Variable weight-hybrid decision evaluation optimization control method based on multifunctional converter

Abstract

The variable weight-hybrid decision evaluation optimization control method based on the multifunctional converter comprises the following steps: designing a phase-locked loop-free compensation instruction and a grid-connected power tracking current instruction based on a control structure of the multifunctional converter; establishing a secondary evaluation index system considering current quality and power factors; evaluating the electric energy quality by adopting a variable weight-hybrid decision evaluation model; a multi-objective collaborative optimization method considering variable weight-hybrid decision evaluation is provided, and an objective function with the minimum power quality and compensation capacity is constructed; and solving an optimal compensation coefficient under a constraint condition based on an optimization method of a multi-target artificial hummingbird algorithm updating mechanism. The invention provides a variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter, and the method has a good convergence effect. The power quality of multiple grid-connected points can be obviously improved, and the power grid treatment cost is reduced.

Description

Technical Field

[0001] The invention belongs to the field of microgrid converter control, and in particular relates to a variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter. Background Art

[0002] The access and operation of a large number of nonlinear and unbalanced loads causes the current at the common coupling point of the distribution network to contain a large number of harmonics, unbalanced and reactive components. The quality of the power grid can be improved by installing equipment such as active power filters at the grid connection point, but the operation and maintenance costs of the power grid are increased. The circuit topology of the multifunctional converter is basically the same as that of the power quality management device, which provides a hardware foundation for the use of multifunctional converters to carry out power quality management. In addition, in order to adapt to the randomness and intermittency of new energy power generation such as photovoltaic and wind power, multifunctional converters are often installed with power margin, and it is impossible to always work at full capacity. Therefore, making full use of the capacity margin of the multifunctional converter for power quality management of the power grid has become a research hotspot.

[0003] As an emerging weight analysis and evaluation method, variable weight comprehensive theory is mainly divided into penalty variable weight, incentive variable weight and hybrid variable weight. Among them, penalty variable weight and incentive variable weight methods only punish or incentivize specific indicators in the comprehensive evaluation, while hybrid variable weight can take into account the penalty and incentive of the same indicator, which is more suitable for the comprehensive evaluation of power quality. Multi-objective optimization algorithm can effectively reduce the complexity of non-inferior sorting genetic algorithm, has the advantages of fast running speed and good solution set convergence, and has been widely used in solving multi-objective problems.

[0004] At present, the use of multifunctional converters for power quality management often considers a single power quality problem. When managing multiple power quality problems, full utilization of the multifunctional converter capacity and comprehensive evaluation of the management effect are not considered enough. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter, which has a good convergence effect; it can significantly improve the power quality of multiple grid-connected points and reduce the cost of grid management.

[0006] The technical solution adopted by the present invention is:

[0007] The variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter includes the following steps:

[0008] Step 1: Based on the control structure of the multifunctional converter, design the phase-locked loop-free compensation command and the grid-connected power tracking current command;

[0009] Step 2: Establish a secondary evaluation index system that considers current quality and power factors;

[0010] Step 3: Evaluate the power quality using a variable weight-hybrid decision-making evaluation model;

[0011] Step 4: Propose a multi-objective collaborative optimization method considering variable weight-hybrid decision-making evaluation, and construct an objective function with the minimum power quality and compensation capacity;

[0012] Step 5: Based on the optimization method of the multi-objective artificial hummingbird algorithm update mechanism, solve the optimal compensation coefficient under the constraint conditions.

[0013] In the said Step 1, the control structure of the multifunctional converter is as Figure 7 shown, where the new energy includes photovoltaic or wind power; the control structure includes a boost circuit, a DC-side capacitor, three sets of single-phase full-bridge inverters, LC filtering, a boost isolation transformer, a grid-connected function and a current compensation control circuit, and a current closed-loop controller. This control topology consists of three identical single-phase systems, which are decoupled from each other and have a strong ability to carry unbalanced loads. Among them, the nonlinear and unbalanced loads are used to simulate the influence of local loads on the power quality at the PCC point.

[0014] In the said Step 1, in the grid-connected power tracking current command, the grid-connected tracking current i g,abc is obtained from the power command P g , Q g and the grid-connected point voltage U s,abc .

[0015] The PLL-free compensation command includes: commands for controlling the harmonic, negative sequence, zero sequence, and reactive power components of the grid-connected point current;

[0016] The compensation current i c,abc is obtained by decomposing the harmonic current, negative sequence current, zero sequence current, and reactive current from the filter current i L,abc , the isolation transformer current i o,abc , the distribution network current i s,abc and the grid-connected point voltage U s,abc using a multi-objective collaborative optimization algorithm considering variable weight hybrid decision-making evaluation.

[0017] First, select the constant power Park transformation matrix:

[0018]

[0019] In Equation (1), C represents the Clark transformation matrix for abc to dq transformation; θ is the phase angle.

[0020] Secondly, based on the instantaneous power theory, by projecting the positive-sequence fundamental current vector i along the direction of the positive-sequence fundamental voltage vector u, the current i pd and the current i pq, the current i pd indicates that the active component of the current is the projection length of the positive-sequence fundamental current vector i in the direction of the voltage vector u; the current i pq indicates that the reactive component of the current is the magnitude of the component of the positive-sequence fundamental current vector i perpendicular to the direction of the positive-sequence fundamental voltage vector u.

[0021]

[0022] In Equation (2), represents the d-axis component of the positive-sequence fundamental voltage, represents the q-axis component of the positive-sequence fundamental voltage, represents the d-axis component of the positive-sequence fundamental current, represents the q-axis component of the positive-sequence fundamental current.

[0023] Thus, the grid-connected power tracking current command is obtained as:

[0024]

[0025] In Equation (3): i refd represents the d-axis component of the reference command current, i refq represents the q-axis component of the reference command current, u d represents the d-axis component of the grid connection point voltage, u q represents the d-axis component of the grid connection point voltage, P g represents the active power, Q g represents the reactive power.

[0026] Combining the above detection principle of the positive-sequence fundamental active current component and the generation method of the grid-connected power tracking current command i g,abc can obtain the PLL-free compensation command.

[0027]

[0028] Equation (3) is the calculation method of the grid-connected tracking command current. The reference currents of the d and q axes of the command current are obtained through Equation (3), and then i g,abc is obtained through dq to abc conversion.

[0029]

[0030] The above formula indicates that i refq , i refd is obtained through dq / abc conversion to get i g,abc .

[0031] The calculation formula for obtaining the grid-connected power tracking current command i g,abc includes:

[0032] Equation (3) and the formula

[0033] To achieve flexible compensation while the compensation capacity is limited, compensation coefficients β are respectively multiplied on the basis of the harmonic component, negative sequence component, zero sequence component and reactive power component compensation components i , i = 1, 2, 3, 4.

[0034]

[0035] In the formula: R bk is the corresponding compensation capacity under the multi-functional converter at node k; U k is the effective value of the voltage at node k; β 1k represents the harmonic distortion rate compensation coefficient at node k, β 2k represents the negative sequence unbalance degree compensation coefficient at node k, β 3k represents the zero sequence unbalance degree compensation coefficient at node k, β 4k represents the reactive power compensation coefficient at node k, I hk represents the harmonic current component, I fk represents the negative sequence current component, I zk represents the zero sequence current component, I dk represents the reactive current component.

[0036] In step 2, aiming at the comprehensive management of the power quality of the current at the grid connection point of the multi-functional converter, a comprehensive power quality evaluation index system is established. This comprehensive evaluation index system is a secondary evaluation index system. The criterion layer in the constructed secondary evaluation index system contains 2 secondary indicators, namely the power factor B 1 and the current quality B 2 ; There are a total of 4 tertiary indicators, namely the harmonic distortion rate C 1 , the negative sequence unbalance degree C 2 , the zero sequence unbalance degree C 3 and the reactive power coefficient C 4 .

[0037] Step 3 includes the following steps:

[0038] Step 3.1: First, based on the improved analytic hierarchy process, determine the subjective weights of each index under each criterion layer. Its advantage is that it does not require consistency verification by introducing the geometric mean super-transitive theory. Specifically as follows:

[0039] Use the improved AHP method to determine the subjective weight matrix w':

[0040] a. Combine the index system and construct a judgment matrix V about the power quality index according to expert opinions.

[0041] V = [v ij m×m

[0042] ​where \(v\) ij is the influence degree value of index \(i\) on index \(j\) given by expert opinion, and \(v\) ii = 1; \(m\) is the number of indices.

[0043] b. Obtain the complementary matrix \(T\) of the judgment matrix k , and its element value-taking method is as follows:

[0044]

[0045] where \(k = 1, 2, \ldots, m\); \(v\) in is the element in the \(i\)-th row and \(n\)-th column of matrix \(V\); is the element in the \(i\)-th row of the \(k\)-th complementary matrix.

[0046] \(v\) i represents the element in the \(i\)-th row of the matrix;

[0047] c. Calculate the super-transitive approximation matrix \(X\):

[0048]

[0049] where is the element in the \(i\)-th row and \(j\)-th column of the complementary matrix \(T\) k .

[0050] d. Construct the column vector of the first column of the super-transitive approximation matrix \(X\) and perform normalization processing, and then obtain the subjective weight matrix \(w'\) corresponding to each index of the judgment matrix.

[0051] Step 3.2: Then, based on the method for determining index weights based on index correlation, determine the objective weights, specifically as follows:

[0052] Determine the objective weight matrix \(w''\) according to CRITIC

[0053] a. Normalize the index matrix formed by the original index data to obtain the standardized matrix \(A\):

[0054] \(A=(a\) ij ) t×m

[0055] where \(a\) ij is the normalized value of all indices, \(i = 1, 2, \ldots, t\); \(j = 1, 2, \ldots, m\); \(t\) is the number of objects to be evaluated.

[0056] b. Calculate the standard deviation of the indices and the correlation coefficient between the indices from the standardized matrix:

[0057]

[0058] where \(\psi\) j is the standard deviation of the \(j\)-th index, is the correlation coefficient between the i-th index and the j-th index; a i and a j are the i-th column and the j-th column of A respectively; is the average value of the elements in the j-th column of the standardized matrix A; p represents the number of observation values, and cov(a i , a j ) represents the covariance of a i and a j .

[0059] c. Calculate the index information amount φ j according to the standard deviation and the correlation coefficient, and then calculate the objective weight w″ j of each index in combination with the index information amount. Finally, obtain the objective weight matrix w″ from the objective weights of each index:

[0060]

[0061] Step 3.3: Use the coefficient of variation to combine the subjective and objective weights to ensure the comprehensiveness of the comprehensive evaluation; specifically as follows:

[0062] Perform optimal combined weighting of the subjective and objective weights based on the coefficient of variation, use the Lagrange extreme value algorithm to obtain the subjective and objective weight assignment coefficients, and finally obtain the comprehensive weight matrix w″′.

[0063] a. Construct the optimal combined weighting objective function:

[0064]

[0065] In the formula, γ and λ are the subjective and objective weight assignment coefficients respectively; Q j represents the optimal combined weighting objective function, V represents the judgment matrix V, and A represents the standardized matrix A.

[0066] b. Use the Lagrange extreme value algorithm to solve for γ and λ, and perform normalization processing to obtain γ* and λ*. Among them, the solution methods for γ and λ are as follows:

[0067]

[0068] And perform normalization processing to obtain γ* and λ*. γ* and λ* represent the weights of subjective evaluation and objective evaluation respectively, and the values are normalized to the interval (0, 1) or (-1, 1).

[0069] The specific calculation expressions of γ* and λ* are as follows:

[0070]

[0071] c. Perform the combination of the subjective and objective weight assignment coefficients to obtain the comprehensive weight value matrix w″′.

[0072] w''' = γ*×w' + λ*×w''

[0073] Step 3.4: Finally, introduce the variable weight theory to dynamically optimize and adjust the weights, so that the obtained weights are more in line with the requirements of power quality assessment. Specifically as follows:

[0074] Based on the dynamic adjustment of the comprehensive weight based on the variable weight theory, the variable weight optimized weight matrix w of each index is obtained. Combining with the actual values of the power quality indexes, the power quality comprehensive index value M under variable weight optimization is obtained.

[0075] a. Perform dynamic adjustment based on the variable weight theory to obtain the variable weight optimization coefficient matrix B, and the solution method of its elements is:

[0076]

[0077] In the formula, μ is the penalty factor, σ is the negation factor; B j represents the jth variable weight optimization coefficient matrix, C j represents the variable weight coefficient of the jth element.

[0078] b. Combine the variable weight coefficient matrix B and the comprehensive weight matrix w''' to obtain the variable weight matrix w:

[0079]

[0080] In the formula, it represents matrix multiplication.

[0081] c. Combine the actual values of the power indexes and the variable weight values to obtain the power quality comprehensive index value D:

[0082]

[0083] The said step 4 includes the following steps:

[0084] Step 4.1: Multiply each current component by the corresponding compensation coefficient β i , i = 1, 2, 3, 4; the current components are the harmonic components, negative sequence components, zero sequence components and reactive power components of the compensation current components obtained in the previous stage.

[0085]

[0086] In the formula, U k is the effective value of the node k voltage; β 1k represents the harmonic distortion rate compensation coefficient of node k, β 2k represents the negative sequence unbalance degree compensation coefficient of node k, β 3k represents the zero sequence unbalance degree compensation coefficient of node k, β 4k represents the reactive power compensation coefficient of node k, I hk represents the harmonic current component, Ifk represents the negative sequence current component, I zk represents the zero sequence current component, I dk represents the reactive current component. Step 4.2: Let the node number where the multifunctional converter is located be k. Then, the single - item power quality index C jik and the comprehensive power quality index D k at this node after compensation are as follows:

[0087]

[0088] In formula (4), C j0k is the single - item power quality index before compensation at node k; w jk represents the weight coefficient of node k, and m represents the number of indexes.

[0089] After determining the harmonic component, negative sequence component, zero sequence component and reactive component of the compensation current, calculate the compensation current i c.abc ;

[0090] The corresponding compensation capacity R bk of the multifunctional converter at node k is as follows:

[0091]

[0092] In formula (5), U k is the effective value of the voltage at node k; β 1k represents the harmonic distortion rate compensation coefficient of node k, β 2k represents the negative sequence unbalance degree compensation coefficient of node k, β 3k represents the zero sequence unbalance degree compensation coefficient of node k, β 4k represents the reactive power compensation coefficient of node k, I hk represents the harmonic current component, I fk represents the negative sequence current component, I zk represents the zero sequence current component, I dk represents the reactive current component.

[0093] When the remaining capacity R k of the multifunctional converter at node k is greater than or equal to the compensation capacity R bk , full compensation is performed for the harmonic component, negative sequence component, zero sequence component and reactive component at the node where it is located;

[0094] On the contrary, the multifunctional converter needs to optimize and select its compensation capacity according to the comprehensive power quality index D k .

[0095] Step 4.4: According to the proposed comprehensive power quality evaluation method, select the compensation coefficients β of the harmonic distortion rate, negative sequence unbalance degree, zero sequence unbalance degree, and reactive power coefficient i(i = 1, 2, 3, 4) is a decision variable to compensate the comprehensive power quality index D k The compensation capacity R of the optimal and multi-functional converter input bk with the minimum as the goal;

[0096] The objective function includes:

[0097]

[0098] In Equation (6), F 1 represents the objective function of the compensation capacity of the multi-functional converter input, and F 2 represents the objective function of the comprehensive power quality index after compensation.

[0099] The constraint conditions corresponding to the objective function are:

[0100]

[0101] In Equation (7), C 10 represents the maximum limit value allowed for the total harmonic distortion rate of the grid-side current after the multi-functional converter is input for compensation, and C 20 represents the maximum limit value allowed for the negative sequence unbalance degree, and C 30 represents the maximum limit value allowed for the zero sequence unbalance degree, and C 40 represents the maximum limit value allowed for the reactive power coefficient.

[0102] In Step 5, the multi-objective artificial hummingbird optimization algorithm is used to optimize and solve the objective function, including the following process steps:

[0103] S5.1: Initialization:

[0104] Input the objective function and constraint conditions, and set the ranges of the harmonic distortion rate, negative sequence unbalance degree, zero sequence unbalance degree, and reactive power coefficient compensation coefficient, as follows:

[0105]

[0106] Among them: C 10 represents the maximum limit value allowed for the total harmonic distortion rate of the grid-side current after the multi-functional converter is input for compensation, and C 20 represents the maximum limit value allowed for the negative sequence unbalance degree, and C 30 represents the maximum limit value allowed for the zero sequence unbalance degree, and C 40 represents the maximum limit value allowed for the reactive power coefficient.

[0107] S5.2: Set the population size, maximum number of iterations, external archive size, population position, and access list of the multi-objective artificial hummingbird optimization algorithm;

[0108] The population size N = 100, and the number of iterations is 6 times.

[0109] ν i (t + 1)= x i,tar (t)+ a·D·(x i (t)- x i,tar (t))

[0110] a ~ N(0, 1)

[0111] where x i (t) represents the position of the i-th hummingbird at time t at the food source, and x i,tar (t) represents the position of the food source that the i-th hummingbird is expected to visit, and a is the guiding factor.

[0112] S5.3: Evaluate the objective function of all hummingbird particles, add the current population to the archive, save all non-dominated solutions in the initial population to the archive, and initialize the access list;

[0113] S5.4: In each iteration, the multi-objective artificial hummingbird optimization algorithm performs guided foraging or territorial foraging with a probability of 50%. When the hummingbird performs guided foraging, it continuously updates its position on the target food source according to the access list and the dominance relationship, as follows:

[0114]

[0115] where: D (i) represents the direction vector of the i-th dimension of the hummingbird, randi([1, d]) represents generating a random integer from 1 to d, and d represents the problem dimension.

[0116] v i (t + 1)= x i,tar (t)+ aiDi(x i (t)- x i,tar (t))

[0117] where: v i (t + 1 represents the position of the i-th hummingbird at the food source at (t + 1), a represents the guiding factor, and D represents the axial flight direction vector.

[0118] a ~ N(0, 1)

[0119]

[0120] where: x i (t + 1) represents the position of the i-th hummingbird at time t + 1, x i (t) represents the position of the i-th hummingbird at time t, f(x i (t)) represents the fitness function value, f(v i (t + 1)) represents the fitness function value, vi (t + 1) represents the food source location where the i-th hummingbird is at the (t + 1)-th moment.

[0121] When foraging in the territory, the hummingbird updates its position to its neighbors within its territory; when a foraging is completed, the solution is updated based on non-dominated sorting, and the access list is updated. Migration foraging is implemented every 2n iterations, the worst front solutions are randomly initialized in the search space, and the access list is changed. Specifically as follows:

[0122] ν i (t + 1) = x i (t) + b·D·x i (t)

[0123] b ∼ N(0, 1)

[0124] S5.5: After each iteration, the non-dominated solutions in the new population are inserted into the archive. If the archive size exceeds its predefined limit, an external archive process based on the crowded distance with dynamic elimination is called, specifically as follows:

[0125]

[0126] S5.6: Repeatedly optimize and execute until the maximum number of iterations is reached.

[0127] S5.7: Obtain the archive of the optimal non-dominated solutions as the pareto optimal front, and obtain the optimal solutions of the harmonic distortion rate, negative sequence unbalance degree, zero sequence unbalance degree, and reactive power coefficient.

[0128] For a variable weight - hybrid decision evaluation optimization control method based on a multi-functional converter according to the present invention, the technical effects are as follows:

[0129] 1) The photovoltaic power generation multi-functional converter control strategy designed by the present invention can effectively adjust the harmonics, unbalances, and reactive currents of multiple nodes locally, significantly improve the power quality of multiple parallel connection points, and reduce its governance cost; it can further achieve multi-objective collaborative optimization control considering the variable weight hybrid decision evaluation of power quality, and reasonably reduce the investment in compensation capacity.

[0130] 2) The comprehensive evaluation of power quality based on variable weight - hybrid decision optimization of the present invention can objectively reflect the weights of each power quality index in real time, and is more applicable to the working conditions when the power quality index changes due to the uncertainty of new energy and the access of nonlinear loads.

[0131] 3) The Pareto optimal front of the optimization model based on the update mechanism of the multi-objective artificial hummingbird optimization algorithm of the present invention has a good convergence effect, and its optimization control effect has better characteristics in terms of economy and governance effect. Description of the Drawings

[0132] The present invention will be further described below in conjunction with the accompanying drawings and examples;

[0133] Figure 1 It is a variable-weight hybrid decision-making control structure diagram based on a multifunctional converter.

[0134] Figure 2 It is a design diagram of the reference current of a multifunctional converter with a current quality management function.

[0135] Figure 3 It is a flowchart of power quality assessment based on variable-weight hybrid decision-making.

[0136] Figure 4 It is a flowchart for solving the optimization model based on the multi-objective artificial hummingbird optimization algorithm.

[0137] Figure 5 It is the measurement result of the dq-axis current of the multifunctional converter without a phase-locked loop.

[0138] Figure 6 It is a schematic diagram of the compensation capacity of the multifunctional converter.

[0139] Figure 7 It is the control structure diagram of the multifunctional converter. Specific implementation manner

[0140] Based on the variable-weight hybrid decision-making evaluation and optimization control method of a multifunctional converter, with the multifunctional converter as the basic control structure, the reference current of the non-phase-locked loop compensation and the grid-connected tracking current reference are given, and a multi-objective collaborative optimization method based on variable-weight hybrid decision-making evaluation is proposed to better adapt to the power quality index fluctuations caused by the uncertainty of new energy and the access of nonlinear loads. A multi-objective function with the best power quality compensation effect and the smallest required compensation capacity is constructed, and an optimization algorithm based on the update mechanism of the multi-objective artificial hummingbird algorithm is used to solve the optimal capacity allocation coefficient for compensating various power quality problems. The correctness and effectiveness of the proposed method are verified through simulations under various scenarios. The method proposed by the present invention has a good convergence effect, can significantly improve the power quality of multiple grid connection points, and reduce the grid governance cost.

[0141] Figure 1 It is a variable-weight hybrid decision-making control structure diagram based on a multifunctional converter.

[0142] Based on the control structure of a multifunctional converter, a design method for compensating instruction current without a phase-locked loop is proposed. A two-level evaluation index system that comprehensively considers current quality and power factor is established. The power quality is evaluated using a variable weight-hybrid decision evaluation model, and a multi-objective collaborative optimization method considering variable weight-hybrid decision evaluation is further proposed. An objective function with the minimum power quality and compensation capacity is constructed, and an optimization algorithm based on the update mechanism of the multi-objective artificial hummingbird optimization algorithm is used to solve the optimal compensation coefficient under the constraint conditions, and its optimal control function is realized through a closed-loop current controller.

[0143] Figure 2 It is a design diagram of the instruction current for a multifunctional converter with a function of governing current quality.

[0144] Using the method of no phase-locked loop based on the synchronous rotating coordinate system, the instruction for grid-connected power tracking current and the instructions for governing the harmonic, negative sequence, zero sequence, and reactive components of the grid-connected point current are designed. Select the constant power Park transformation matrix as shown below:

[0145]

[0146] In the formula, θ = wt + θ 0 , θ 0 is the initial phase.

[0147] Based on the instantaneous power theory, the projection of the positive-sequence fundamental current vector i along the direction of the positive-sequence fundamental voltage vector u can obtain the currents i pd and i pq :

[0148]

[0149] Among them, are the d / q axis components of the positive-sequence fundamental voltage and current after low-pass filtering, respectively.

[0150] The instruction for grid-connected power tracking current is:

[0151]

[0152] Combining the above detection principle of the positive-sequence fundamental active current component and the generation algorithm of the grid-connected power tracking current instruction i g,abc a method for generating the compensation current instruction can be obtained. To achieve flexible compensation while the compensation capacity is limited, the compensation coefficient β i , i = 1, 2, 3, 4 is multiplied on the basis of each compensation component.

[0153] Figure 3 It is a flowchart of power quality evaluation based on variable weight-hybrid decision evaluation.

[0154] The uncertainty of new energy output and the access of nonlinear loads will cause serious deterioration of the power quality indicators at the grid connection point. It is difficult for traditional fixed-weight evaluation methods to ensure the accuracy and scientificity of power quality evaluation. The present invention proposes a comprehensive evaluation method of variable weight-hybrid decision optimization. First, the subjective weights of each index under each criterion layer are determined based on the improved analytic hierarchy process, and its advantage lies in that it does not require consistency verification by introducing the geometric mean hypertransfer theory. Then, the objective weights are determined based on the index weight determination method related to the index, and the coefficient of variation is used to combine the subjective and objective weights to ensure the comprehensiveness of the comprehensive evaluation. Finally, the variable weight theory is introduced to optimize the dynamic adjustment of the weights, so that the obtained weights are more in line with the evaluation requirements.

[0155] Figure 4 It is the flow chart for solving the optimization model based on the multi-objective artificial hummingbird optimization algorithm.

[0156] The optimization process of the compensation coefficient of the PV multi-functional converter based on the multi-objective artificial hummingbird optimization algorithm is as follows:

[0157] 1) Parameter initialization, input the objective function and constraint conditions, set the ranges of the harmonic distortion rate, negative sequence unbalance degree, zero sequence unbalance degree, and reactive power coefficient compensation coefficient; set the population size, maximum number of iterations, external archive size, population position, and access list of the multi-objective artificial hummingbird algorithm.

[0158] 2) Evaluate the objective function of all hummingbird particles, add the current population to the archive, save all non-dominated solutions in the initial population to the archive, and initialize the access table.

[0159] 3) In each iteration, the multi-objective artificial hummingbird algorithm performs guided foraging or territorial foraging with a probability of 50%. When the hummingbird performs guided foraging, it continuously updates its position on the target food source according to the visit table and the dominance relationship. When performing territorial foraging, the hummingbird updates its position to its neighbors within its territory. After a foraging is completed, the solution is updated based on non-dominated sorting, and the access table is updated.

[0160] 4) Migratory foraging is implemented every 2n iterations. Randomly initialize the worst front solution in the search space and change the access table. After each iteration, the non-dominated solutions in the new population are inserted into the archive. If the archive size exceeds its predefined limit, an external archive process based on the crowded distance with dynamic elimination is called.

[0161] 5) Execute repeatedly until the maximum number of iterations is reached. Output the archive with the optimal non-dominated solution as the pareto optimal front, and obtain the optimal compensation coefficients of the harmonic distortion rate, negative sequence unbalance degree, zero sequence unbalance degree, and reactive power coefficient.

[0162] Figure 5It is the measurement result of dq-axis current of the multi-functional converter without a phase-locked loop. When the converter is used for power quality management, it needs to work under various harmonic interferences. Figure 5 The measurement effect of dq-axis current without a phase-locked loop shown can well meet the stability requirements.

[0163] Figure 6 It is the schematic diagram of the compensation capacity of the multi-functional converter.

[0164] The variable weight-hybrid decision optimization method obtains the comprehensive evaluation value D for each time period. 2 They are 0.1083 / 0.0334 / 0.0146 respectively, and the compensation capacity R b2 The inputs are 4.3039 / 5.2877 / 5.6730 respectively, and the comprehensive evaluation value D obtained by combined weighting 2 They are 0.1097 / 0.0376 / 0.0151 respectively, and the compensation capacity R b2 The inputs are 4.3865 / 5.3615 / 5.7106 respectively.

Claims

1. A variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter, characterized in that The following steps are involved: Step 1: Based on the control structure of the multifunctional converter, design the phase-locked loop-free compensation command and the grid-connected power tracking current command; Step 2: Establish a secondary evaluation index system that considers current quality and power factors; Step 3: Use variable weight-hybrid decision evaluation model to evaluate power quality; Step 4: Propose a multi-objective collaborative optimization method taking into account variable weight-hybrid decision evaluation and construct an objective function that minimizes power quality and compensation capacity; Step 5: Based on the optimization method of the multi-objective artificial hummingbird algorithm update mechanism, solve the optimal compensation coefficient under the constraints.

2. The variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter according to claim 1 is characterized in that: In step 1, the control structure of the multifunctional converter includes a boost circuit, a DC side capacitor, three sets of single-phase full-bridge inverters, an LC filter, a boost isolation transformer, a grid-connected function and a current compensation control circuit, and a current closed-loop controller. The control topology consists of three completely identical single-phase systems, and the phases are decoupled from each other, and the ability to carry unbalanced loads is strong; among them, nonlinear and unbalanced loads are used to simulate the impact of local loads on the power quality of the PCC point.

3. The variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter according to claim 1 is characterized in that: In the step 1, in the grid-connected power tracking current instruction, the grid-connected tracking current i g,abc By power command P g , Q g And grid connection point voltage U s,abc get; The non-phase-locked loop compensation instructions include: instructions for controlling the harmonics, negative sequence, zero sequence and reactive components of the grid-connected point current; Compensation current i c,abc The filter current i L,abc , Isolation transformer current i o,abc , distribution network current i s,abc And grid connection point voltage U s,abc Decomposition of harmonic current, negative sequence current, zero sequence current and reactive current is performed, and the results are obtained by adopting a multi-objective collaborative optimization algorithm taking into account variable weight hybrid decision evaluation; First, select the constant power Park transformation matrix: In formula (1), C represents the Clark transformation abc to dq transformation matrix; θ is the phase angle; Secondly, based on the instantaneous power theory, the current i can be obtained by projecting the positive-sequence fundamental current vector i along the direction of the positive-sequence fundamental voltage vector u. pd and current i pq , current i pd Indicates that the active component of current is the projection length of the positive sequence fundamental current vector i in the direction of the voltage vector u; the current i pq It indicates that the reactive component of current is the component size perpendicular to the positive sequence fundamental current vector i and the positive sequence fundamental voltage vector u; In formula (2), represents the d-axis component of the positive sequence fundamental voltage, represents the q-axis component of the positive sequence fundamental voltage, represents the d-axis component of the positive sequence fundamental current, Represents the q-axis component of the positive-sequence fundamental current; The grid-connected power tracking current instruction is obtained as follows: In formula (3): i refd Indicates the d-axis component of the reference command current, i refq Indicates the q-axis component of the reference command current, u d Indicates the d-axis component of the grid-connected point voltage, u q Represents the d-axis component of the grid-connected point voltage, P g Indicates active power, Q g Indicates reactive power; Combining the above positive sequence fundamental active current component detection principle and grid-connected power tracking current instruction i g,abc A generation method can obtain a phase-locked loop-free compensation instruction; Formula (3) is the calculation method of grid-connected tracking command current. The command current d is obtained by formula (3), and the q-axis reference current is obtained by converting dq to abc. g,abc ; The above formula represents i refq ,i refd Get i through dq / abc g,abc ; In order to achieve flexible compensation when the compensation capacity is limited, the compensation coefficient β is multiplied by the compensation coefficient β on the basis of the compensation components of harmonic component, negative sequence component, zero sequence component and reactive component. i , i=1,2,3,4.

4. The variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter according to claim 1 is characterized in that: In the step 2, a comprehensive evaluation index system for power quality is established with the goal of comprehensive management of the current and power quality at the grid-connected point of the multifunctional converter. The comprehensive evaluation index system is a two-level evaluation index system. The criterion layer in the constructed two-level evaluation index system contains two second-level indicators, namely, power factor B1 and current quality B2; and a total of four third-level indicators, namely, harmonic distortion rate C1, negative sequence imbalance C2, zero sequence imbalance C3 and reactive power coefficient C4.

5. The variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter according to claim 1 is characterized in that: The step 3 comprises the following steps: Step 3.1: First, the subjective weights of each indicator under each criterion layer are determined based on the improved analytic hierarchy process, as follows: The subjective weight matrix w′ is determined using the improved AHP method: a. Combine the index system and construct the judgment matrix V about power quality index according to expert opinions; V=[v ij ] m×m In the formula, v ij is the degree to which indicator i, given by expert opinion, affects indicator j, and v ii =1; m is the number of indicators; b. Obtain the complementary matrix T of the judgment matrix k , the element value method is: Where, k = 1, 2, ..., m; v in is the element in the i-th row and n-th column of the matrix V; is the i-th row element of the k-th complementary matrix; v i Represents the i-th row element of the matrix; c. Find the super-transfer approximation matrix X: In the formula, is the complementary matrix T k The element in row i and column j; d. Construct the first column vector of the super transfer approximation matrix X and normalize it, and then obtain the subjective weight matrix w′ corresponding to each indicator of the judgment matrix; Step 3.2: Then, the objective weights are determined based on the indicator weight determination method based on indicator relevance, as follows: Determine the objective weight matrix w″ based on CRITIC a. Normalize the indicator matrix formed by the original indicator data to obtain the standardized matrix A: A=(a ij ) t×m In the formula, a ij is the normalized value of all indicators, i = 1, 2, ..., t; j = 1, 2, ..., m; t is the number of objects to be evaluated; b. Calculate the standard deviation of the indicators and the correlation coefficient between the indicators by the standardized matrix: In the formula, ψ j is the standard deviation of the jth indicator, is the correlation coefficient between the i-th indicator and the j-th indicator; a i and a j are the i-th and j-th columns of A respectively; is the average value of the elements in the jth column of the standardized matrix A; p represents the number of observations, cov(a i ,a j ) means a i and a j The covariance of c. Calculate the index information φ based on the standard deviation and correlation coefficient j , and then calculate the objective weight w″ of each indicator based on the indicator information j , and finally the objective weight matrix w″ is obtained from the objective weights of each indicator: Step 3.3: Combine the subjective and objective weights using the coefficient of variation, as follows: Based on the coefficient of variation, the subjective and objective weights are optimally combined and weighted, and the Lagrange extreme value algorithm is used to obtain the subjective and objective weight assignment coefficients, and finally the comprehensive weight matrix w″′ is obtained; a. Construct the optimal combination weighted objective function: In the formula, γ and λ are the subjective and objective weight assignment coefficients respectively; Q j represents the optimal combination weighted objective function, V represents the judgment matrix V, and A represents the standardized matrix A; b. Use the Lagrange extremum algorithm to solve γ and λ, and normalize them to get γ* and λ*. The solution methods for γ and λ are as follows: The normalized values ​​are obtained by γ* and λ*, where γ* and λ* represent the weight of subjective evaluation and the weight of objective evaluation respectively, and the values ​​are reduced to the interval of (0,1) or (-1,1); The specific calculation expressions of γ* and λ* are as follows: c. Combine the subjective and objective weight assignment coefficients to obtain a comprehensive weight value matrix w″′; w”'=γ*×w'+λ*×w” Step 3.4: Finally, variable weight theory is introduced to dynamically optimize and adjust the weights so that the obtained weights are more in line with the power quality assessment requirements; the details are as follows: Based on the dynamic adjustment of the comprehensive weight of the variable weight theory, the variable weight optimization weight matrix w of each indicator is obtained. Combined with the actual value of the power quality indicator, the comprehensive power quality indicator value M under variable weight optimization is obtained; a. Based on the variable weight theory, dynamic adjustment is performed to obtain the variable weight optimization coefficient matrix B, and the element solution method is: In the formula, μ is the penalty factor, σ is the negation factor; B j represents the jth variable weight optimization coefficient matrix, C j represents the variable weight coefficient of the jth element; b. Combine the variable weight coefficient matrix B and the comprehensive weight matrix w″′ to obtain the variable weight matrix w: Where, represents matrix multiplication; c. Combine the actual value of the power index and the variable weight value to obtain the comprehensive power quality index value D:

6. The variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter according to claim 1 is characterized in that: The step 4 comprises the following steps: Step 4.1: Multiply each current component by the corresponding compensation coefficient β i , i=1,2,3,4; the current components are the compensation current components, harmonic components, negative sequence components, zero sequence components and reactive components obtained in the previous stage; Step 4.2: Assume that the node number of the multifunctional converter is k, then the single power quality index C of the node after compensation is jik And its comprehensive power quality index D k for: In formula (4), C j0k is the single power quality index of node k before compensation; w jk represents the k-node weight coefficient, and m represents the number of indicators; Step 4.3: After determining the corresponding harmonic components, negative sequence components, zero sequence components and reactive components of the compensation current, calculate the compensation current i c.abc ; The corresponding compensation capacity R of the multifunctional converter at node k bk for: In formula (5), U k is the effective value of the voltage at node k; β 1k represents the harmonic distortion compensation coefficient of node k, β 2k represents the negative sequence imbalance compensation coefficient of node k, β 3k represents the zero-sequence unbalance compensation coefficient of node k, β 4k represents the reactive power compensation coefficient of node k, I hk represents the harmonic flow component, I fk Represents the negative sequence current component, I zk Represents the zero-sequence current component, I dk Represents the reactive current component; When the remaining capacity R of the multifunctional converter at node k k Greater than or equal to compensation capacity R bk When the harmonic component, negative sequence component, zero sequence component and reactive component of the node are fully compensated; On the contrary, the multifunctional converter needs to be based on the comprehensive power quality index D k Optimize the selection of its compensation capacity; Step 4.4: Select the compensation coefficient β for harmonic distortion rate, negative sequence unbalance, zero sequence unbalance and reactive power coefficient i (i=1,2,3,4) is the decision variable, and the comprehensive index of power quality after compensation D k Optimal and multifunctional converter input compensation capacity R bk Minimum is the goal; The objective function includes: In formula (6), F1 represents the compensation capacity objective function of the multifunctional converter, and F2 represents the comprehensive index objective function of the power quality after compensation.

7. The variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter according to claim 6 is characterized in that: The constraints corresponding to the objective function are: In formula (7), C 10 It indicates the maximum limit of the total harmonic distortion rate of the grid-side current after the multi-function converter is put into compensation, C 20 Indicates the maximum limit allowed for negative sequence imbalance, C 30 Indicates the maximum limit allowed for zero-sequence unbalance, C 40 Indicates the maximum limit allowed for the reactive power factor.

8. The variable weight-hybrid decision evaluation optimization control method based on a multifunctional converter according to claim 7 is characterized in that: In step 5, a multi-objective artificial hummingbird optimization algorithm is used to optimize and solve the objective function, including the following process steps: S5.1: Initialization: Input the objective function and constraints, and set the range of harmonic distortion rate, negative sequence imbalance, zero sequence imbalance, and reactive power compensation coefficient; S5.2: Set the population size, maximum number of iterations, external archive size, population location, and access list of the multi-objective artificial hummingbird optimization algorithm. The details are as follows: The population size is N = 100, and the number of iterations is 6; ν i (t+1)=x i,tar (t)+a·D·(x i (t)-x i,tar (t)) a~N(0,1) Among them, x i (t) represents the location of the food source where the i-th hummingbird is located at time t, x i,tar (t) represents the estimated The location of the food source visited, a is the guiding factor; S5.3: Evaluate the objective function of all hummingbird particles, add the current population to the archive, save all non-dominated solutions in the initial population to the archive, and initialize the visit list; S5.4: In each iteration, the multi-objective artificial hummingbird optimization algorithm performs guided foraging or territorial foraging with a probability of 50%; when the hummingbird performs guided foraging, it continuously updates its position on the target food source based on the visit list and dominance relationship, as follows: Where: D (i) represents the direction vector of the i-th dimension of the hummingbird, randi([1,d]) represents the generation of random integers from 1 to d, where d represents the problem dimension; v i (t+1)=x i,tar (t)+aiDi(x i (t)-x i,tar (t)) Where: v i (t+1 represents the food source location of the i-th hummingbird at time (t+1), a represents the guidance factor, and D represents the axial flight direction vector; a~N(0,1) Where: x i (t+1) represents the position of the i-th hummingbird at time t+1, x i (t) represents the position of the i-th hummingbird at time t, f(x i (t)) represents the fitness function value, f(v i (t+1)) represents the fitness function value, v i (t+1) represents the location of the food source where the i-th hummingbird is located at time (t+1); When foraging in a territory, a hummingbird will update its location to its neighbors in the territory. After a foraging session is completed, the solution is updated based on non-dominated sorting, and the access list is updated. Migration foraging is performed every 2n iterations, the worst front-end solution is randomly initialized in the search space, and the access list is changed. The details are as follows: ν i (t+1)=x i (t)+b·D·x i (t) b~N(0,1) S5.5: After each iteration, the non-dominated solutions in the new population are inserted into the archive; if the archive size exceeds its predefined limit, an external archiving procedure based on dynamically de-crowding distance is called as follows: S5.6: Repeat the optimization until the maximum number of iterations is reached; S5.7: The optimal non-dominated solution is archived as the Pareto optimal frontier, and the optimal solutions of harmonic distortion rate, negative sequence unbalance, zero sequence unbalance and reactive power coefficient are obtained.

Images