A variable switching frequency control method for semiconductor devices
By dynamically adjusting the weight and step size of the LMS adaptive filter, the contradiction between steady-state error and convergence speed caused by fixed step size is solved, more accurate current control and frequency management are achieved, and the performance of variable switching frequency control is improved.
Patent Information
- Application Number
- CN202411265994.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-11
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-09-11
AI Technical Summary
LMS adaptive filtering uses a fixed step size, making it difficult to coordinate steady-state error and convergence speed, resulting in inaccurate control of the variable switching frequency of semiconductor devices.
By presetting the basic step size of the filter, and during each iteration, the filter weight and step size are dynamically adjusted according to the distribution differences and error changes of the current data to optimize the filter parameters.
It realizes more accurate current signal tracking, reduces noise interference, improves the dynamic response performance and steady-state accuracy of the control system, and optimizes the variable switching frequency control performance of semiconductor devices.
Smart Images

Figure CN118783748B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing technology, and more particularly to a variable switching frequency control method for a semiconductor device. Background Art
[0002] Pulse converters are widely used in power electronics, motor control, marine systems, and other power applications. They convert and control electrical energy through semiconductor devices such as IGBTs (insulated-gate bipolar transistors), thyristors, and diodes, enabling power transmission and utilization. The selection of semiconductor device switching frequencies involves considerations of thermal management and device lifespan. Excessively high switching frequencies can increase device heat loss, requiring more complex heat dissipation designs for effective cooling. Excessively low switching frequencies not only affect system energy efficiency and waveform quality but can also reduce the stability and response speed of the pulse converter, increasing its size, cost, and electromagnetic compatibility. Therefore, it is necessary to balance the effects of switching frequency on these factors to ensure that the pulse converter meets design requirements and maintains good performance.
[0003] Patent document CN110943602B discloses a method and control system for controlling the switching frequency of a converter, including first detecting the actual harmonic content of the converter, then comparing the actual harmonic content with a set content threshold, and controlling the converter to increase the switching frequency when the actual harmonic content is greater than the set content threshold and the converter is in a seven-segment control mode; and controlling the converter to decrease the switching frequency when the actual harmonic content is less than or equal to the set content threshold and the converter switching frequency is greater than the minimum allowable switching frequency. This control method can control the operating mode and switching frequency of the converter's switching devices according to changes in the harmonic content output at the grid connection point, while ensuring that the converter can be connected to the grid and operate stably. This ensures that the converter has a low harmonic content within the full power output range, achieves low harmonic and high-efficiency operation within the converter's full power range, and improves the converter's operating efficiency.
[0004] However, the converter itself has accuracy limitations and errors, and electromagnetic interference may exist in the current measurement circuit, resulting in inaccurate monitored current, which in turn affects the control of the switching frequency of the semiconductor device. The LMS adaptive filtering algorithm is typically used to filter the output current of the pulse converter. However, the step size of the LMS adaptive filter is fixed. Although a smaller step size can achieve a smaller steady-state error, the convergence speed is relatively slow. A larger step size can increase the convergence speed of the filter, but it will increase the steady-state error. Therefore, the use of a fixed step size in the LMS adaptive filter makes it difficult to reconcile the contradiction between steady-state error and convergence speed, resulting in inaccurate control of the variable switching frequency of the semiconductor device by the output current data. Summary of the Invention
[0005] In order to solve the problem that the fixed step size used in LMS adaptive filtering makes it difficult to coordinate the contradiction between steady-state error and convergence speed, resulting in inaccurate control of the variable switching frequency of semiconductor devices by output current data, the present invention proposes a variable switching frequency control method for semiconductor devices, which includes the following steps:
[0006] The basic step size of the filter is preset, the current data is subjected to LMS adaptive filtering through the filter, and the variable switching frequency of the semiconductor device is controlled according to the filtered current data; in each round of filtering iteration, the filter weight of the next moment is determined, including: determining the importance of each previous moment according to the distribution difference between the current data of each previous moment and the current moment, and the importance is negatively correlated with the distribution difference; for any moment, the difference between the current data at that moment and the expected value is determined as the error at that moment, and the cost value is determined according to the error at that moment, and the cost value is positively correlated with the error; in response to the cost value at that moment being not greater than the current moment Cost value, the weight of the moment is determined according to the importance of the moment and the difference between the cost value of the moment and the current moment, and the weight is positively correlated with the importance and the difference; the basic step size of the filter is adjusted according to the weights and errors of the previous moments, and the adjusted step size is positively correlated with the weights, errors and basic step size of the previous moments; the filter weight of the next moment is determined according to the filter weight, error, current data and adjusted step size at the current moment, and the filter weight of the next moment is positively correlated with the filter weight, error, current data and adjusted step size at the current moment; based on the filter weight of the next moment, the next round of filtering is performed.
[0007] The beneficial effects are: through LMS adaptive filtering technology, the filter weight can be dynamically adjusted according to the actual changes in current data, so as to more accurately track the current signal, reduce noise interference, and improve the dynamic response performance and steady-state accuracy of the control system; variable switching frequency control can optimize switching operations according to real-time operating conditions, avoid unnecessary high-frequency switching actions, reduce switching losses, and thus improve the energy efficiency of the overall system; by considering the distribution differences and errors of current data at historical moments, the filter weight is dynamically adjusted, so that the control strategy can better adapt to changes in operating conditions and external disturbances, thereby enhancing the stability and robustness of the system; the variable switching frequency strategy can flexibly adjust the switching frequency according to system requirements, avoid resonance or interference problems that may be caused by fixed frequency operation, and optimize the system's utilization of frequency resources; through more precise current control and frequency management, the electromagnetic interference generated by the system can be effectively suppressed, meeting electromagnetic compatibility requirements, and improving the reliability and environmental friendliness of the equipment.
[0008] Furthermore, the basic step size of the filter satisfies the following relationship: Where, is the basic step size of the filter, is the maximum value in the current data, is the minimum value in the current data, The first The current value of the data point, is the number of data points in the current data, The first A maximum value, The first A minimum value, is the number of maxima in the current data, is the number of minimum values in the current data.
[0009] The beneficial effect is that by combining the inverse of the variance of the current data and the extreme value deviation, adaptive adjustment of the filter learning rate is achieved, which not only enhances the adaptability to the dynamic range of the current signal, but also improves the filtering performance and control accuracy, and effectively copes with the current fluctuation characteristics under different working conditions.
[0010] Furthermore, the number of data of the current data at each moment is consistent with the number of orders of the filter.
[0011] Furthermore, the importance of each previous moment satisfies the following relationship: Where, For the previous The importance of a moment, is the mean value of the current data corresponding to the error at the current moment, is the standard deviation of the current data corresponding to the error at the current moment, For the previous The mean value of the current data corresponding to the error at each moment, For the previous The standard deviation of the current data corresponding to the error at each moment is: is the vector of current data corresponding to the error at the current moment, For the previous The vector of current data corresponding to the error at each moment is composed of, is the number of errors at previous moments, is the cosine similarity function, It is an exponential function with the natural constant e as its base.
[0012] The beneficial effect is that by fusing the difference between the mean and standard deviation of the error with the cosine similarity, a refined evaluation of the historical error is achieved, which improves the learning efficiency in the adaptive filtering process and the sensitivity to changes in the current signal characteristics, thereby achieving better adaptability and control accuracy in variable switching frequency control.
[0013] Furthermore, the cost value satisfies the following relationship: Where, is the cost value at the current moment, is the number of data points in the error, is the first data point in the error at the current moment, is the second data point in the error at the current moment, is the error at the current moment data points, is the error at the current moment data points.
[0014] Furthermore, Where, For the previous The weight of the moment, For the previous The importance of a moment, is the cost value at the current moment, For the previous The cost value at this moment, For the previous The importance of a moment, For the previous The cost value at this moment, is the number of errors at previous moments.
[0015] The beneficial effects are: by comparing the cost and importance of the current and previous moments, the contribution of each previous moment to the filtering process is dynamically adjusted, ensuring the optimal distribution of filter weights, enhancing the response speed to significant error changes, and improving the adaptability and accuracy of variable switching frequency control.
[0016] Furthermore, the adjusted step length satisfies the following relationship: , Where, is the adjusted step size, is the first data point of the adjusted step size, is the second data point of the adjusted step size, The adjusted step size data points, The adjusted step size data points, is the basic step size of the filter, For the previous The weight of the error at each moment, For the previous The error at the moment data points.
[0017] Furthermore, the filter weight at the next moment satisfies the following relationship: Where, is the filter weight at the next moment, The filter weight at the current moment, is the adjusted step size, is the error at the current moment, is the current data at the current moment.
[0018] The beneficial effect is that by combining the adjusted step size, current error and current data, adaptive updating of the filter weights is achieved, which promotes rapid convergence while ensuring accurate tracking of current signal changes, and further optimizes the variable switching frequency control performance of semiconductor devices.
[0019] The present invention has the following beneficial effects:
[0020] The present invention adaptively optimizes the filtering parameters according to the current data distribution differences and error changes, which not only accelerates the algorithm convergence speed but also ensures the filtering accuracy, thereby achieving better dynamic response and energy efficiency in variable switching frequency control; by dynamically adjusting the filter weight and step size, the contradiction between the steady-state error and the convergence speed caused by the fixed step size in the LMS adaptive filtering is effectively resolved, making the output current data more accurate in controlling the variable switching frequency of the semiconductor device. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an illustrative and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:
[0022] Figure 1 The present invention is a flowchart of a method for controlling a variable switching frequency of a semiconductor device according to an embodiment of the present invention.
[0023] Figure 2 This is a flow chart of the steps of performing LMS adaptive filtering on current data through a filter in a variable switching frequency control method for a semiconductor device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.
[0025] It should be understood that when the terms "first," "second," and the like are used in the claims, description, and drawings of the present invention, they are merely used to distinguish between different objects, rather than to describe a specific order. The terms "comprise" and "comprising" used in the description and claims of the present invention indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or combinations thereof.
[0026] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0027] The embodiment of the present invention discloses a method for controlling the switching frequency of a semiconductor device, referring to Figure 1 , including steps S1 to S4:
[0028] S1: Collect current data.
[0029] The current data of the pulse converter is collected through a current sensor. Since the output current directly reflects the load demand and the output performance of the converter, in order to make the control of the semiconductor switching frequency better respond to the changes in the pulse converter load and maintain the stable output of the pulse converter, a current sensor is installed at the output end of the pulse converter to monitor the output current of the pulse converter in real time and obtain the current data of the pulse converter.
[0030] S2: The basic step size of the preset filter.
[0031] The basic step size of the filter satisfies the following relationship:
[0032] ;
[0033] Where, is the basic step size of the filter, is the maximum value in the current data, is the minimum value in the current data, The first The current value of the data point, is the number of data points in the current data, The first A maximum value, The first A minimum value, is the number of maxima in the current data, is the number of minimum values in the current data.
[0034] is the difference between the maximum value in the current data and the i-th maximum value in the current data. The larger the value, the greater the difference between the maximum value and the maximum value in the current data. The smaller the value, the smaller the difference between the maximum value and the maximum value in the current data. It is the average level of the gap between the maximum value and all the maximum values in the current data. The larger the value is, the larger the gap between the maximum value in the current data and the maximum value in the current data is. The current data is more likely to fluctuate around the average value of the current data as a whole, and the smaller the upward change of the current data is. The smaller the value is, the smaller the gap between the maximum value in the current data and the maximum value in the current data is. The more data in the current data is close to the maximum current value, the larger the upward change of the current data is. is the difference between the minimum value in the current data and the jth minimum value in the current data. The larger the value, the greater the difference between the minimum value and the minimum value in the current data. The smaller the value, the smaller the difference between the minimum value and the minimum value in the current data. It is the average level of the gap between the minimum value and all the minimum values in the current data. The larger the value, the larger the gap between the minimum value and the minimum value in the current data. The more likely the current data is to fluctuate around the average value of the current data as a whole. The smaller the value, the smaller the gap between the minimum value and the minimum value in the current data. The more data close to the minimum value in the current data, the larger the downward variation of the current data. In order to facilitate calculation, we use and remove ,get and , = is the overall change amplitude of the current data. The larger the value, the greater the overall change amplitude of the current data. The smaller the value, the smaller the overall change amplitude of the current data. The larger the overall change amplitude of the current data, the faster the current signal may change and the more likely the current data is to contain high-frequency components. In this case, a larger step size should be used to update the filter weight coefficients with a larger amplitude, allowing the LMS adaptive filtering algorithm to adapt to changes in the current signal more quickly and adjust to the appropriate state more quickly, thereby improving the convergence speed. The smaller the overall change amplitude of the current data, the slower the current signal may change and the more likely the current data is to contain mainly low-frequency components. In this case, a smaller step size should be used to update the filter weight coefficients with a smaller amplitude. A smaller step size helps reduce steady-state error and improve the accuracy of the filter in a stable state. is the mean square value of the current data. The larger the value, the more energy and fluctuation the current signal contains. In order to ensure the stability of the LMS adaptive filtering algorithm, a smaller step size should be used to prevent the weight update amplitude from being too large, resulting in unstable filtering results. The smaller the value, the less energy the current signal contains. A larger step size should be used to speed up the convergence.
[0035] S3: Perform LMS adaptive filtering on the current data through the filter.
[0036] Reference Figure 2 In each round of filtering iteration, the filter weight at the next moment is determined, including steps S31 to S36, as follows:
[0037] S31: Determine the importance of each previous moment.
[0038] The importance of each previous moment is determined according to the distribution difference between the current data at each previous moment and the current data at the current moment, and the number of data of the current data at each moment is consistent with the number of orders of the filter.
[0039] The embodiment personnel can select the order number of the filter according to the actual operation situation, for example, the order number is 5.
[0040] In the present invention and embodiments Indicates the current moment.
[0041] The importance of each previous moment satisfies the following relationship:
[0042] ;
[0043] Where, Before the current moment The importance of a moment, is the mean value of the current data corresponding to the error at the current moment, is the standard deviation of the current data corresponding to the error at the current moment, Before the current moment The mean value of the current data corresponding to the error at each moment, Before the current moment The standard deviation of the current data corresponding to the error at each moment is: is the vector of current data corresponding to the error at the current moment, Before the current moment The vector of current data corresponding to the error at each moment is composed of, is the number of errors at previous moments, is the cosine similarity function, It is an exponential function with the natural constant e as its base.
[0044] The error corresponding to the current data obtained at the current moment is the average value of the current data before the current moment. The error obtained at each moment corresponds to the difference between the average values of the current data. The smaller the value, the greater the difference between the average values of the current data before the current moment. The more likely the current data features corresponding to the error obtained at the previous moment are similar to the current data features corresponding to the error obtained at the current moment, the similar data features mean that the filter encountered similar problems when processing the data corresponding to these moments. At this time, the error at the previous moment can provide more valuable information to guide the update of the filter weight. The more important the error is, the greater the value is. The greater the difference between the current data characteristics corresponding to the error obtained at the previous moment and the current data characteristics corresponding to the error obtained at the current moment, the greater the difference in data characteristics means that the filter is more likely to encounter different problems when processing the data corresponding to these moments. At this time, the error at the previous moment is less likely to provide valuable information to guide the update of the filter weight. The less important the error is, the easier it is to calculate the right Normalize it and get . The standard deviation of the current data corresponding to the error obtained at the current moment and the standard deviation of the current data before the current moment The error obtained at each moment corresponds to the difference between the standard deviations of the current data. The smaller the value, the greater the difference between the first and second moments before the current moment. The more likely the current data characteristics corresponding to the error obtained at the previous moment are similar to the current data characteristics corresponding to the error obtained at the current moment, the more likely the current data characteristics corresponding to the error obtained at the previous moment are similar to the current data characteristics corresponding to the error obtained at the current moment. The more important the error is, the greater the value is. The greater the difference between the current data characteristics corresponding to the error obtained at the previous moment and the current data characteristics corresponding to the error obtained at the current moment, the greater the difference between the current data characteristics corresponding to the error obtained at the previous moment and the current data characteristics corresponding to the error obtained at the current moment. The less important the error is, the easier it is to calculate the right Normalize it and get . It is the similarity between the current data corresponding to the two errors measured from the perspective of mean and standard deviation. The smaller the value, the more similar the current data corresponding to the error obtained at the current moment is to the current data before the current moment. The more likely the current data corresponding to the error obtained at each moment is similar, the more likely the current data before the current moment is similar. The more likely the error obtained at the moment is important, the more The higher the importance of the moment, the greater the value is, the more the current data corresponding to the error obtained at the current moment is different from the current data before the current moment. The larger the gap between the current data corresponding to the error obtained at each moment, the greater the difference between the current data corresponding to the error obtained at each moment. The more likely the error obtained at the moment is unimportant, the more likely the error is before the current moment. The lower the importance obtained at each moment. Since the variance and standard deviation of the data measure the similarity of the current data corresponding to the errors obtained at two moments from the perspective of the central tendency of the data and the degree of dispersion of the data, but two data series with similar central tendencies and degrees of dispersion do not necessarily have similar change trends (for example, the mean of the two data series A[1,2,3,4,5] and the standard deviation of the two data series B[1,5,2,4,3] are both 3, and the standard deviation is 1.58. The first data series shows a straight-line upward trend, and the second data series shows an oscillating trend. There is a big difference between the two data series in terms of the characteristics of the change trend. Therefore, two data series with the same mean and standard deviation may have different change trends). In order to make the first data series before the current moment The similarity between the current data corresponding to the error obtained at the moment and the current data corresponding to the error obtained at the current moment can better express the The importance obtained at the moment is obtained by The vector of the current data corresponding to the error obtained at the moment The vector corresponding to the current data corresponding to the error obtained at the current moment The cosine similarity between right Correction (for the convenience of calculation, Adjust by adding 1 and dividing by 2 scope of action), The bigger, The smaller the right The greater the expansion degree of , the more similar the change trend characteristics between the current data corresponding to the error obtained at the kth moment before the current moment and the current data corresponding to the error obtained at the current moment are, the more similar the change trend characteristics between the current data corresponding to the error obtained at the kth moment before the current moment are. The more importance a moment gets, The smaller, The larger the The closer it is to 1, the The value depends more on , then the first The error obtained at the moment (relative to The larger the value, the smaller the
[0045] S32: Determine the cost value at each moment.
[0046] At any moment, the difference between the current data at that moment and the expected value is determined as the error at that moment, and the cost value is determined according to the error at that moment.
[0047] The cost value satisfies the following relationship:
[0048] ;
[0049] Where, is the cost value at the current moment, is the number of data points in the error, is the first data point in the error at the current moment, is the second data point in the error at the current moment, is the error at the current moment data points, is the error at the current moment data points.
[0050] S33: Determine the weight of each moment.
[0051] In response to the cost value of the moment being not greater than the cost value of the current moment, a weight of the moment is determined according to the importance of the moment and the difference between the cost values of the moment and the current moment:
[0052] ;
[0053] Where, Before the current moment The weight of the moment, Before the current moment The importance of a moment, is the cost value at the current moment, Before the current moment The cost value at this moment, Before the current moment The importance of a moment, Before the current moment The cost value at this moment, is the number of errors at each moment before the current moment.
[0054] The error at the current moment is Compare Big, No. Time to At this moment, the filter weight update is not updated in the direction of the optimal filter weight, and the error at this time is more meaningful for the update of the filter weight. The error at the current moment is Compare Small, first Time to The weight update of the moment filter is updated in the direction of the optimal filter weight, and there is no need to refer too much to it. The corresponding error To update the filter weights, When the real-time personnel can select the value according to the actual situation, A very small value can be used, for example, .
[0055] S34: Adjust the basic step size of the filter.
[0056] The basic step size of the filter is adjusted according to the weights and errors at previous moments.
[0057] The adjusted step size satisfies the following relationship:
[0058] ;
[0059] ;
[0060] Where, is the adjusted step size, is the first data point of the adjusted step size, is the second data point of the adjusted step size, The adjusted step size data points, The adjusted step size data points, is the basic step size of the filter, Before the current moment The weight of the error at each moment, Before the current moment The error at the moment data points.
[0061] when The larger the value, the Time error The more important it is, the more important it is before the current moment when updating the filter weights. Time error The higher the weight of The smaller the hour, the more Time error The less important it is, the more important it is before the current moment when updating the filter weights. Time error The weight should be lower, and at the same time, the error at the current moment The error between the kth moment before the current moment The gap between It also determines to a certain extent whether the first time before the current moment should be referenced when updating the filter weight. Time error , The larger the The lower the effect of the filter weight update before the moment, The smaller the The better the filter weight update effect before the moment, the less need to use the filter weight before the current moment. Time error , then when updating the filter weights The lower the weight of As this time The weight of For basic step length Adjustments were made.
[0062] S35: Determine the filter weight at the next moment.
[0063] The filter weight at the next moment is determined based on the filter weight, error, current data and the adjusted step size at the current moment.
[0064] The filter weight at the next moment satisfies the following relationship:
[0065] ;
[0066] Where, is the filter weight at the next moment, is the filter weight at the current moment, is the adjusted step size, is the error at the current moment, is the current data at the current moment.
[0067] S36: Perform the next round of filtering.
[0068] Based on the filter weights at the next moment, the next round of filtering is performed.
[0069] S4: Perform variable switching frequency control.
[0070] Implementers can determine the convergence rules based on actual conditions, including: convergence based on a threshold of error change, for example, the error is less than 0.001 for five consecutive iterations.
[0071] The variable switching frequency of the semiconductor device is controlled according to the filtered current data.
[0072] Since the filtered output current data of the pulse converter is closer to the actual output current data of the pulse converter, the switching frequency of the semiconductor devices of the pulse converter can be controlled according to the actual condition and output requirements of the pulse converter, making the control of the switching frequency of the semiconductor devices of the pulse converter more reasonable and better ensuring the performance requirements of the pulse converter.
[0073] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A variable switching frequency control method for a semiconductor device, characterized in that: include: Preset the basic step size of the filter, perform LMS adaptive filtering on the current data through the filter, and perform variable switching frequency control of the semiconductor device according to the filtered current data; In each round of filtering iteration, the filter weights for the next moment are determined, including: The importance of each previous moment is determined based on the distribution difference between the current data at each previous moment and the current moment, satisfying the following relationship: ; In the formula, For the previous The importance of a moment, is the mean value of the current data corresponding to the error at the current moment, is the standard deviation of the current data corresponding to the error at the current moment, For the previous The mean value of the current data corresponding to the error at each moment, For the previous The standard deviation of the current data corresponding to the error at each moment is: is the vector of current data corresponding to the error at the current moment, For the previous The vector of current data corresponding to the error at each moment, is the number of errors at previous moments, is the cosine similarity function, is an exponential function with the natural constant e as the base, and the importance is negatively correlated with the distribution difference; for any moment, the difference between the current data at that moment and the expected value is determined as the error at that moment, and the cost value is determined according to the error at that moment, satisfying the following relationship: ; In the formula, is the cost value at the current moment, is the number of data points in the error, is the first data point in the error at the current moment, is the second data point in the error at the current moment, is the error at the current moment data points, is the error at the current moment data points, the cost value is positively correlated with the error; In response to the cost value at the moment being not greater than the cost value at the current moment, the error weight at the moment is determined according to the importance of the moment and the difference between the cost value at the moment and the cost value at the current moment, and the error weight satisfies the following relationship: ; In the formula, For the previous The error weight at each moment is For the previous The importance of a moment, is the cost value at the current moment, For the previous The cost value at that moment, For the previous The importance of a moment, For the previous The cost value at that moment, is the number of errors at previous moments, and the error weight is positively correlated with the importance and the difference; The basic step size of the filter is adjusted according to the error weights and errors at previous moments, and the adjusted step size is positively correlated with the error weights, errors and basic step size at previous moments; the filter weights at the next moment are determined according to the filter weights, errors, current data and adjusted step size at the current moment, and the filter weights at the next moment are positively correlated with the filter weights, errors, current data and adjusted step size at the current moment; based on the filter weights at the next moment, the next round of filtering is performed.
2. The variable switching frequency control method of a semiconductor device according to claim 1, characterized in that: The basic step size of the filter satisfies the following relationship: ; In the formula, is the basic step size of the filter, is the maximum value in the current data, is the minimum value in the current data, is the first The current value of the data point, is the number of data points in the current data, is the first A maximum value, is the first A minimum value, is the number of maxima in the current data, is the number of minimum values in the current data.
3. The variable switching frequency control method of a semiconductor device according to claim 1, characterized in that: The number of data of the current data at each moment is consistent with the number of orders of the filter.
4. The variable switching frequency control method of a semiconductor device according to claim 1, characterized in that: The adjusted step length satisfies the following relationship: , ; In the formula, is the adjusted step size, is the first data point of the adjusted step size, is the second data point of the adjusted step size, is the adjusted step size data points, is the adjusted step size data points, is the basic step size of the filter, For the previous The error weight at each moment is For the previous The error at the moment data points.
5. The variable switching frequency control method of a semiconductor device according to claim 1, characterized in that: The filter weight at the next moment satisfies the following relationship: ; In the formula, is the filter weight at the next moment, is the filter weight at the current moment, is the adjusted step size, is the error at the current moment, is the current data at the current moment.
Citation Information
Patent Citations
A converter switching frequency control method and control system
CN110943602B
Synchronization method and apparatus for grid-connected photovoltaic power generation system
CN105470988A
Filtering x minimum mean square method of adaptive active power filter with delay notch
CN115589216A
