A hybrid modulation continuous adaptive PI control method for LLC resonant converter

By adopting a hybrid modulation continuous adaptive PI control method in the LLC resonant converter, the group learning factor of the particle swarm algorithm is adjusted by using electrical variable curves and wavelet transformation analysis, the problem of particle swarm algorithm being trapped in the local optimal solution is solved, and the problem of particle swarm algorithm being trapped in the local optimal solution is improved. The efficiency and performance of the system are achieved.

CN118868570BActive Publication Date: 2024-11-29BEIJING HUACHUANG QIXING MICROELECTRONICS CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411337846.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-25
Publication Date
2024-11-29
Estimated Expiration
2044-09-25

AI Technical Summary

Technical Problem

The particle swarm algorithm is prone to fall into local optimal solutions during the mixed modulation process of the LLC resonant converter, resulting in low system efficiency and poor performance.

Method used

The LLC resonant converter hybrid modulation continuous adaptive PI control method is adopted to adjust the population learning factor of the particle swarm algorithm by obtaining electrical variable curves, discrete wavelet transform decomposition spectrum diagrams, and calculating the response value and complexity of the jump in the window to realize PI Continuous adaptive modulation of the controller.

Benefits of technology

It enhances the efficiency and robustness of the algorithm, avoids local optimal solutions, improves the global exploration ability of the system, and improves the efficiency and performance of the resonant converter.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118868570B_ABST
    Figure CN118868570B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of electric variable analysis, and in particular to a mixed modulation continuous adaptive PI control method for LLC resonant converter. The method comprises: obtaining different electric variable curves to form a comprehensive output spectrum, using a window traversal to the comprehensive output spectrum and collecting the difference of adjacent data values ​​to obtain the jump response value of the window; decomposing the electric variable curve to obtain a spectrum diagram, using a sliding window traversal for each spectrum diagram, and collecting the judgment extreme values ​​of the data points to obtain the complexity of the electric variable curve; obtaining an improved group learning factor according to the complexity and the jump response value, and completing continuous adaptive modulation. The present invention enhances the global exploration ability of particles and the efficiency and robustness of the algorithm by analyzing different electric variable curves.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of electric variable analysis, and in particular to a mixed modulation continuous adaptive PI control method for an LLC resonant converter. Background Art

[0002] LLC resonant converter is a topology for efficient energy conversion. Through continuous adaptive PI control, the energy conversion efficiency can be further improved. In power electronics applications, improving energy conversion efficiency helps reduce energy waste and system heat loss, thereby improving overall system performance. And the introduction of continuous adaptive PI control enables the system to automatically adjust control parameters to adapt to different working conditions and load changes. Because the actual working environment often changes, and the system needs to maintain an efficient and stable operating state, so for power electronics systems, the reference to PI control is very important; the LLC resonant converter, hybrid modulation and adaptive PI control are combined to improve the stability of the system.

[0003] In renewable energy systems, such as solar or wind energy systems, since the output of renewable energy is usually AC, resonant converters can be used to convert this AC power into DC power for DC loads or for charging batteries, and in electric vehicle charging systems, resonant converters can be used to convert the power of DC power sources (such as batteries) into AC power for charging electric vehicles. However, continuous hybrid modulation causes the objective function of the system to have multiple local optimal solutions, that is, there are multiple relatively good solutions in the parameter space. The particle swarm algorithm is prone to fall into the local optimal solution in such a situation and cannot find the global optimal solution. Summary of the invention

[0004] In order to solve the technical problem that the particle swarm algorithm falls into a local optimal solution, the present invention provides a hybrid modulation continuous adaptive PI control method for an LLC resonant converter, and the technical solution adopted is as follows:

[0005] The present invention proposes a hybrid modulation continuous adaptive PI control method for an LLC resonant converter, the method comprising the following steps:

[0006] Obtaining electrical variable curves;

[0007] All electrical variable curves are used to form a comprehensive output spectrum, and the window of the comprehensive output spectrum and the number of mutation points on each electrical variable curve in the window are obtained; the jump value of the data point is obtained according to the data value difference between adjacent data points in each electrical variable curve; the jump response value of the window is obtained according to the jump value of the data point, the number of mutation points in the window and the data value of the data point in the window;

[0008] For each electric variable curve, a number of frequency spectra are obtained by using discrete wavelet transform decomposition, and a number of sliding windows are obtained for each frequency spectra; the judgment extreme value of the data point in the sliding window is obtained according to the data value difference of the data point in the sliding window; the complexity of the electric variable curve is obtained according to the data value difference of adjacent data points in all sliding windows and the judgment extreme value of the data point;

[0009] An improved group learning factor is obtained according to the complexity of each electric variable curve, the jump response value of the window and the preset group learning factor;

[0010] The particle swarm algorithm is used to adjust the control parameters of the PI controller according to the improved group learning factor to achieve continuous adaptive modulation of the PI controller.

[0011] Preferably, the method of forming a comprehensive output map from all electrical variable curves and obtaining a window of the comprehensive output map and the number of mutation points on each electrical variable curve in the window is:

[0012] The electrical variable curves include time current curve, time voltage curve and time power curve;

[0013] The time power curve, the time voltage curve and the time current curve are set in a coordinate system, the abscissa of the coordinate system is time, the ordinate is the data value of the data point, and the coordinate system is recorded as a comprehensive output spectrum;

[0014] A window is set in the comprehensive output graph, the length of the window is a preset value, and the width of the window is the maximum data value in the comprehensive output graph; the bottom edge of the window coincides with the horizontal axis of the coordinate system, and the window moves from the leftmost side to the rightmost side, each time it moves one unit of time, and the number of windows is obtained;

[0015] The mutation point detection algorithm is used to detect the number of mutation points of each electrical variable curve within the window.

[0016] Preferably, the method for obtaining the jump value of a data point according to the data value difference between adjacent data points in each electric variable curve is:

[0017] For each electrical variable curve, the absolute value of the difference between the data values ​​of two adjacent data points on the electrical variable curve is calculated, and the average of all the absolute values ​​of the difference is used as the jump threshold; for any data point, if the absolute value of the difference between the data value of the data point and its adjacent next data point is greater than or equal to the jump threshold, then the absolute value of the difference between the vertical coordinates of the data point and its adjacent next data point is recorded as the jump value of the data point, otherwise, the jump value is 0.

[0018] Preferably, the expression of the jump response value of the window is:

[0019] , represents the rth data point in the i-th window on the s-th electric variable curve, represents the jump value of the rth data point in the i-th window on the s-th electrical variable curve, represents the number of mutation points in the ith window on the sth electric variable curve, represents the maximum value of the ordinate in the i-th window on the s-th electric variable curve, represents the minimum value of the ordinate in the i-th window on the s-th electric variable curve, represents the sigmoid function, represents the adjustment parameter, represents the number of data points on the sth electrical variable curve, Represents the jump response value of the i-th window.

[0020] Preferably, the method of using discrete wavelet transform to decompose each electric variable curve to obtain a plurality of frequency spectra, and obtaining a plurality of sliding windows for each frequency spectra is as follows:

[0021] Each electric variable curve is used as input to use discrete wavelet transform to obtain three high-scale wavelet coefficients. Each scale corresponds to a spectrum graph. A sliding window is set in the spectrum graph. The maximum absolute value of the difference between the frequencies corresponding to adjacent peaks and troughs is used as the length of the sliding window. The highest peak represents the width of the sliding window. The bottom edge of the sliding window coincides with the horizontal axis of the spectrum graph. The sliding window moves from the leftmost side to the rightmost side. The moving step is a unit length. The number of sliding windows increases by one each time it moves.

[0022] Preferably, the method for obtaining the judgment extreme value of the data points in the sliding window according to the data value difference of the data points in the sliding window is:

[0023] For any data point in the sliding window, the data value difference between the data point and the previous data point is recorded as the first difference, and the data value difference between the data point and the next data point is recorded as the second difference. If the product of the first difference and the second difference is less than or equal to 0, the judgment extreme value of the data point is recorded as 0; if the product of the first difference and the second difference is greater than 0, the judgment extreme value of the data point is recorded as 1.

[0024] Preferably, the method for obtaining the complexity of the electric variable curve according to the data value differences of adjacent data points in all sliding windows and the determined extreme values ​​of the data points is:

[0025] Each spectrum graph corresponds to a scaling curve; the expression of complexity is:

[0026] , represents the rth data point in the vth sliding window on the uth scaling curve, represents the judgment extreme value of the rth data point in the vth sliding window on the uth scale curve, Represents the square value of the data value of the rth data point in the vth sliding window on the uth scale curve, Represents the square value of the data value of the r-1th data point in the vth sliding window on the uth scale curve, represents the maximum data value in the vth sliding window on the uth scale curve, represents the minimum data value in the vth sliding window on the uth scale curve, represents the length of the sliding window, represents the complexity of the vth sliding window on the uth scaling curve;

[0027] The complexity of the sliding window is accumulated to obtain the complexity of the electric variable curve.

[0028] Preferably, the method of accumulating the complexity of the sliding window to obtain the complexity of the electric variable curve is:

[0029] The complexity of all sliding windows on the scale curve corresponding to each scale is accumulated and recorded as the first accumulated sum, and the first accumulated sum of the scale curves of all scales corresponding to each electric variable curve is accumulated again to obtain the complexity of the electric variable curve.

[0030] Preferably, the method for obtaining the improved group learning factor according to the complexity of each electric variable curve, the jump response value of the window and the preset group learning factor is:

[0031] The complexity of each electric variable curve is added up and recorded as the first drift index, the jump response value of each window is added up and recorded as the first response factor, the product of the first drift index and the first response factor is normalized using the sigmoid function, the normalized value is multiplied by the preset group learning factor and then by the adjustment parameter to obtain the improved group learning factor.

[0032] Preferably, the method of adjusting the control parameters of the PI controller using the particle swarm algorithm according to the improved group learning factor is:

[0033] The improved group learning factor is introduced into the particle swarm algorithm to obtain the speed calculation formula, preset the fitness function, and take the minimum value of the fitness function as the control parameter of this control. Based on this, the proportional parameter and integral parameter of the PI controller are adjusted.

[0034] The present invention has the following beneficial effects: In order to solve the problems of low efficiency and poor performance of resonant converters during hybrid modulation, the present application adopts a particle swarm algorithm to perform parameter tuning on a PI controller, and then, in order to solve the problem that the particle swarm algorithm may fall into a local optimal solution, a local jump influencing factor is constructed based on window analysis and the Pettitt algorithm to judge the degree of jump of the output parameter caused by falling into the local optimal solution, and then, the complexity is constructed based on the wavelet transform, and the stability of the electric variable curve and the parameter drift characteristics are further analyzed. The group learning factor of the particle swarm algorithm is adaptively improved in real time by combining the above two indicators, so that the algorithm can enhance the global exploration ability of the particles in real time according to the jump of the output parameters caused by the local optimal solution and the parameter drift characteristics, thereby enhancing the efficiency and robustness of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 A flow chart of a hybrid modulation continuous adaptive PI control method for an LLC resonant converter provided by an embodiment of the present invention;

[0037] Figure 2 A flowchart of an implementation method of a hybrid modulation continuous adaptive PI control method for an LLC resonant converter provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of a mixed modulation continuous adaptive PI control method for an LLC resonant converter proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0039] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0040] An embodiment of a hybrid modulation continuous adaptive PI control method for LLC resonant converter:

[0041] The specific scheme of the mixed modulation continuous adaptive PI control method of LLC resonant converter provided by the present invention is described in detail below with reference to the accompanying drawings.

[0042] See also Figure 1 , which shows a flow chart of a mixed modulation continuous adaptive PI control method for an LLC resonant converter provided by an embodiment of the present invention, the method comprising the following steps:

[0043] Step S001, obtaining different electrical variable curves.

[0044] A programmable digital power supply is connected to the input end of the LLC resonant converter, and the output voltage of the power supply is adjusted by software through a computer or a digital interface to achieve uniform and continuous changes in the power supply voltage. A power analyzer is used to collect the output power of the resonant converter under different power supply voltage configurations, and a current probe and a voltage probe are used to collect the output voltage and output current of the resonant converter under different configurations.

[0045] The collected resonant converter output power, output voltage and output current are uniformly sampled at intervals of time t, where t takes an empirical value of 0.1s, and a visualization tool such as the Matplotlib library in Python is used to draw the power curve, voltage curve and current curve. The three curves use time as the horizontal axis and power, voltage and current as the vertical axis to show the law of output parameters changing over time under different power supply voltage configurations. The three images are aligned with time as the alignment unit, and the sampled output power, output current and voltage are Z-Score normalized to obtain the final time power curve, time voltage curve and time current curve. The time power curve, time voltage curve and time current curve are collectively referred to as electrical variable curves.

[0046] So far, different electrical variable curves have been obtained.

[0047] Step S002, the electric variable curve is used to form a comprehensive output spectrum, and the comprehensive output spectrum is traversed by a window and the difference between adjacent data values ​​is collected to obtain the jump response value of the window.

[0048] Resonant converters are usually designed to achieve larger current and voltage amplitudes at the resonant frequency, which is achieved by adjusting the resonant elements (such as inductors and capacitors) in the circuit. Near the resonant frequency, the amplitudes of the input current and voltage will increase significantly, which helps to improve the efficiency of the mixing modulation. The control parameter tuning of the traditional PI controller usually adopts the trial and error method or the frequency domain method, which takes a long time and the performance of the designed controller is not ideal. The frequency domain method has a large amount of calculation and needs to obtain the transfer function of the PI controller. Therefore, in order to improve the efficiency and obtain the best control performance of the PI controller, the particle swarm algorithm is used to tune the control parameters of the PI controller online. The design goal of the resonant converter is usually to achieve the maximum amplitude of current and voltage near the resonant frequency, which means that when the particle swarm algorithm tunes the control parameters of the PI controller online, it will look for those control parameters that can maximize the current and voltage amplitudes. By adjusting the PI controller, the dynamic response of the PI controller can be affected, thereby affecting the waveform of the output current and voltage.

[0049] When the particle swarm algorithm is at the local optimal solution, the PI controller will converge to a local optimal current and voltage, resulting in smaller current and voltage amplitudes or non-ideal resonance conditions in certain frequency ranges, and the output power is affected by the local optimal solution, resulting in low power efficiency or unsatisfactory spectrum distribution; conversely, when the particle swarm algorithm is at the global optimal solution, the PI controller will reach the designed resonant frequency, maximize the current and voltage amplitudes, and thus help improve the efficiency of the mixing modulation. The global optimal solution will produce higher power efficiency and more ideal spectrum characteristics, and the response of the PI controller at the global optimal solution is closer to the design goal.

[0050] All electrical variable curves are set in a coordinate system as a comprehensive output map; since the position of each particle in the particle swarm algorithm corresponds to a control parameter of a controller, when the particle falls into a local optimal solution, the control parameter will undergo a parameter mutation.

[0051] On the comprehensive output map, establish a The window of N is ,in Indicates the maximum ordinate value among the three curves. Move the window from the leftmost side of the comprehensive output spectrum to the rightmost side, with a moving step of 1 unit time. , where the bottom edge of the window coincides with the horizontal axis. This is to ensure that the window contains the complete changes of 5 unit times in the three curves. Each time the window moves, the number of windows is increased by 1, and the number of windows obtained is recorded as .

[0052] For each electric variable curve, calculate the absolute value of the difference between the ordinates of two adjacent data points on the electric variable curve, and take the average of all the absolute values ​​of the difference as the jump threshold For any data point on any curve, if the absolute value of the difference between the ordinates of the data point and its adjacent next data point is greater than or equal to the jump threshold, the absolute value of the difference between the ordinates of the data point and its adjacent next data point is recorded as the jump value of the data point; if the absolute value of the difference between the ordinates of the data point and its adjacent next data point is less than the jump threshold, the jump value of the data point is 0.

[0053] Use the Pettitt mutation point detection algorithm to detect the number of mutation points in the window of the sth electric variable curve , assuming that there are Bt data points in each electrical variable curve, the jump response value of each window is obtained according to the jump value of the data point, the number of mutation points in the window of the electrical variable curve, and the maximum and minimum values ​​of the vertical axis in the window. The formula is as follows:

[0054] ,

[0055] In the formula, represents the rth data point in the i-th window on the s-th electric variable curve, represents the jump value of the rth data point in the i-th window on the s-th electrical variable curve, represents the number of mutation points in the ith window on the sth electric variable curve, represents the maximum value of the ordinate in the i-th window on the s-th electric variable curve, represents the minimum value of the ordinate in the i-th window on the s-th electric variable curve, represents the sigmoid function, It represents the adjustment parameter, and the empirical value is 0.01. represents the number of data points on the sth electrical variable curve, Represents the jump response value of the i-th window.

[0056] in, By calculating the jump degree of two adjacent data points on the sth electric variable curve in the sliding window, it is considered that the resonant converter output jump is caused by the particle swarm algorithm falling into the local optimum and the parameter mutation after threshold screening. The larger the value, the greater the dynamic escape feature of the PI controller caused by the particle swarm algorithm falling into the local optimal solution, the stronger the locality of the particle swarm algorithm, and the jump response value. As it gets bigger; By reflecting the mutation point on the sth electric variable curve in the ith window, the stable characteristics of the output parameters of the resonant converter within the time range represented by the window are reflected. The larger the value, the more unstable the output of the resonant converter is within the time represented by the window. At this time, it is believed that the PI controller has undergone significant changes between different operating points. The unstable output will cause the objective function to have multiple local minimum values. The particle swarm algorithm is more likely to fall into the local optimal solution in the multi-mode objective function, especially when the search space is complex and multiple local minimum values ​​are close. At this time The larger the value, the more adjustment parameters are introduced. It can prevent calculation errors caused by no mutation points in the window; the denominator The jump judgment function is constrained by the range of the sth electric variable curve in the window. The larger its value is, the larger the value range of the sth electric variable curve in the window is, and the weaker the reference of the control parameter jump is. Conversely, the smaller the value is, the stronger the reference of the control parameter jump is.

[0057] At this point, the jump response value of each window is obtained.

[0058] Step S003, decomposing the electric variable curve to obtain a frequency spectrum, traversing each frequency spectrum using a sliding window, and collecting the extreme values ​​of the data points to obtain the complexity of the electric variable curve.

[0059] When the particle swarm algorithm falls into a local optimal solution, it will lead to unstable adjustment of the control parameters, making the system more complex and unstable in terms of nonlinearity and dynamic response, which in turn affects the stability of the output parameters. When the algorithm falls into a local minimum, it represents the local minimum of the objective function at that point, which means that the system may experience oscillations or instability of certain frequencies near the local minimum. Since the gradient of the objective function at these locations is small, it may be difficult for the algorithm to escape from these local minima.

[0060] In this regard, this embodiment takes each electric variable curve as input and performs discrete wavelet transform on it. The wavelet transform decomposes each electric variable curve into a low-scale and three high-scale wavelet coefficients. For the three high-scale wavelet coefficients obtained, in the spectrum diagram, the frequency value of each high-scale wavelet coefficient is in a range, and the frequency value range of the j-th high-scale wavelet coefficient in the spectrum diagram is In the spectrum graph corresponding to the j-th high scale, a sliding window is set, the length of the sliding window is represented by the absolute value of the difference between the frequencies corresponding to adjacent peaks and troughs, and the maximum absolute value of the difference between the frequencies corresponding to adjacent peaks and troughs is used as the length of the sliding window, which is to ensure that the length of the sliding window contains at least one peak and trough; the width of the sliding window is represented by the highest peak, the bottom edge of the sliding window coincides with the horizontal axis of the spectrum graph, the sliding window moves from the leftmost side to the rightmost side, the moving step is a unit length, and the number of sliding windows increases by one each time it moves.

[0061] For the sliding window, for any data point in the sliding window, the data value difference between the data point and the previous data point is recorded as the first difference, and the data value difference between the data point and the next data point is recorded as the second difference. If the product of the first difference and the second difference is less than or equal to 0, the judgment extreme value of the data point is recorded as 0; if the product of the first difference and the second difference is greater than 0, the judgment extreme value of the data point is recorded as 1.

[0062] The complexity of each sliding window is obtained according to the difference in data values ​​of adjacent data points in each sliding window in the scaling curve, the difference between the maximum and minimum values ​​in the sliding window, and the judgment extreme value of each data point. The formula is as follows:

[0063] ,

[0064] In the formula, represents the rth data point in the vth sliding window on the uth scaling curve, represents the judgment extreme value of the rth data point in the vth sliding window on the uth scale curve, Represents the square value of the data value of the rth data point in the vth sliding window on the uth scale curve, Represents the square value of the data value of the r-1th data point in the vth sliding window on the uth scale curve, represents the maximum data value in the vth sliding window on the uth scale curve, represents the minimum data value in the vth sliding window on the uth scale curve, represents the length of the sliding window, represents the complexity of the vth sliding window on the uth scaling curve.

[0065] The complexity of each sliding window of each scale on each electrical variable curve is calculated, and the complexity of all sliding windows corresponding to all scales is accumulated as the complexity of the electrical variable curve.

[0066] In the calculation of complexity, the trend and mutation in the curve are determined by dividing the scale of wavelet transform and calculating the sliding window in different scales. By defining a function to determine whether a data point in the sliding window is an extreme point, when the product of the first product and the second product is greater than 0, it means that the two adjacent data points of the rth data point are on the same side, that is, the rth data point is an extreme point. The larger its value is, the more extreme points of the wavelet coefficients in the sliding window there are. The energy difference between the rth data point and its previous adjacent data point is calculated, and the difference between the maximum and minimum values ​​of the curve in the sliding window is used as a constraint condition. The extreme point often corresponds to the inflection point, trend change or mutation point of the curve, so The larger it is, the worse the stability of the curve reflected by the high-frequency details in the curve is. It means that the cumulative energy of the wavelet transform has changed significantly, and the integration of high-frequency information reflects the trend of the u-th scale curve. The larger the value, the greater the trend change and mutation of the u-th scale curve under the high-frequency information. Finally, the windows in different frequency ranges and the sum of the three high-frequency scales are performed to reflect the comprehensive mutation and drift characteristics of each electrical variable curve.

[0067] At this point, the complexity of each electrical variable curve is obtained.

[0068] Step S004, obtaining an improved group learning factor according to the complexity of each electric variable curve, the jump response value of the window and the preset group learning factor.

[0069] For the particle swarm algorithm, the particle swarm parameters need to be initialized when used. The particle swarm parameters include: particle swarm scale, particle dimension, number of iterations, inertia weight, learning factor and iteration step range. This embodiment optimizes the learning factor in combination with the complexity of each electric variable curve and the jump response value of each window so that it will not fall into the local optimal solution. The remaining particle swarm parameters are the conventional values ​​used by the particle swarm algorithm and are not adjusted here.

[0070] The improved group learning factor is obtained according to the complexity of each electric variable curve, the jump response value of each window and the preset group learning factor. The formula is as follows:

[0071] ,

[0072] In the formula, represents the jump response value of the i-th window, represents the complexity of the sth electric variable curve, represents the preset group learning factor, Indicates the number of windows, represents the sigmoid function, represents the improved group learning factor, Represents the adjustment parameter, and the empirical value is 2.

[0073] When the particle swarm algorithm is used to perform online tuning of the control parameters of the PI controller, if the particle swarm algorithm falls into a local optimal solution due to the multi-modal output response generated by the power current and voltage connected to the resonant converter, the initial group learning factor is adaptively improved in real time. The greater the parameter jump degree, parameter chaos and parameter drift characteristics generated in the past, the greater the group learning factor of the particle swarm algorithm for online tuning of the PI controller at the latest moment, and the greater the probability that the particles will find the global optimal solution through other particles in the group, slowing down the convergence speed of the particles, making it more likely to jump out of the local optimal solution and improving the chance of global search.

[0074] At this point, an improved group learning factor is obtained.

[0075] Step S005, using the particle swarm algorithm to adjust the control parameters of the PI controller according to the improved group learning factor, so as to realize continuous adaptive modulation of the PI controller.

[0076] The improved group learning factor is introduced into the speed calculation formula of the particle swarm algorithm, and based on this, the fitness function of the particle swarm algorithm is obtained. The fitness function is the fitness function in the original particle swarm algorithm. A number of particles are preset, and each particle corresponds to a proportional parameter and an integral parameter, thereby calculating the fitness value of each particle. The output voltage of the resonant converter is used as the control target, and the error signal between the desired output voltage (reference voltage) and the actual output voltage of the resonant converter is used as the input. The control parameter corresponding to the minimum value of the fitness function is used as the control parameter of this control. Based on this, the proportional parameter (Proportional) and the integral parameter (Integral) of the PI controller are adjusted to minimize the system error or achieve other specified performance indicators, so that the control performance of the system is optimized, and finally the control signal is output. The signal is used to adjust the switching strategy of the converter (such as changing the switching frequency or adjusting the duty cycle of the switching device) to control the output voltage of the converter to reach the desired output value. Continuous adaptive modulation of the PI controller is realized. The implementation flow chart of adaptive modulation is as follows Figure 2 shown.

[0077] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0078] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A hybrid modulation continuous adaptive PI control method for LLC resonant converter, characterized in that: The method comprises the following steps: Obtaining electrical variable curves; All electrical variable curves are used to form a comprehensive output spectrum, and the window of the comprehensive output spectrum and the number of mutation points on each electrical variable curve in the window are obtained; the jump value of the data point is obtained according to the data value difference between adjacent data points in each electrical variable curve; the jump response value of the window is obtained according to the jump value of the data point, the number of mutation points in the window and the data value of the data point in the window; For each electric variable curve, a number of frequency spectra are obtained by using discrete wavelet transform decomposition, and a number of sliding windows are obtained for each frequency spectra; the judgment extreme value of the data point in the sliding window is obtained according to the data value difference of the data point in the sliding window; the complexity of the electric variable curve is obtained according to the data value difference of adjacent data points in all sliding windows and the judgment extreme value of the data point; An improved group learning factor is obtained according to the complexity of each electric variable curve, the jump response value of the window and the preset group learning factor; The particle swarm algorithm is used to adjust the control parameters of the PI controller according to the improved group learning factor to achieve continuous adaptive modulation of the PI controller. The method of forming a comprehensive output map from all electric variable curves and obtaining the window of the comprehensive output map and the number of mutation points on each electric variable curve in the window is: The electrical variable curves include time current curve, time voltage curve and time power curve; The time power curve, the time voltage curve and the time current curve are set in a coordinate system, the abscissa of the coordinate system is time, the ordinate is the data value of the data point, and the coordinate system is recorded as a comprehensive output spectrum; A window is set in the comprehensive output graph, the length of the window is a preset value, and the width of the window is the maximum data value in the comprehensive output graph; the bottom edge of the window coincides with the horizontal axis of the coordinate system, and the window moves from the leftmost side to the rightmost side, each time it moves one unit of time, and the number of windows is obtained; Use mutation point detection algorithm to detect the number of mutation points of each electric variable curve within the window; The method for obtaining the jump value of a data point according to the data value difference of adjacent data points in each electric variable curve is: For each electric variable curve, the absolute value of the difference between the data values ​​of two adjacent data points on the electric variable curve is calculated, and the average of all the absolute values ​​of the difference is used as the jump threshold; for any data point, if the absolute value of the difference between the data value of the data point and its next adjacent data point is greater than or equal to the jump threshold, the absolute value of the difference between the vertical coordinates of the data point and its next adjacent data point is recorded as the jump value of the data point, otherwise, the jump value is 0; The method of using discrete wavelet transform to decompose each electric variable curve to obtain a plurality of frequency spectra, and obtaining a plurality of sliding windows for each frequency spectra is as follows: Each electric variable curve is used as input to use discrete wavelet transform to obtain three high-scale wavelet coefficients. Each scale corresponds to a spectrum graph. A sliding window is set in the spectrum graph. The maximum absolute value of the difference between the frequencies corresponding to adjacent peaks and troughs is used as the length of the sliding window. The highest peak represents the width of the sliding window. The bottom edge of the sliding window coincides with the horizontal axis of the spectrum graph. The sliding window moves from the leftmost side to the rightmost side. The moving step is one unit length. The number of sliding windows increases by one each time it moves. The method for obtaining the judgment extreme value of the data point in the sliding window according to the data value difference of the data point in the sliding window is: For any data point in the sliding window, the data value difference between the data point and the previous data point is recorded as the first difference, and the data value difference between the data point and the next data point is recorded as the second difference. If the product of the first difference and the second difference is less than or equal to 0, the judgment extreme value of the data point is recorded as 0. If the product of the first difference and the second difference is greater than 0, the judgment extreme value of the data point is recorded as 1. The method for obtaining the complexity of the electric variable curve according to the data value differences of adjacent data points in all sliding windows and the judgment extreme values ​​of the data points is: Each spectrum graph corresponds to a scaling curve; the expression of complexity is: , represents the rth data point in the vth sliding window on the uth scaling curve, represents the judgment extreme value of the rth data point in the vth sliding window on the uth scale curve, Represents the square value of the data value of the rth data point in the vth sliding window on the uth scale curve, Represents the square value of the data value of the r-1th data point in the vth sliding window on the uth scale curve, represents the maximum data value in the vth sliding window on the uth scale curve, represents the minimum data value in the vth sliding window on the uth scale curve, represents the length of the sliding window, represents the complexity of the vth sliding window on the uth scaling curve; The complexity of the sliding window is accumulated to obtain the complexity of the electric variable curve.

2. A hybrid modulation continuous adaptive PI control method for LLC resonant converter according to claim 1, characterized in that: The expression of the jump response value of the window is: , represents the rth data point in the i-th window on the s-th electric variable curve, represents the jump value of the rth data point in the i-th window on the s-th electrical variable curve, represents the number of mutation points in the ith window on the sth electric variable curve, represents the maximum value of the ordinate in the i-th window on the s-th electric variable curve, represents the minimum value of the ordinate in the i-th window on the s-th electric variable curve, represents the sigmoid function, represents the adjustment parameter, represents the number of data points on the sth electrical variable curve, Represents the jump response value of the i-th window.

3. A hybrid modulation continuous adaptive PI control method for LLC resonant converter according to claim 1, characterized in that: The method of accumulating the complexity of the sliding window to obtain the complexity of the electric variable curve is: The complexity of all sliding windows on the scale curve corresponding to each scale is accumulated and recorded as the first accumulated sum, and the first accumulated sum of the scale curves of all scales corresponding to each electric variable curve is accumulated again to obtain the complexity of the electric variable curve.

4. A hybrid modulation continuous adaptive PI control method for LLC resonant converter according to claim 1, characterized in that: The method for obtaining the improved group learning factor according to the complexity of each electric variable curve, the jump response value of the window and the preset group learning factor is: The complexity of each electric variable curve is added up and recorded as the first drift index, the jump response value of each window is added up and recorded as the first response factor, the product of the first drift index and the first response factor is normalized using the sigmoid function, the normalized value is multiplied by the preset group learning factor and then by the adjustment parameter to obtain the improved group learning factor.

5. A hybrid modulation continuous adaptive PI control method for LLC resonant converter according to claim 1, characterized in that: The method for adjusting the control parameters of the PI controller using the particle swarm algorithm according to the improved group learning factor is: The improved group learning factor is introduced into the particle swarm algorithm to obtain the speed calculation formula, preset the fitness function, and take the minimum value of the fitness function as the control parameter of this control. Based on this, the proportional parameter and integral parameter of the PI controller are adjusted.

Citation Information

Patent Citations

  • Parameter optimization method of LLC module in DC / DC circuit

    CN109600047A

  • LLC resonant converter parameter optimization method, PI control method and device

    CN114785144A