A parallel active power filter parameter identification method based on chaos improved hunter-prey algorithm

Abstract

The application discloses a parallel active power filter parameter identification method based on a chaos improved hunter-prey algorithm, relates to the technical field of electric power, and introduces a chaos improved hunter-prey optimization algorithm (CHPO algorithm) to perform parameter identification on a parallel active power filter. Since the hunter-prey optimization algorithm can simultaneously search and develop, the CHPO algorithm is used to search and develop the prey when the position of the prey is updated, so that the prey is prevented from losing multiple properties and falling into a local optimal solution. Meanwhile, the CHPO algorithm is introduced on the basis of the hunter-prey algorithm to improve the ergodicity of the algorithm. The CHPO algorithm has sufficient exploration and development capability for solving unimodal and multimodal problems. Compared with other optimization algorithms, the performance of the CHPO algorithm is more superior, the convergence speed of a global optimal solution is greatly improved, and the CHPO algorithm has great significance for parallel active power filter parameter identification.

Description

Technical Field

[0001] This invention relates to the field of power technology, and specifically to a method for parameter identification of parallel active power filters based on a chaotic improved hunter-prey algorithm. Background Technology

[0002] The concept of microgrids was formally proposed in the early 21st century. As a new direction for power grid development that has emerged in the last decade, research on it continues. Currently, due to the increasingly frequent connection of nonlinear loads to microgrids, harmonic problems are becoming more common. Harmonics affect the normal operation of power equipment, increase losses, reduce equipment lifespan, and may even cause resonance, leading to abnormal instrument readings and erroneous operation of relay protection devices, thus significantly impacting the safety and normal operation of the power grid. Because power electronic devices themselves have nonlinear load characteristics, a large amount of harmonic and reactive current enters the power grid. Excessive harmonic pollution lowers the power factor of the grid, increases the burden on the power system, and harmonic content exceeding certain standards will seriously threaten the safe and stable operation of electrical equipment. Therefore, research on harmonic suppression in microgrids is of great practical significance in improving power quality. Parallel active power filters (SAPFs) can effectively suppress harmonics and compensate reactive power in real time, and have become one of the most commonly used power quality management devices. When a parallel active power filter (APF) performs harmonic compensation, the voltage across the DC-side capacitor C changes. For example, losses during system operation can also cause the DC-side voltage to decrease. This unstable voltage fluctuation affects the SAPF's output compensation current, leading to suboptimal harmonic compensation results. Therefore, a method is needed to control the SAPF's DC-side voltage and stabilize it near a reference value to ensure accurate harmonic compensation. Active power filters (APFs) can dynamically suppress harmonics in power electronic devices and track and compensate for constantly changing reactive power, thus finding widespread application in grid harmonic mitigation. The core component of a parallel active power filter is its control system, and the performance of the control system has a decisive impact on the performance of the active power filter. Summary of the Invention

[0003] To address the issue of slow convergence speed in existing methods, this invention aims to provide a parameter identification method for parallel active power filters based on a chaotic improved hunter-prey algorithm. This method introduces a chaotic improved hunter-prey algorithm to identify parameters of the parallel active power filter model. Since the hunter-prey algorithm can perform both searching and development simultaneously, and a chaotic algorithm is introduced on top of the hunter-prey algorithm, the ergodicity of the algorithm is improved, significantly increasing the convergence speed of the global optimal solution. This method is of great significance for parameter identification of parallel active power filter models.

[0004] The present invention specifically adopts the following technical solution:

[0005] A method for parameter identification of parallel active power filters based on a chaotic improved hunter-prey algorithm includes the following steps:

[0006] Step 1: Initialize the hunter-prey population, setting the initial total random value to... Then the objective function for all members of the population is calculated as follows: The maximum number of iterations is MaxIt, and the maximum population size is nPop;

[0007] Step 2: The position of each member in the initial group is randomly generated in the search space by the following formula;

[0008] X i =Rand(1,d).*(ub-lb)+lb

[0009] In the formula: X i lb is the position of the hunter or prey, ub is the lower bound of the variable, d is the upper bound of the variable, and d is the number of variables.

[0010] The following formula defines the lower and upper bounds of the search space. The upper and lower bounds of variables in the algorithm may be the same or different:

[0011] lb = {lb1, lb2, ..., lb} d}, ub={ub1, ub2,···,ub d}

[0012] Step 3: After generating the initial population and determining the position of each member, use the objective function O i =f(x) calculates the fitness value for each solution;

[0013] Step 4: The hunter's search mechanism consists of two steps: exploration mode and development mode; the mathematical model of the search mechanism:

[0014] x ij (t+1)=x ij (t)+0.5[(2CZP pos(j) -x ij (t))+(2(1-C)Zμ(j)-x ij (t))]

[0015] In the formula: x ij (t) represents the current hunter position, x ij (t+1) is the hunter's position in the next iteration, P posHere, μ is the prey location, Z is the calculated adaptive parameter, and C is the balance parameter between exploration and exploitation modes.

[0016]

[0017] Where it is the current iteration number;

[0018]

[0019] in and It is a random vector between [0,1], where P is the index value, R2 is a random number between 0 and 1, and IDX is a vector that satisfies the equation requirement. The index value;

[0020] Step 5: Calculate the prey P pos To determine the location of the prey, first calculate the average value μ for all locations, and then calculate the distance between each prey location and the average value μ.

[0021]

[0022] Then calculate its Euclidean distance:

[0023]

[0024] The position with the largest average distance from the location is considered the prey P. pos :

[0025]

[0026] Based on the hunting scenario, when the hunter captures prey at position x(t), the prey dies, and the hunter will move to a new prey position x(t+1) next time. Therefore, a decreasing mechanism is considered to obtain the current best search position:

[0027] kbest = round(C × N)

[0028] Where N is the number of prey, and the prey location is calculated as follows:

[0029]

[0030] At the start of the algorithm, the value of kbest, which is the number of best positions to search, is equal to N. Therefore, the last prey that is furthest from the average position μ of the prey being searched is captured by the hunter.

[0031] Step 6: Record the prey with the best fitness in the retained population;

[0032] Step 7: Assuming the safest position is the optimal global position, the prey position needs to be updated. By combining the prey search formula of the gray wolf algorithm, we get:

[0033] x ij (t+1)=0.5*(α pos(j) -A*D α +β pos(j) -A*D β )+α levi

[0034] In the formula, x ij (t+1) is the position of the next iteration, α pos(j) ,β pos(j) These are the hunter's optimal and second-best positions, It is the distance between prey and hunter. This means entering a global search. This means entering a local search. and These are the optimal and suboptimal solutions, respectively, and they improve performance during the development phase. α levi It utilizes Lévy flight to avoid the algorithm search getting trapped in local extrema;

[0035] β = 2 * rand()

[0036] σ u =((γ(1+β)*sin(π*β / 2)) / (γ((1+β / 2)*β*2 (0.5(β-1)) )) (1 / β)

[0037] u = normrnd(0, σ) u )

[0038] v = normrnd(0, 1)

[0039] α levi =0.01*u / |v| -β *(HPpos(i,j)-HPpos(i,:))

[0040]

[0041]

[0042]

[0043]

[0044] In the formula, parameter β is a random number between 0 and 2, and σ uIt is the upper limit of the number u that satisfies the normal distribution, where u and v are both numbers that satisfy the normal distribution, γ is a random distribution variable generated by the gamma function on 1+β, HPpos(i,j) is the hunter's current position, and HPpos(i,:) is the hunter's position at the previous moment.

[0045] Step 8: Development Mode: To select the hunter and the prey, combining the search formula above, we arrive at:

[0046]

[0047] In the formula, R5 is a random number between 0 and 1, and ω is an adjustment parameter;

[0048] Step 9: Determine if the termination condition is met. If it is met, output the optimal solution and end the program.

[0049] The termination condition is that the fitness function meets the requirements:

[0050]

[0051] Where e1(t) is the DC side voltage U of SAPF dc and reference voltage U ref The difference between them.

[0052] Conversely, determine whether the global optimal solution has been improved. If it has been improved, repeat steps Step 2 to Step 8; otherwise, apply equation (1) to the conventional variable Cx. i (t) is transformed by mapping, and the resulting chaotic variable x i (t) is between [0, 1], and equation (2) is used to analyze the chaotic variable x. i (t) is used to perform chaotic mapping to obtain x i (t+1), and then through equation (3) the chaotic variable x i j Perform a mapping transformation at (t+1) to obtain the regular variable Cx in the next iteration. i (t+1), repeat steps Step2 to Step8;

[0053]

[0054]

[0055]

[0056] In the formula x k It is the chaotic mapping value obtained at the previous time t.

[0057] Step 10: Based on the optimal result obtained in Step 9, and in conjunction with the SAPF control parameters, determine the PI controller parameters Kp and Ki of the DC-side voltage control loop of the parallel active filter.

[0058] Furthermore, in Step 10, based on the voltage change characteristic curve and circuit model of the parallel active power filter, a PI controller is used to control the DC-side voltage, and the actual value U of the DC-side voltage is... dc and reference value U def Subtraction yields the desired error signal ΔU dc After passing through the DC-side PI controller, the output current signal Δi is obtained. p This signal is then superimposed with the positive-sequence active component of the fundamental current in the reference current detection stage to obtain the reference current indication signal. This signal, after passing through the current tracking stage, drives the device to output a compensation current of a certain magnitude, which flows into the microgrid system, exchanging energy between the AC and DC sides, thus maintaining the DC side voltage stable near the reference voltage value. The mathematical expression is:

[0059] e1(t)=ΔU dc =U ref -U dc

[0060]

[0061] Furthermore, in Step 10, the voltage and current relationship of the active power filter is obtained according to Kirchhoff's laws:

[0062]

[0063]

[0064]

[0065] Among them, u ca u cb u cc i ca i cb i cc These are the output voltage and compensation current of the active power filter, respectively.

[0066] The beneficial effects of this invention are:

[0067] This invention introduces a chaotic improved hunter-prey algorithm for parameter identification of a parallel active power filter model. Because the chaotic improved hunter-prey algorithm can perform both local and global searches simultaneously, it has a fast convergence speed, is simple and easy to understand, and can be widely applied to function optimization problems, showing broad application prospects. At the same time, the introduction of a chaotic algorithm on the basis of the hunter-prey algorithm improves the ergodicity of the algorithm, greatly increasing the convergence speed of the global optimal solution, which is of great significance for parameter identification of the parallel active power filter model. Attached Figure Description

[0068] Figure 1 A flowchart of a parallel active power filter parameter identification method based on a chaotic improved hunter-prey algorithm;

[0069] Figure 2 This is a schematic diagram of a parallel active power filter model;

[0070] Figure 3 This is a schematic diagram illustrating the algorithm's iterative accuracy.

[0071] Figure 4 This is a comparison chart of the accuracy of CHPO and CICSO.

[0072] Figure 5 The output voltage is SAPF, and an active power filter is connected in parallel to compensate for the current and U. dc Waveform diagram.

[0073] Figure 6 This is a schematic diagram of the waveform of the harmonic source current.

[0074] Figure 7 This is a schematic diagram of the load terminal voltage, load current, and harmonic detection current waveforms. Detailed Implementation

[0075] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0076] like Figure 1-2 As shown, one embodiment of the present invention discloses a method for parameter identification of a parallel active power filter based on a chaotic improved hunter-prey algorithm, comprising the following steps:

[0077] Step 1: Initialize the hunter-prey population, setting the initial total random value to... Then the objective function for all members of the population is calculated as follows: The maximum number of iterations is MaxIt = 500, and the maximum population size is nPop = 30.

[0078] Step 2: The position of each member in the initial group is randomly generated in the search space by the following formula;

[0079] X i 1 represents the position of the hunter or prey, lb represents the minimum value (lower bound) of the variable, ub represents the maximum value (upper bound) of the variable, and d represents the number of variables (dimension).

[0080] X i =Rand(1,d).*(ub-lb)+lb

[0081] In the formula: X i 1 represents the position of the hunter or prey, lb represents the minimum value (lower bound) of the variable, ub represents the maximum value (upper bound) of the variable, and d represents the number of variables (dimension).

[0082] The following formula defines the lower and upper bounds of the search space. The upper and lower bounds of variables in the algorithm may be the same or different:

[0083] lb = {lb1, lb2, ..., lb} d}, ub={ub1, ub2,···,ub d}

[0084] Step 3: After generating the initial population and determining the position of each member, use the objective function. Calculate the fitness value for each solution. F(x) can be a maximum value (efficiency, performance, etc.) or a minimum value (cost, time, etc.).

[0085] Step 4: The hunter's search mechanism consists of two steps: exploration mode and development mode. Exploration mode: Exploration refers to the algorithm's tendency towards highly random behavior, resulting in significant variations in the solution. These significant variations in the solution prompt the hunter to further explore the search space and discover promising areas. Mathematical model of the hunter's search mechanism:

[0086] x ij (t+1)=x ij (t)+0.5[(2CZP pos(j) -x ij (t))+(2(1-C)Zμ(j)-x ij (t))]

[0087] In the formula: x ij (t) represents the current hunter position, x ij (t+1) is the hunter's position in the next iteration, P pos Here, μ is the prey location, Z is the calculated adaptive parameter, and C is the balance parameter between exploration and exploitation modes.

[0088]

[0089] Where it is the current iteration number.

[0090]

[0091] in and It is a random vector between [0,1], where P is the index value, R2 is a random number between 0 and 1, and IDX is a vector that satisfies the equation requirement. The index value.

[0092] Step 5: Calculate the prey P pos To determine the location of the prey, first calculate the average value μ for all locations, and then calculate the distance between each prey location and the average value μ.

[0093]

[0094] Then calculate its Euclidean distance:

[0095]

[0096] The position with the largest average distance from the location is considered the prey P. pos :

[0097]

[0098] If we consider the maximum distance between the search position and the average position (μ) in each iteration, the algorithm will have delayed convergence. Based on the hunting scenario, we infer that when our hunter captures prey (at position x(t), the prey dies, and the hunter will move to the new prey position x(t+1) next time. Therefore, we consider a decreasing mechanism to obtain its current best search position:

[0099] kbest = round(C × N)

[0100] Where N is the number of prey, and based on this, we can calculate the prey's location as follows:

[0101]

[0102] At the start of the algorithm, the value of kbest, which is the number of best positions to search, is equal to N. Therefore, the last prey that is furthest from the average position μ of the prey being searched is captured by the hunter.

[0103] Step 6: Record the prey with the best fitness in the retained population;

[0104] Step 7: Assuming the safest position is the optimal global position, the prey position needs to be updated. By combining the prey search formula of the gray wolf algorithm, we get:

[0105] x ij (t+1)=0.5*(α pos(j) -A*D α +β pos(j) -A*D β )+αlevi

[0106] In the formula, x ij (t+1) is the position of the next iteration, α p os(j), β pos(j) These are the hunter's optimal and second-best positions, It is the distance between prey and hunter. This means entering a global search. This means entering a local search. and These are the optimal and suboptimal solutions, respectively, and they improve performance during the development phase. α levi It utilizes Lévy flight to avoid the algorithm search getting trapped in local extrema;

[0107] β = 2 * rand()

[0108] σ u =((γ(1+β)*sin(π*β / 2)) / (γ((1+β / 2)*β*2 (0.5(β-1)) )) (1 / β)

[0109] u = normrnd(0, σ) u )

[0110] v = normrnd(0, 1)

[0111] α levi =0.01*u / v -β *(HPpos(i,j)-HPpos(i,:))

[0112]

[0113]

[0114]

[0115]

[0116] In the formula, parameter β is a random number between 0 and 2, and σ u It is the upper limit of the number u that satisfies the normal distribution, where u and v are both numbers that satisfy the normal distribution, γ is a random distribution variable generated by the gamma function on 1+β, HPpos(i,j) is the hunter's current position, and HPpos(i,:) is the hunter's position at the previous moment.

[0117] Step 8: Development Mode: To select the hunter and the prey, combining the search formula above, we arrive at:

[0118]

[0119] In the formula, R5 is a random number between 0 and 1, and ω is the adjustment parameter; Step 9: Determine whether the termination condition is met. If it is, output the optimal solution and end the program; The termination condition is that the fitness function meets the requirements:

[0120]

[0121] Where e1(t) is the DC side voltage U of SAPF dc and reference voltage U ref The difference between them.

[0122] Conversely, determine whether the global optimal solution has been improved. If it has been improved, repeat steps Step 2 to Step 8; otherwise, apply equation (1) to the conventional variable Cx. i (t) is transformed by mapping, and the resulting chaotic variable x i (t) is between [0, 1], and equation (2) is used to analyze the chaotic variable x. i (t) is used to perform chaotic mapping to obtain x i (t+1), and then through equation (3) the chaotic variable x i j Perform a mapping transformation at (t+1) to obtain the regular variable Cx in the next iteration. i (t+1), repeat steps Step2 to Step8;

[0123]

[0124]

[0125]

[0126] In the formula x k It is the chaotic mapping value obtained at the previous time t.

[0127] Furthermore, the chaotic mapping described in Step 9 is the Tent chaotic mapping. The Tent mapping, also known as the tent map, is a piecewise linear one-dimensional mapping. Compared to Logistic Regression, it has uniform power spectral density, probability density, and ideal correlation characteristics, as well as a faster iteration speed. Its mathematical expression is:

[0128] x n+1 =a-1-a|x n |,a∈(1,2)

[0129] In the formula, x n It is a number distributed in the nth random order;

[0130] When a ≤ 1, the Tent chaotic mapping is in a stable state; when a > 1, it is in a chaotic state; when a = 2, it is a central Tent mapping, and its mathematical expression is:

[0131]

[0132] Step 10: Based on the optimal position results Kp and Ki obtained in Step 9, calculate U in the circuit model respectively. dc U ref ΔU dc In order to obtain the output current signal Δi p This signal is then superimposed with the positive-sequence active component of the fundamental current in the reference current detection stage to obtain the reference current indication signal. This signal, after passing through the current tracking stage, drives the device to output a compensation current of a certain magnitude, which flows into the microgrid system. This exchange of energy between the AC and DC sides ensures that the DC side voltage remains stable near the reference voltage value. The mathematical expression is:

[0133] e1(t)=ΔU dc =U ref -U dc

[0134]

[0135] Furthermore, in Step 10, the voltage and current relationship of the active power filter is obtained according to Kirchhoff's laws:

[0136]

[0137]

[0138]

[0139] Among them, u ca u cb u cc i ca i cb i cc These are the output voltage and compensation current of the active power filter, respectively.

[0140] from Figure 3 As can be seen from this, the algorithm has found the optimal value when the number of iterations is around 170. Figure 4It can be seen that the Chaotic Improved Hunter-Prey Optimization Algorithm (CHPO algorithm) has a faster convergence speed, higher accuracy, and the advantage of not getting trapped in local optima compared to the Chaotic Improved Cat Swarm Algorithm (CICSO algorithm). It is clear that the CHPO algorithm has a better effect on parameter identification of SAPF and can make SAPF have a better effect on improving power quality.

[0141] Figure 5 The above figure shows the output voltage u of the active power filter. ca u cb u cc The middle value is the compensation current I output by SAPF. apf The one below is U. dc , Figure 6 It is the detected harmonic current Is. Figure 7 The top image shows the voltage at the load end, the middle image shows the load current, and the bottom image shows the detected harmonic current.

[0142] Finally, it should be noted that the above description is only for specific embodiments of the present invention. However, the present invention is not limited to the specific embodiments described above. Equivalent modifications and substitutions made to the present invention by those skilled in the art are also within the scope of the present invention. Therefore, all equivalent changes and modifications made without departing from the spirit and scope of the present invention are covered within the scope of the present invention.

Claims

1. A method for parameter identification of parallel active power filters based on a chaotic improved hunter-prey algorithm, characterized in that: Includes the following steps: Step 1: Initialize the hunter-prey population, setting the initial total random value to... ={ , ... }, and then calculate the objective function for all members of the population as ( )={ , , ···, The maximum number of iterations is MaxIT, and the maximum population size is nPop; Step 2: The position of each member in the initial group is randomly generated in the search space by the following formula; X i =Rand(1,d)*(ub-lb)+lb, In the formula: X i lb is the position of the hunter or prey, ub is the lower bound of the variable, d is the upper bound of the variable, and d is the number of variables. The following formula defines the lower and upper bounds of the search space, where the upper and lower bounds of variables in the algorithm may be the same or different: lb={lb1,lb2,···,lb d },ub={ub1,ub2,···,ub d }, Step 3: After generating the initial population and determining the position of each member, use the objective function. i =f( Calculate the fitness value for each solution; Step 4: The hunter's search mechanism consists of two steps: exploration mode and development mode; the mathematical model of the search mechanism: x ij (t+1)=x ij (t)+0.5[(2CZP pos(j) -x ij (t))+(2(1-C)Zμ(j)-x ij (t))], In the formula: x ij (t) represents the current hunter position, x ij (t+1) is the hunter's position in the next iteration, P pos Here, μ is the prey location, Z is the calculated adaptive parameter, and C is the balance parameter between exploration and exploitation modes. ； Where it is the current iteration number; ;IDX=(P==0); , in and It is a random vector between [0,1], where P is the index value, R2 is a random number between 0 and 1, and IDX is a vector that satisfies the equation requirement. The index value; Step 5: Calculate the prey P pos To determine the location of the prey, first calculate the average value μ for all locations, and then calculate the distance between each prey location and the average value μ. ， Then calculate its Euclidean distance: ， The position with the largest average distance from the location is considered the prey P. pos : |i is index of Max(end) sort(D euc ), Based on the hunting scenario, when the hunter captures prey at position x(t), the prey dies, and the hunter will move to a new prey position x(t+1) next time. Therefore, a decreasing mechanism is considered to obtain the best current search position: kbest=round(C N), Where N is the number of prey, and the prey location is calculated as follows: |i is sort D euc (kbest), At the start of the algorithm, the value of kbest, which is the number of best positions to search, is equal to N. Therefore, the last prey that is furthest from the average position μ of the prey being searched is captured by the hunter. Step 6: Record the prey with the best fitness in the retained population; Step 7: Assuming the safest position is the optimal global position, the prey position needs to be updated. By combining the prey search formula of the gray wolf algorithm, we get: ， In the formula, x ij (t+1) is the position of the next iteration. , These are the hunter's optimal and second-best positions, It is the distance between prey and hunter. This means entering a global search. This means entering a local search. and These are the optimal and suboptimal solutions, respectively, and they aim to improve performance during the development phase. It utilizes Lévy flight to avoid the algorithm search getting trapped in local extrema; β=2*rand() ， ， ， ， ， ， ， ， In the formula, the parameter It is a random number between 0 and 2. It is the upper bound of the number u that satisfies a normal distribution, where u and v are both numbers that satisfy a normal distribution. Using the gamma function to apply 1+ The generated random distribution variable, This is the hunter's current location. That was the hunter's position a moment ago; Step 8: Development Mode: To select the hunter and the prey, combining the search formula above, we arrive at: In the formula, R5 is a random number between 0 and 1, and ω is an adjustment parameter; Step 9: Determine if the termination condition is met. If it is met, output the optimal solution and end the program. The termination condition is that the fitness function meets the requirements: ， Where e1(t) is the DC side voltage U of SAPF dc and reference voltage U ref The difference between them; Conversely, determine whether the global optimal solution has been improved. If it has been improved, repeat steps Step 2 to Step 8; otherwise, apply equation (1) to the conventional variable Cx. i (t) Perform a mapping transformation, and the resulting chaotic variable x i (t) is between [0 1], and equation (2) is used to analyze the chaotic variable x. i (t) is used to perform chaotic mapping to obtain x. i (t+1), and then through equation (3) the chaotic variable x i j Perform a mapping transformation at (t+1) to obtain the regular variable Cx in the next iteration. i (t+1), repeat steps Step 2 through Step 8; (1) Equation (2) Equation (3) Equation In the formula It is the chaotic mapping value obtained at the previous time t; Step 10: Based on the optimal result obtained in Step 9, and in conjunction with the SAPF control parameters, determine the PI controller parameters Kp and Ki of the DC-side voltage control loop of the parallel active filter.

2. The method for parameter identification of parallel active power filters based on a chaotic improved hunter-prey algorithm according to claim 1, characterized in that: Step 10 involves using a PI controller to control the DC-side voltage based on the voltage change characteristic curve and circuit model of the parallel active power filter, thus controlling the actual DC-side voltage U. dc and reference value U def Subtraction yields the desired error signal After passing through the DC-side PI controller, the output current signal is obtained. Then, this signal is superimposed with the positive sequence active component of the fundamental current of the reference current detection stage to obtain the reference current indication signal. After passing through the current tracking stage, the drive equipment outputs a certain amount of compensation current into the microgrid system, which exchanges the energy between the AC and DC sides, so that the DC side voltage remains stable near the reference voltage value. The mathematical expression is: ， 。 3. The method for parameter identification of parallel active power filters based on a chaotic improved hunter-prey algorithm according to claim 1, characterized in that: Step 10 uses Kirchhoff's laws to derive the voltage and current relationship of the active power filter: +R·i ca +u ca -u sa =0, +R·i cb +u cb -u sb =0, +R·i cc +u cc -u sc =0, Among them, u ca u cb u cc i ca i cb i cc These are the output voltage and compensation current of the active power filter, respectively.

Images