites into risk prediction models can lead to problems such as unstable training and large prediction biases. Therefore, extracting key features related to faults and using historical anomaly feature values ​​to train the risk prediction model is crucial. Historical anomaly feature values ​​are easier for neural networks to learn and generalize, accelerating the update process of neural network model parameters, making training more efficient, and thus improving the model's convergence speed and stability. At the same time, using historical anomaly feature values ​​to train the risk prediction model can avoid the influence of noise or data defects in the raw data, enabling the model to learn patterns that are truly related to faults in the data. This allows for more accurate predictions when faced with new data, improving the model's prediction accuracy and generalization ability, and enhancing the system's practicality.

[0036] Specifically, a risk prediction model is constructed using a BP neural network (Back-Propagation neural network, a multi-layer feedforward network trained by the error backpropagation algorithm) classification model, and the sigmoid function is used as the activation function. The historical abnormal feature values ​​of each fault feature factor at all historical emergency stop times are sequentially input into the risk prediction model for training, so as to obtain the model parameters corresponding to each fault feature factor.

[0037] Step S5: Obtain the current abnormal feature values ​​corresponding to each fault feature factor at the current emergency stop of the motor, and obtain the fault risk probability of each fault feature factor based on the risk prediction model and the corresponding model parameters.

[0038] Specifically, the system acquires and preprocesses the operating data sequence of the motor during the current emergency stop within a preset time period to obtain the current time series data. The data acquisition and preprocessing methods for the current emergency stop are the same as those for historical operating data. Based on the current time series data, the system uses the same calculation process as for historical abnormal feature values ​​to obtain the current abnormal feature values ​​corresponding to each fault feature factor. Based on the risk prediction model and the model parameters corresponding to each fault feature factor, the current abnormal feature values ​​corresponding to each fault feature factor are input into the risk prediction model to obtain the fault risk probability of each fault feature factor. In addition, the fault risk probabilities of each fault characteristic factor are sorted in descending order, and the order of the fault characteristic factors is used as the priority order for troubleshooting motor faults in practical applications.

[0039] Step S6: Calculate the overall risk value of the motor at the current moment based on the fault risk probability and importance weight of each fault characteristic factor.

[0040] Motor failures are usually the result of multiple factors working together. A single factor assessment is prone to being one-sided. By combining the importance weights with the failure risk probability output by the model, the contributions of all failure characteristic factors can be integrated to obtain the overall risk value of the motor. This ensures that important factors have a greater impact on the results, thereby improving the comprehensiveness and accuracy of the assessment.

[0041] Specifically, the sum of the importance weights of all fault characteristic factors is taken as the total weight value. The ratio between the importance weight of a certain fault characteristic factor and the total weight value is taken as the normalized importance weight of that fault characteristic factor. The normalized importance weight of each fault characteristic factor is calculated in turn. The sum of the products between the normalized importance weight of each fault characteristic factor and the corresponding fault risk probability is taken as the overall risk value of the motor at the current moment. The formula for calculating the overall risk value of the electric motor is as follows: In the formula, R represents the overall risk value of the motor at the current moment, and N represents the number of types of fault characteristic factors. In this embodiment of the invention, N=4. This represents the importance weight of the Fth type of fault characteristic factor. This represents the normalized importance weight of the Fth type of fault characteristic factor. This represents the failure risk probability of the Fth type of failure characteristic factor.

[0042] Step S7: Calculate the restart power of the motor based on the overall risk value and execute the corresponding restart control strategy.

[0043] Specifically, the rated power of the motor is obtained, the overall risk value range is divided into three risk levels: low, medium, and high, and a piecewise calculation function for the motor's restart power is set; for the low-risk level, the rated power of the motor is... The restart power of the motor is used as the restart power; for medium-risk levels, an exponential function is used to reverse-map the overall risk value, and the product of the reverse-mapped value and the rated power of the motor is used as the restart power of the motor; for high-risk levels, the rated power of the motor is used as the restart power. The restart power of the motor is used as the starting power of the motor. Based on the overall risk value of the motor at the current moment, the restart power of the motor at the current moment is calculated and converted into the corresponding speed control command to control the motor to execute. In this embodiment of the invention, the range of the overall risk value is as follows: The overall risk value is in The time period is classified as low risk level, and the overall risk value is in The risk level is classified as medium, with an overall risk value of [missing information]. The time frame is classified as high-risk; the piecewise calculation function for the motor restart power is as follows: In the formula, This indicates the restarting power of the motor. This indicates the rated power of the electric motor. This indicates the overall risk value of the electric motor.

[0044] This invention also discloses a speed control system for a high-voltage frequency converter, used to implement the speed control method for the aforementioned high-voltage frequency converter. The system structure is as follows: Figure 2 As shown, since the speed control system of the high-voltage frequency converter of the present invention is applicable to various types of motors, but the system does not include a motor structure, therefore... Figure 2The diagram represents an "electric motor," with its border indicated by dashed lines. The specific structure of the system includes a processor, a memory, a communication interface, and a data acquisition device. The processor stores computer program instructions for implementing the speed control method of the aforementioned high-voltage frequency converter. The data acquisition device consists of sensors installed on the electric motor, including voltage sensors, current sensors, and temperature sensors. The communication interface is connected to both the data acquisition device and the electric motor to receive information data from the electric motor acquired by the data acquisition device and to transmit speed control commands to the electric motor.

[0045] The embodiments included in this invention are merely descriptions of preferred embodiments of the invention and are not limited to the precise structures described above and shown in the accompanying drawings. Various modifications and changes can be made without departing from the scope of protection. Any variations and improvements made by those skilled in the art to the technical solutions of this invention without departing from the design concept of this invention should fall within the scope of protection of this invention.

Claims

1. A speed control method for a high-voltage frequency converter, characterized in that: Historical operating data of the motor within a preset period before the historical emergency stop time is sampled and preprocessed to obtain the corresponding historical time series dataset; various fault characteristic factors related to motor faults are extracted, and historical abnormal characteristic values ​​of each fault characteristic factor are calculated based on the historical time series dataset; Based on historical abnormal feature values, the importance weight of each fault feature factor is calculated; a risk prediction model is constructed and trained based on historical abnormal feature values ​​to obtain the model parameters corresponding to each fault feature factor; the current abnormal feature value corresponding to each fault feature factor at the current emergency stop of the motor is obtained, and the fault risk probability of each fault feature factor is obtained based on the risk prediction model and the corresponding model parameters. Based on the fault risk probability and importance weight of each fault characteristic factor, calculate the overall risk value of the motor at the current moment; calculate the restart power of the motor based on the overall risk value, and execute the corresponding restart control strategy. Based on the historical abnormal characteristic values ​​of each fault characteristic factor, the fuzzy weight and reliability weight of each fault characteristic factor are first calculated, and then the importance coefficients of the fuzzy weight and reliability weight are set. The sum of the products of the fuzzy weight and reliability weight with the corresponding importance coefficients is used as the importance weight of the corresponding fault characteristic factor.

2. The speed control method for a high-voltage frequency converter according to claim 1, characterized in that, The sampling process acquires historical operating data within a preset time period prior to each historical emergency stop of the motor and preprocesses it to obtain a corresponding historical time series dataset. This includes: acquiring operating data sequences corresponding to each sensor within a preset time period prior to each motor emergency stop using voltage, current, temperature, and speed sensors installed on the motor at the same sampling frequency; collecting voltage, current, winding temperature, and speed value sequences; taking all emergency stop times prior to the current emergency stop of the motor as historical emergency stop times; taking the motor operating data sequences corresponding to historical emergency stop times as historical data sequences; acquiring historical data sequences corresponding to multiple historical emergency stop times, as well as the motor normal or motor fault judgment labels corresponding to each historical emergency stop time; filling missing values ​​in each historical data sequence using mean interpolation; aligning the timestamps of all sampling points in each historical data sequence to obtain the time series data corresponding to each historical data sequence; and taking the set of time series data corresponding to all historical emergency stop times as the historical time series dataset.

3. The speed control method for a high-voltage frequency converter according to claim 1, characterized in that, The process involves extracting various fault characteristic factors related to motor faults and calculating historical abnormal characteristic values ​​for each fault characteristic factor based on a historical time series dataset. This includes: obtaining the rated operating parameters of the motor, including rated voltage, rated current, and the standard temperature range during normal operation; extracting various fault characteristic factors related to motor faults based on the actual application scenario and historical fault types, including winding temperature characteristics, voltage deviation characteristics, current deviation characteristics, and load mutation characteristics; sequentially extracting data corresponding to each fault characteristic factor from the historical time series dataset, which contains time series data for multiple historical emergency stops. The time series data corresponding to each historical emergency stop includes voltage time series data, current time series data, winding temperature time series data, and speed time series data; obtaining power time series data based on voltage and current time series data; and obtaining torque time series data based on power and speed time series data.

4. The speed control method for a high-voltage frequency converter according to claim 3, characterized in that, The calculation of historical abnormal feature values ​​for each fault characteristic factor further includes: based on the voltage value time series data corresponding to the voltage deviation characteristics at each historical emergency stop time, calculating the sum of the squares of the differences between the data values ​​of each sampling point in a certain voltage value time series and the rated voltage value, and obtaining the mean value based on the number of sampling points in the voltage value time series, taking the square root of the mean value as the historical abnormal feature value of the voltage deviation characteristics at the corresponding historical emergency stop time, and then calculating the historical abnormal feature value of the voltage deviation characteristics at each historical emergency stop time in sequence; similarly, based on the current value time series data corresponding to the current deviation characteristics at each historical emergency stop time, obtaining the historical abnormal feature value of the current deviation characteristics at each historical emergency stop time; Based on the winding temperature time series data corresponding to the winding temperature characteristics at each historical emergency stop, sampling points in the winding temperature time series where the data value exceeds the upper limit of the standard temperature range are extracted as winding temperature characteristic sample points at the corresponding historical emergency stop. The difference between the data value of each winding temperature characteristic sample point and the upper limit of the standard temperature range is calculated in turn. The difference between the upper and lower limits of the standard temperature range is used as the benchmark value. Based on the number of winding temperature characteristic sample points, the cumulative value of the ratio between the difference and the benchmark value is used as the historical abnormal characteristic value of the winding temperature characteristics at the corresponding historical emergency stop. Then, the historical abnormal characteristic value of the winding temperature characteristics at each historical emergency stop is calculated in turn. Based on the time series data of torque values ​​corresponding to the load mutation characteristics at each historical emergency stop, the absolute value of the difference between the current sampling point and the previous sampling point in a certain torque value time series is calculated sequentially to obtain the torque value absolute difference sequence. The overall standard deviation of all data in the torque value absolute difference sequence is used as the historical abnormal feature value of the load mutation characteristics at the corresponding historical emergency stop. Then, the historical abnormal feature value of the load mutation characteristics at each historical emergency stop is calculated sequentially.

5. The speed control method for a high-voltage frequency converter according to claim 2, characterized in that, The step of calculating the importance weight of each fault characteristic factor based on historical abnormal feature values ​​includes: extracting the judgment label of whether the motor is normal or faulty at each historical emergency stop time; taking the historical abnormal feature values ​​of a certain fault characteristic factor at all historical emergency stop times as samples, and dividing them into a normal sample set and a fault sample set based on the judgment label; sorting the historical abnormal feature values ​​in the normal sample set and the fault sample set in ascending order; taking the overlapping interval between the normal sample set and the fault sample set as the transition interval of the fault characteristic factor; taking the maximum value of the historical abnormal feature values ​​in the normal sample set as the upper limit of the transition interval; taking the minimum value of the historical abnormal feature values ​​in the fault sample set as the lower limit of the transition interval; counting the number of samples in the transition interval as the interval width; taking the total number of all samples of the fault characteristic factor as the total sample width; and taking the ratio between the interval width and the total sample width as the fuzzy weight of the fault characteristic factor. Similarly, the transition interval and fuzzy weight of each fault characteristic factor are obtained. For a transition interval of a certain fault characteristic factor, samples with different data values ​​within the transition interval are sequentially used as judgment thresholds. Judgment conditions are set: when the historical abnormal characteristic value of the fault characteristic factor is greater than or equal to the judgment threshold, it is marked as a fault; when it is less than the judgment threshold, it is marked as normal. For a certain judgment threshold, based on the judgment conditions and the judgment labels corresponding to each historical emergency stop time, the ratio between the number of samples that meet the judgment conditions and the interval width is used as the accuracy rate of the judgment threshold. The accuracy rate of each judgment threshold is calculated sequentially, and the average of the judgment threshold accuracy rates is used as the reliability weight of the fault characteristic factor. Similarly, the reliability weight of each fault characteristic factor is obtained. The importance coefficients of fuzzy weight and reliability weight are set according to the actual application scenario and usage time of the motor. Based on the values ​​of the fuzzy weight and reliability weight of each fault characteristic factor, the importance weight of each fault characteristic factor is calculated sequentially.

6. The speed control method for a high-voltage frequency converter according to claim 1, characterized in that, The process of constructing a risk prediction model and training it based on historical abnormal feature values ​​to obtain model parameters corresponding to each fault feature factor includes: constructing a risk prediction model using a BP neural network classification model and using the sigmoid function as the activation function; sequentially inputting the historical abnormal feature values ​​of each fault feature factor at all historical emergency stop times into the risk prediction model for training to obtain model parameters corresponding to each fault feature factor.

7. The speed control method for a high-voltage frequency converter according to any one of claims 1 to 6, characterized in that, The process of obtaining the current abnormal feature values ​​corresponding to each fault feature factor at the current emergency stop of the motor, and obtaining the fault risk probability of each fault feature factor based on the risk prediction model and the corresponding model parameters, includes: obtaining the operating data sequence of the motor at the current emergency stop within a preset time period and performing preprocessing to obtain the current time series data. The data acquisition method and preprocessing process for the current emergency stop are the same as those for historical operating data. Based on the current time series data, the same calculation process as for historical abnormal feature values ​​is used to obtain the current abnormal feature values ​​corresponding to each fault feature factor. Based on the risk prediction model and the model parameters corresponding to each fault feature factor, the current abnormal feature values ​​corresponding to each fault feature factor are input into the risk prediction model to obtain the fault risk probability of each fault feature factor.

8. The speed control method for a high-voltage frequency converter according to claim 7, characterized in that, The calculation of the overall risk value of the motor at the current moment based on the fault risk probability and importance weight of each fault characteristic factor includes: summing the importance weights of all fault characteristic factors as the total weight value; taking the ratio between the importance weight of a certain fault characteristic factor and the total weight value as the normalized importance weight of that fault characteristic factor; calculating the normalized importance weight of each fault characteristic factor in turn; and summing the products of the normalized importance weight of each fault characteristic factor and the corresponding fault risk probability as the overall risk value of the motor at the current moment.

9. The speed control method for a high-voltage frequency converter according to claim 8, characterized in that, The process of calculating the motor's restart power based on the overall risk value and executing the corresponding restart control strategy includes: obtaining the motor's rated power; dividing the overall risk value range into three risk levels: low, medium, and high; and setting a piecewise calculation function for the motor's restart power; for the low risk level, calculating the motor's rated power... The restart power of the motor is used as the restart power; for medium-risk levels, an exponential function is used to reverse-map the overall risk value, and the product of the reverse-mapped value and the rated power of the motor is used as the restart power of the motor; for high-risk levels, the rated power of the motor is used as the restart power. The restart power of the motor is used as the starting power. Based on the overall risk value of the motor at the current moment, the restart power of the motor at the current moment is calculated and converted into the corresponding speed control command to control the motor to execute.

10. A speed control system for a high-voltage frequency converter, characterized in that: It includes a processor, a memory, a communication interface, and a data acquisition device. The processor stores computer program instructions for implementing the speed control method of a high-voltage frequency converter according to any one of claims 1 to 9. The data acquisition device is a sensor installed on the motor. The communication interface is communicatively connected to the data acquisition device and the motor.

Images