Improved droop control method and system based on adaptive virtual impedance

Abstract

The invention relates to an improved droop control method and system based on adaptive virtual impedance, and the method specifically comprises the steps: building a parallel operation model of multiple inverters in a micro-grid system in which the inverters operate in parallel and are connected to an AC bus to supply power to a load, and building an inverter droop controller; collecting reactive power output by each inverter, and solving a self-adaptive virtual impedance value for adjusting line equivalent impedance through a radial basis function neural network according to a difference value between the reactive power output by each inverter and the reactive power output by an adjacent inverter communicating with the inverter; inputting a self-adaptive virtual impedance value into an inverter droop controller, calculating a voltage drop on the virtual impedance, and adding the voltage drop into an original output voltage reference value to obtain an output voltage reference value considering the virtual impedance; and in each sampling period, the virtual impedance value is updated according to the reactive power output difference value of each inverter, and a switching tube driving signal is generated according to the output voltage reference value considering the virtual impedance.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to an improved droop control method and system based on adaptive virtual impedance. Background Art

[0002] With the gradual advancement of the "dual carbon" goal, the "double high" characteristics of my country's power system, with a high proportion of renewable energy and a high proportion of power electronic equipment, have become more prominent. AC microgrid is an independent power generation and distribution system composed of distributed power sources, power electronic equipment and loads. It is widely used because of its flexible and efficient use of distributed energy. Droop control simulates the droop characteristics of synchronous generators during operation and realizes the automatic distribution of load power among inverters through the linear relationship between the active power-frequency and reactive power-voltage output of distributed power sources. However, the line impedance in low-voltage microgrids is not inductive, and the traditional droop control strategy cannot be directly applied through power decoupling. In addition, due to the mismatch between the impedance of the output port of each inverter to the AC bus line and the output reactive power, there is an error in the output voltage of each inverter, which makes the power distribution between inverters unreasonable and generates circulating current. The virtual impedance method is used in droop control to improve the above problems by changing the equivalent impedance of the line. However, the existing virtual impedance method has problems such as complex parameter setting and the control effect is affected by the oscillation of the inverter output power.

[0003] In the prior art CN111130375B, a method for accurately adjusting virtual impedance and a single-machine inverter and an inverter parallel system disclose a Thevenin equivalent circuit based on an inverter, obtain the no-load voltage gain at the fundamental frequency f0, the equivalent output impedance at the fundamental frequency f0, and propose an improved virtual impedance value Z for adding v0 , which can compensate for the impact of the inverter's no-load voltage gain not being 1 and accurately adjust the equivalent output impedance.

[0004] In the prior art, the equivalent output impedance of the inverter is accurately adjusted based on the Thevenin equivalent circuit of the inverter, but the parameter adjustment process is relatively complicated and the efficiency is relatively low. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention proposes an improved droop control method and system based on adaptive virtual impedance.

[0006] The technical solution of the present invention is as follows:

[0007] On the one hand, the present invention provides an improved droop control method based on adaptive virtual impedance, and the specific steps include:

[0008] In a microgrid system where inverters are connected in parallel to the AC bus to supply power to the load, a model for parallel operation of multiple inverters is established, and an inverter droop controller is built;

[0009] The reactive power output by each inverter is collected, and according to the difference between the reactive power output by each inverter and the adjacent inverter communicating with it, the adaptive virtual impedance value used to adjust the line equivalent impedance is obtained through the radial basis function neural network;

[0010] Inputting an adaptive virtual impedance value into an inverter droop controller, calculating a voltage drop on the virtual impedance, and adding the voltage drop to an original output voltage reference value to obtain an output voltage reference value taking into account the virtual impedance;

[0011] In each sampling period, the virtual impedance value is updated according to the reactive power output difference of each inverter, the switch tube drive signal is generated according to the output voltage reference value considering the virtual impedance, and the reactive power distribution of each inverter is adjusted so that the output port voltage of each inverter tends to be equal.

[0012] As a preferred implementation, the step of collecting the reactive power output by each inverter is specifically as follows:

[0013] Consider an inverter as a node and form the communication adjacency matrix U between node i and its communication node j. G :

[0014]

[0015] Where n is the number of inverters, and the element u in the matrix n，n The value is 0 or 1;

[0016] Calculate the reactive power distribution error ΔQ between node i and its adjacent node j Qi

[0017]

[0018] Among them, N i is the set of nodes connected to node i, n i is the reactive power-voltage droop coefficient of the i-th inverter, Q i Output reactive power for the i-th inverter.

[0019] As a preferred implementation, the step of obtaining the adaptive virtual impedance value for adjusting the line equivalent impedance through the radial basis function neural network is specifically as follows:

[0020] Estimating line resistance from line length Perform static virtual impedance value setting, where the virtual static resistance Used to reduce line resistance, the calculation formula is:

[0021]

[0022]

[0023] in, is the estimated line resistance; l is the line length, ρ line is the line resistance parameter per unit length;

[0024] Static virtual inductance To increase the line inductance, the calculation formula is:

[0025]

[0026] ω n =2πf n

[0027] Among them, ω n is the power system angular frequency, f n is the power system frequency.

[0028] Then the dynamic virtual inductance L of the i-th inverter is calculated through the radial basis function neural network module via and dynamic virtual resistance R via , the reactive power distribution error ΔQ Qi and reactive power sharing error derivative As the input information of the neural network module, the radial basis function neural network module is used to fit this nonlinear relationship. The specific calculation formula is as follows:

[0029] Dynamic virtual inductance L of the i-th inverter via :

[0030]

[0031] Dynamic virtual resistance R of the i-th inverter via :

[0032] R via =k QR L via

[0033] In the formula, k QR is the dynamic virtual resistance gain coefficient;

[0034] Using the dynamic virtual inductance L of the i-th inverter via and dynamic virtual resistance R via , update the adaptive virtual inductor L used to adjust the line equivalent impedance vi and adaptive virtual resistor R vi, the specific calculation formula is as follows:

[0035]

[0036] The adaptive virtual inductor L vi and adaptive virtual resistor R vi The value that constitutes the adaptive virtual impedance used to adjust the line equivalent impedance.

[0037] As a preferred implementation, the radial basis function neural network module is used to calculate the dynamic virtual inductance L of the i-th inverter. via The specific steps are:

[0038] Establish a radial basis function neural network, including input layer, hidden layer and output layer; the input layer input is the reactive power distribution error ΔQ of the i-th inverter Qi and the derivative of reactive power distribution error of the i-th inverter The output of the input layer is x1 = ΔQ Qi ,

[0039] The input of the hidden layer is:

[0040]

[0041] The output of the hidden layer is:

[0042]

[0043] in, is the center vector of the nth neuron in the hidden layer, b n is the width of the nth neuron in the hidden layer;

[0044] The input y of the output layer out for:

[0045]

[0046] Among them, w n is the weight of the nth neuron in the hidden layer;

[0047] The output of the output layer is the dynamic virtual impedance L of the i-th inverter via , the specific formula is:

[0048]

[0049] Among them, u L is the upper limit of dynamic virtual impedance;

[0050] Since the inverter output power will produce transient oscillation and then gradually adjust to a new steady state after the line equivalent impedance changes, in order to reduce the impact of the oscillation process on the neural network fitting effect, the performance index function of the neural network is selected as:

[0051]

[0052] in: is the reactive power allocation error ΔQ of the i-th inverter Qi The sliding average of the reactive power distribution error of the i-th inverter at the t-th moment is The specific calculation formula is:

[0053]

[0054] Where t represents the current sampling time, ΔQ Qi,t is the reactive power allocation error of the i-th inverter at the t-th moment, β is the sliding average weight, and the value range of β is [0,1).

[0055] As a preferred implementation, the steps of inputting an adaptive virtual impedance value into the inverter droop controller, calculating the voltage drop on the virtual impedance, adding the voltage drop to the original output voltage reference value, and obtaining the output voltage reference value considering the virtual impedance are specifically as follows:

[0056] First, calculate the voltage drop V on the adaptive virtual impedance νdi 、V νqi :

[0057]

[0058] Among them, V νdi is the voltage drop of the virtual impedance of the i-th inverter on the d-axis, V νqi is the voltage drop of the virtual impedance of the i-th inverter on the q axis, I di is the d-axis component of the ABC phase current output by the i-th inverter after Park transformation, I qi is the q-axis component obtained after Park transformation of the ABC phase current output by the i-th inverter;

[0059] Then update the output voltage reference value considering the voltage drop on the adaptive virtual impedance:

[0060]

[0061] Among them, V odi and V oqi are the d-axis and q-axis voltages input to the voltage outer loop of the voltage-current dual closed-loop control by the i-th inverter; V di and V qiThey are respectively the d-axis and q-axis components of the output voltage reference values ​​obtained by power calculation in the droop controller of the i-th inverter when the virtual impedance is ignored.

[0062] On the other hand, the present invention proposes an improved droop control system based on adaptive virtual impedance, comprising:

[0063] A droop controller building module is used to build a parallel operation model of multiple inverters and build an inverter droop controller in a microgrid system where the inverters are connected to the AC bus to supply power to the load.

[0064] The adaptive virtual impedance calculation module collects the reactive power output by each inverter, and obtains the adaptive virtual impedance value for adjusting the line equivalent impedance through a radial basis function neural network based on the difference in reactive power output between each inverter and the adjacent inverter communicating with it;

[0065] an output voltage reference value calculation module, inputting an adaptive virtual impedance value into an inverter droop controller, calculating a voltage drop on the virtual impedance, adding the voltage drop to an original output voltage reference value, and obtaining an output voltage reference value taking the virtual impedance into account;

[0066] The drive signal adjustment module updates the virtual impedance value according to the reactive power output difference of each inverter in each sampling period, generates the switch tube drive signal according to the output voltage reference value considering the virtual impedance, and adjusts the reactive power distribution of each inverter so that the output port voltage of each inverter tends to be equal.

[0067] As a preferred implementation, the step of collecting the reactive power output by each inverter is specifically as follows:

[0068] Consider an inverter as a node and form the communication adjacency matrix U between node i and its communication node j. G :

[0069]

[0070] Where n is the number of inverters, and the element u in the matrix n，n The value is 0 or 1;

[0071] Calculate the reactive power distribution error ΔQ between node i and its adjacent node j Qi

[0072]

[0073] Among them, N i is the set of nodes connected to node i, n i is the reactive power-voltage droop coefficient of the i-th inverter, Q iOutput reactive power for the i-th inverter.

[0074] As a preferred implementation, the step of obtaining the adaptive virtual impedance value for adjusting the line equivalent impedance through the radial basis function neural network is specifically as follows:

[0075] Estimating line resistance from line length Perform static virtual impedance value setting, where the virtual static resistance Used to reduce line resistance, the calculation formula is:

[0076]

[0077] in, is the estimated line resistance; l is the line length, ρ line is the line resistance parameter per unit length;

[0078] Static virtual inductance To increase the line inductance, the calculation formula is:

[0079]

[0080] Among them, ω n is the power system angular frequency, f n is the power system frequency.

[0081] Then the dynamic virtual inductance L of the i-th inverter is calculated through the radial basis function neural network module via and dynamic virtual resistance R via , the reactive power distribution error ΔQ Qi and reactive power sharing error derivative As the input information of the neural network module, the radial basis function neural network module is used to fit this nonlinear relationship. The specific calculation formula is as follows:

[0082] Dynamic virtual inductance L of the i-th inverter via :

[0083]

[0084] Dynamic virtual resistance R of the i-th inverter via :

[0085] R via =k QR L via

[0086] In the formula, k QR is the dynamic virtual resistance gain coefficient;

[0087] Using the dynamic virtual inductance L of the i-th inverter via and dynamic virtual resistance Rvia , update the adaptive virtual inductor L used to adjust the line equivalent impedance vi and adaptive virtual resistor R vi , the specific calculation formula is as follows:

[0088]

[0089] The adaptive virtual inductor L vi and adaptive virtual resistor R vi The value that constitutes the adaptive virtual impedance used to adjust the line equivalent impedance.

[0090] As a preferred implementation, the radial basis function neural network module is used to calculate the dynamic virtual inductance L of the i-th inverter. via The specific steps are:

[0091] Establish a radial basis function neural network, including input layer, hidden layer and output layer; the input layer input is the reactive power distribution error ΔQ of the i-th inverter Qi and the derivative of reactive power distribution error of the i-th inverter The output of the input layer is x1 = ΔQ Qi ,

[0092] The input of the hidden layer is:

[0093]

[0094] The output of the hidden layer is:

[0095]

[0096] in, is the center vector of the nth neuron in the hidden layer, b n is the width of the nth neuron in the hidden layer;

[0097] The input y of the output layer out for:

[0098]

[0099] Among them, w n is the weight of the nth neuron in the hidden layer;

[0100] The output of the output layer is the dynamic virtual impedance L of the i-th inverter via , the specific formula is:

[0101]

[0102] Among them, u L is the upper limit of dynamic virtual impedance;

[0103] Since the inverter output power will produce transient oscillation and then gradually adjust to a new steady state after the line equivalent impedance changes, in order to reduce the impact of the oscillation process on the neural network fitting effect, the performance index function of the neural network is selected as:

[0104]

[0105] in: is the reactive power allocation error ΔQ of the i-th inverter Qi The sliding average of the reactive power distribution error of the i-th inverter at the t-th moment is The specific calculation formula is:

[0106]

[0107] Where t represents the current sampling time, ΔQ Qi,t is the reactive power allocation error of the i-th inverter at the t-th moment, β is the sliding average weight, and the value range of β is [0,1).

[0108] As a preferred implementation, the steps of inputting an adaptive virtual impedance value into the inverter droop controller, calculating the voltage drop on the virtual impedance, adding the voltage drop to the original output voltage reference value, and obtaining the output voltage reference value considering the virtual impedance are specifically as follows:

[0109] First, calculate the voltage drop V on the adaptive virtual impedance νdi 、V νqi :

[0110]

[0111] Among them, V νdi is the voltage drop of the virtual impedance of the i-th inverter on the d-axis, V νqi is the voltage drop of the virtual impedance of the i-th inverter on the q axis, I di is the d-axis component of the ABC phase current output by the i-th inverter after Park transformation, I qi is the q-axis component obtained after Park transformation of the ABC phase current output by the i-th inverter;

[0112] Then update the output voltage reference value considering the voltage drop on the adaptive virtual impedance:

[0113]

[0114] Among them, V odi and V oqi are the d-axis and q-axis voltages input to the voltage outer loop of the voltage-current dual closed-loop control by the i-th inverter; V di and V qiThey are respectively the d-axis and q-axis components of the output voltage reference values ​​obtained by power calculation in the droop controller of the i-th inverter when the virtual impedance is ignored.

[0115] The present invention has the following beneficial effects:

[0116] 1. The present invention improves the droop control power distribution problem and circulating current problem caused by the line impedance not being inductive and the mismatch between the line impedance and the rated reactive power of the inverter in the microgrid system through adaptive virtual impedance.

[0117] 2. The present invention uses the sliding average term of reactive power distribution error as a performance indicator function, thereby reducing the influence of transient oscillation of inverter output power after the line equivalent inductance changes on the neural network fitting effect.

[0118] 3. The present invention utilizes radial basis function neural network to generate adaptive virtual impedance, which has a faster convergence speed and avoids complex parameter setting; at the same time, it reduces the influence of transient oscillation of inverter output power after the line equivalent inductance changes on the neural network fitting effect, and effectively improves the power distribution effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0119] Figure 1 It is a schematic diagram of the steps of the present invention;

[0120] Figure 2 This is a circuit diagram of a microgrid with two inverters running in parallel;

[0121] Figure 3 It is the output reactive power waveform of two inverters;

[0122] Figure 4 It is the output voltage amplitude waveform of the two inverters;

[0123] Figure 5 This is the control principle block diagram of the circuit. DETAILED DESCRIPTION

[0124] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0125] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0126] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0127] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0128] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0129] Embodiment 1:

[0130] See also Figure 1 , an improved droop control method based on adaptive virtual impedance, the specific steps include:

[0131] In a microgrid system where inverters are connected in parallel to the AC bus to supply power to the load, a model for parallel operation of multiple inverters is established, and an inverter droop controller is built;

[0132] The reactive power output by each inverter is collected, and according to the difference between the reactive power output by each inverter and the adjacent inverter communicating with it, the adaptive virtual impedance value used to adjust the line equivalent impedance is obtained through the radial basis function neural network;

[0133] Inputting an adaptive virtual impedance value into an inverter droop controller, calculating a voltage drop on the virtual impedance, and adding the voltage drop to an original output voltage reference value to obtain an output voltage reference value taking into account the virtual impedance;

[0134] In each sampling period, the virtual impedance value is updated according to the reactive power output difference of each inverter, the switch tube drive signal is generated according to the output voltage reference value considering the virtual impedance, and the reactive power distribution of each inverter is adjusted so that the output port voltage of each inverter tends to be equal.

[0135] As a preferred implementation of this embodiment, the step of collecting the reactive power output by each inverter is specifically as follows:

[0136] Consider an inverter as a node and form the communication adjacency matrix U between node i and its communication node j. G :

[0137]

[0138] Where n is the number of inverters, and the element u in the matrix n，n The value is 0 or 1;

[0139] Calculate the reactive power distribution error ΔQ between node i and its adjacent node j Qi

[0140]

[0141] Among them, N i is the set of nodes connected to node i, n i is the reactive power-voltage droop coefficient of the i-th inverter, Q i Output reactive power for the i-th inverter.

[0142] In this embodiment, the element u in the matrix i,j The value of is determined by whether there is communication between nodes i and j. If there is communication between nodes i and j, then u i,j =1, if there is no communication between node i and node j, then u i,j =0.

[0143] As a preferred implementation of this embodiment, the step of obtaining the adaptive virtual impedance value for adjusting the line equivalent impedance through the radial basis function neural network is specifically as follows:

[0144] Estimating line resistance from line length Perform static virtual impedance value setting, where the virtual static resistance Used to reduce line resistance, the calculation formula is:

[0145]

[0146]

[0147] in, is the estimated line resistance; l is the line length, ρ line is the line resistance parameter per unit length;

[0148] In this embodiment, since the load and operating state of the low-voltage microgrid are not fixed, and the actual line impedance is affected by environmental factors and line aging, the line impedance is difficult to calculate accurately; therefore, the line resistance is estimated by the line length. Set the static virtual impedance value.

[0149] Static virtual inductance To increase the line inductance, the calculation formula is:

[0150]

[0151] ωn =2πf n

[0152] Among them, ω n is the power system angular frequency, f n is the power system frequency.

[0153] In this embodiment, the static virtual inductance A positive value is used to increase the line inductance component and the static virtual resistance. The value is negative to weaken the line resistance component. Since the microgrid is a low-voltage line, the line is not inductive. Through the static virtual impedance setting, the line impedance can be made approximately inductive, realizing the decoupling of active power and reactive power control.

[0154] Then the dynamic virtual inductance L of the i-th inverter is calculated through the radial basis function neural network module via and dynamic virtual resistance R via , the reactive power distribution error ΔQ Qi and reactive power sharing error derivative As the input information of the neural network module, the radial basis function neural network module is used to fit this nonlinear relationship. The specific calculation formula is as follows:

[0155] Dynamic virtual inductance L of the i-th inverter via :

[0156]

[0157] Dynamic virtual resistance R of the i-th inverter via :

[0158] R via =k QR L via

[0159] In the formula, k QR is the dynamic virtual resistance gain coefficient;

[0160] In this embodiment, the dynamic virtual inductor L via It is used to adjust the line impedance to reduce the reactive power distribution error. It is a parameter about the reactive power distribution error ΔQ. Qi The nonlinear function of .

[0161] Using the dynamic virtual inductance L of the i-th inverter via and dynamic virtual resistance R via , update the adaptive virtual inductor L used to adjust the line equivalent impedance vi and adaptive virtual resistor R vi , the specific calculation formula is as follows:

[0162]

[0163] The adaptive virtual inductor L vi and adaptive virtual resistor R vi The value that constitutes the adaptive virtual impedance used to adjust the line equivalent impedance.

[0164] As a preferred implementation of this embodiment, the radial basis function neural network module is used to calculate the dynamic virtual inductance L of the i-th inverter. via The specific steps are:

[0165] Establish a radial basis function neural network, including input layer, hidden layer and output layer; the input layer input is the reactive power distribution error ΔQ of the i-th inverter Qi and the derivative of reactive power distribution error of the i-th inverter The output of the input layer is x1 = ΔQ Qi ,

[0166] The input of the hidden layer is:

[0167]

[0168] The output of the hidden layer is:

[0169]

[0170] in, is the center vector of the nth neuron in the hidden layer, b n is the width of the nth neuron in the hidden layer;

[0171] The input y of the output layer out for:

[0172]

[0173] Among them, w n is the weight of the nth neuron in the hidden layer;

[0174] The output of the output layer is the dynamic virtual impedance L of the i-th inverter via , the specific formula is:

[0175]

[0176] Among them, u L is the upper limit of dynamic virtual impedance;

[0177] Since the inverter output power will produce transient oscillation and then gradually adjust to a new steady state after the line equivalent impedance changes, in order to reduce the impact of the oscillation process on the neural network fitting effect, the performance index function of the neural network is selected as:

[0178]

[0179] in: is the reactive power allocation error ΔQ of the i-th inverter Qi The sliding average of the reactive power distribution error of the i-th inverter at the t-th moment is The specific calculation formula is:

[0180]

[0181] Where t represents the current sampling time, ΔQ Qi,t is the reactive power allocation error of the i-th inverter at the t-th moment, β is the sliding average weight, and the value range of β is [0,1).

[0182] As a preferred implementation of this embodiment, the steps of inputting an adaptive virtual impedance value into the inverter droop controller, calculating the voltage drop on the virtual impedance, adding the voltage drop to the original output voltage reference value, and obtaining the output voltage reference value considering the virtual impedance are specifically as follows:

[0183] First, calculate the voltage drop V on the adaptive virtual impedance νdi 、V νqi :

[0184]

[0185] Among them, V νdi is the voltage drop of the virtual impedance of the i-th inverter on the d-axis, V νqi is the voltage drop of the virtual impedance of the i-th inverter on the q axis, I di is the d-axis component of the ABC phase current output by the i-th inverter after Park transformation, I qi is the q-axis component obtained after Park transformation of the ABC phase current output by the i-th inverter;

[0186] Then update the output voltage reference value considering the voltage drop on the adaptive virtual impedance:

[0187]

[0188] Among them, V odi and V oqi are the d-axis and q-axis voltages input to the voltage outer loop of the voltage-current dual closed-loop control by the i-th inverter; V di and V qi They are respectively the d-axis and q-axis components of the output voltage reference values ​​obtained by power calculation in the droop controller of the i-th inverter when the virtual impedance is ignored.

[0189] Embodiment 2:

[0190] A simulation experiment is conducted based on the improved droop control method based on adaptive virtual impedance mentioned above:

[0191] Firstly, a power transmission model based on droop control for a multi-inverter microgrid is constructed. Figure 2 Taking the parallel operation model of two inverters as an example, the inverter is equivalent to a DC voltage source, U dci represents the output port voltage of the i-th inverter, L fi , C fi They represent the inductance and capacitance of the LC filter of the i-th inverter, U oi ,I oi They represent the output voltage and output current of the LC filter port of the i-th inverter, Z li represents the line impedance of the ith inverter connected to the AC bus, U pcc Represents the voltage of the AC bus. The i-th inverter outputs active power P i , reactive power Q i for:

[0192]

[0193] In the formula, θ i is the voltage phase output by the i-th inverter, δ i is the line impedance angle connecting the i-th inverter to the AC bus.

[0194] Since this method introduces static virtual impedance By equivalently reducing the line resistance and increasing the line inductance, the line impedance in the low-voltage microgrid can be approximated to be inductive, making δ i ≈90°; and since θ i Small, can be approximated as sinθ i ≈θ i , cosθ i ≈1, then the i-th inverter outputs active power P i , reactive power Q i Can be simplified to:

[0195]

[0196] From the simplified output power expression, we can see that P i With θ i , Q i with U oi It is a first-order linear relationship, and there is approximately no coupling between active power and reactive power. The droop control equation can be established:

[0197]

[0198] in are the output angular frequency and output voltage reference value of the i-th inverter ignoring the virtual impedance, ω ni 、E ni are the output angular frequency and output voltage rated reference values ​​of the i-th inverter, m i 、n i are the active power droop coefficient and reactive power droop coefficient of the i-th inverter, P ni , Q ni are the rated active power and reactive power of the i-th inverter respectively. It can be concluded that in order to realize that inverters of different capacities in the microgrid share the power load according to the capacity ratio, the droop coefficient must be inversely proportional to the power:

[0199]

[0200] According to the droop control equation, the output reactive power expression can be rewritten as:

[0201]

[0202] It can be seen that it is only necessary to make the line inductance proportional to the droop factor, for example:

[0203]

[0204] At this time, the condition that inverters of different capacities share reactive load according to their capacity ratio can be met.

[0205] After the adaptive virtual impedance is introduced in this method, the line impedance value can be equivalently changed by adding a virtual voltage drop in the control link, and the impedance is adjusted to meet the following requirements:

[0206] (X l1 +L v1 )Q1=(X l2 +L v2 )Q2=…=(X l3 +L v3 )Q i

[0207] In order to verify the effect of the improved droop control strategy based on adaptive virtual impedance, a MATLAB\simulink platform is built as follows: Figure 2 The two inverters shown in the figure are connected in parallel to run the simulation model. The main parameters of the simulation model are shown in Table 1:

[0208] Table 1 Main parameters of simulation model

[0209]

[0210]

[0211] The rated output reactive power of inverter 1 and inverter 2 is the same. The AC bus is connected to a reactive load of 10.5 MVar. The adaptive virtual impedance proposed by the present invention is not introduced before 0.15 s, and only the traditional droop control is used. The adaptive virtual impedance proposed by the present invention is added after 0.15 s. The output reactive power waveforms of the two inverters and the inverter output voltage amplitudes are shown as follows: Figure 3 , Figure 4 As shown, Q1 and Q2 are the reactive powers output by inverter 1 and inverter 2 respectively, and U1 and U2 are the output voltage amplitudes of inverter 1 and inverter 2 respectively.

[0212] Depend on Figure 3 , Figure 4 The simulation results show that before 0.15s, since the line impedance in the low-voltage microgrid system is not inductive and the line impedance does not match the reactive capacity of the inverter, although the rated output reactive power of the two inverters is the same and the same reactive droop coefficient is set, the actual output reactive power cannot be distributed according to the rated output reactive power ratio, and there is a difference in output voltage. After 0.15s, the adaptive virtual impedance is added using the method proposed in this paper. After the transient regulation process, after 0.35s, the output reactive power of the two inverters is distributed in a 1:1 ratio, and the actual output voltage U of the two inverters is 1:1. o1 , U o2 The droop control method based on adaptive virtual impedance proposed in this paper can effectively improve the power distribution effect, reduce the voltage difference between inverters, and achieve the effect of suppressing circulating current.

[0213] The control principle block diagram based on this embodiment is as follows Figure 5 shown.

[0214] Embodiment three:

[0215] An improved droop control system based on adaptive virtual impedance, comprising:

[0216] A droop controller building module is used to build a parallel operation model of multiple inverters and build an inverter droop controller in a microgrid system where the inverters are connected to the AC bus to supply power to the load.

[0217] The adaptive virtual impedance calculation module collects the reactive power output by each inverter, and obtains the adaptive virtual impedance value for adjusting the line equivalent impedance through a radial basis function neural network based on the difference in reactive power output between each inverter and the adjacent inverter communicating with it;

[0218] an output voltage reference value calculation module, inputting an adaptive virtual impedance value into an inverter droop controller, calculating a voltage drop on the virtual impedance, adding the voltage drop to an original output voltage reference value, and obtaining an output voltage reference value taking the virtual impedance into account;

[0219] The drive signal adjustment module updates the virtual impedance value according to the reactive power output difference of each inverter in each sampling period, generates the switch tube drive signal according to the output voltage reference value considering the virtual impedance, and adjusts the reactive power distribution of each inverter so that the output port voltage of each inverter tends to be equal.

[0220] As a preferred implementation of this embodiment, the step of collecting the reactive power output by each inverter is specifically as follows:

[0221] Consider an inverter as a node and form the communication adjacency matrix U between node i and its communication node j. G :

[0222]

[0223] Where n is the number of inverters, and the element u in the matrix n，n The value is 0 or 1;

[0224] Calculate the reactive power distribution error ΔQ between node i and its adjacent node j Qi

[0225]

[0226] Among them, N i is the set of nodes connected to node i, n i is the reactive power-voltage droop coefficient of the i-th inverter, Q i Output reactive power for the i-th inverter.

[0227] As a preferred implementation of this embodiment, the step of obtaining the adaptive virtual impedance value for adjusting the line equivalent impedance through the radial basis function neural network is specifically as follows:

[0228] Estimating line resistance from line length Perform static virtual impedance value setting, where the virtual static resistance Used to reduce line resistance, the calculation formula is:

[0229]

[0230] in, is the estimated line resistance; l is the line length, ρ line is the line resistance parameter per unit length;

[0231] Static virtual inductance To increase the line inductance, the calculation formula is:

[0232]

[0233] ω n =2πf n

[0234] Among them, ω n is the power system angular frequency, f n is the power system frequency.

[0235] Then the dynamic virtual inductance L of the i-th inverter is calculated through the radial basis function neural network module via and dynamic virtual resistance R via , the reactive power distribution error ΔQ Qi and reactive power sharing error derivative As the input information of the neural network module, the radial basis function neural network module is used to fit this nonlinear relationship. The specific calculation formula is as follows:

[0236] Dynamic virtual inductance L of the i-th inverter via :

[0237]

[0238] Dynamic virtual resistance R of the i-th inverter via :

[0239] R via =k QR L via

[0240] In the formula, k QR is the dynamic virtual resistance gain coefficient;

[0241] Using the dynamic virtual inductance L of the i-th inverter via and dynamic virtual resistance R via , update the adaptive virtual inductor L used to adjust the line equivalent impedance vi and adaptive virtual resistor R vi , the specific calculation formula is as follows:

[0242]

[0243] The adaptive virtual inductor L vi and adaptive virtual resistor R vi The value that constitutes the adaptive virtual impedance used to adjust the line equivalent impedance.

[0244] As a preferred implementation of this embodiment, the radial basis function neural network module is used to calculate the dynamic virtual inductance L of the i-th inverter. via The specific steps are:

[0245] Establish a radial basis function neural network, including input layer, hidden layer and output layer; the input layer input is the reactive power distribution error ΔQ of the i-th inverter Qi and the derivative of reactive power distribution error of the i-th inverter The output of the input layer is x1 = ΔQ Qi ,

[0246] The input of the hidden layer is:

[0247]

[0248] The output of the hidden layer is:

[0249]

[0250] Among them, is the center vector of the nth neuron in the hidden layer, b n is the width of the nth neuron in the hidden layer;

[0251] The input y of the output layer out for:

[0252]

[0253] Among them, w n is the weight of the nth neuron in the hidden layer;

[0254] The output of the output layer is the dynamic virtual impedance L of the i-th inverter via , the specific formula is:

[0255]

[0256] Among them, u L is the upper limit of dynamic virtual impedance;

[0257] Since the inverter output power will produce transient oscillation and then gradually adjust to a new steady state after the line equivalent impedance changes, in order to reduce the impact of the oscillation process on the neural network fitting effect, the performance index function of the neural network is selected as:

[0258]

[0259] in: is the reactive power allocation error ΔQ of the i-th inverter Qi The sliding average of the reactive power distribution error of the i-th inverter at the t-th moment is The specific calculation formula is:

[0260]

[0261] Where t represents the current sampling time, ΔQQi,t is the reactive power allocation error of the i-th inverter at the t-th moment, β is the sliding average weight, and the value range of β is [0,1).

[0262] As a preferred implementation of this embodiment, the steps of inputting an adaptive virtual impedance value into the inverter droop controller, calculating the voltage drop on the virtual impedance, adding the voltage drop to the original output voltage reference value, and obtaining the output voltage reference value considering the virtual impedance are specifically as follows:

[0263] First, calculate the voltage drop V on the adaptive virtual impedance νdi 、V νqi :

[0264]

[0265] Among them, V νdi is the voltage drop of the virtual impedance of the i-th inverter on the d-axis, V νqi is the voltage drop of the virtual impedance of the i-th inverter on the q axis, I di is the d-axis component of the ABC phase current output by the i-th inverter after Park transformation, I qi is the q-axis component obtained after Park transformation of the ABC phase current output by the i-th inverter;

[0266] Then update the output voltage reference value considering the voltage drop on the adaptive virtual impedance:

[0267]

[0268] Among them, V odi and V oqi are the d-axis and q-axis voltages input to the voltage outer loop of the voltage-current dual closed-loop control by the i-th inverter; V di and V qi They are respectively the d-axis and q-axis components of the output voltage reference values ​​obtained by power calculation in the droop controller of the i-th inverter when the virtual impedance is ignored.

[0269] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An improved droop control method based on adaptive virtual impedance, characterized in that: The specific steps include: In a microgrid system where inverters are connected in parallel to the AC bus to supply power to the load, a model for parallel operation of multiple inverters is established, and an inverter droop controller is built; The reactive power output by each inverter is collected, and according to the difference between the reactive power output by each inverter and the adjacent inverter communicating with it, the adaptive virtual impedance value used to adjust the line equivalent impedance is obtained through the radial basis function neural network; Inputting an adaptive virtual impedance value into an inverter droop controller, calculating a voltage drop on the virtual impedance, and adding the voltage drop to an original output voltage reference value to obtain an output voltage reference value taking into account the virtual impedance; In each sampling period, the virtual impedance value is updated according to the reactive power output difference of each inverter, the switch tube drive signal is generated according to the output voltage reference value considering the virtual impedance, and the reactive power distribution of each inverter is adjusted so that the output port voltage of each inverter tends to be equal.

2. The improved droop control method based on adaptive virtual impedance according to claim 1, characterized in that: The step of collecting the reactive power output by each inverter is specifically as follows: Consider an inverter as a node and form the communication adjacency matrix U between node i and its communication node j. G : Where n is the number of inverters, and the element u in the matrix n，n The value is 0 or 1; Calculate the reactive power distribution error ΔQ between node i and its adjacent node j Qi Among them, N i is the set of nodes connected to node i, n i is the reactive power-voltage droop coefficient of the i-th inverter, Q i Output reactive power for the i-th inverter.

3. The improved droop control method based on adaptive virtual impedance according to claim 1, characterized in that: The specific steps of obtaining the adaptive virtual impedance value for adjusting the line equivalent impedance through the radial basis function neural network are as follows: Estimating line resistance from line length Perform static virtual impedance value setting, where the virtual static resistance Used to reduce line resistance, the calculation formula is: in, is the estimated line resistance; l is the line length, ρ line is the line resistance parameter per unit length; Static virtual inductance To increase the line inductance, the calculation formula is: oh n =2πf n Among them, ω n is the power system angular frequency, f n is the power system frequency. Then the radial basis function neural network module is used to calculate the dynamic virtual inductance L of the i-th inverter via and dynamic virtual resistance R via , the reactive power distribution error ΔQ Qi and reactive power sharing error derivative As the input information of the neural network module, the radial basis function neural network module is used to fit this nonlinear relationship. The specific calculation formula is as follows: Dynamic virtual inductance L of the i-th inverter via : Dynamic virtual resistance R of the i-th inverter via : R via =k QR L via In the formula, k QR is the dynamic virtual resistance gain coefficient; Using the dynamic virtual inductance L of the i-th inverter via and dynamic virtual resistance R via , update the adaptive virtual inductor L used to adjust the line equivalent impedance vi and adaptive virtual resistor R vi , the specific calculation formula is as follows: The adaptive virtual inductor L vi and adaptive virtual resistor R vi The value that constitutes the adaptive virtual impedance used to adjust the line equivalent impedance.

4. The improved droop control method based on adaptive virtual impedance according to claim 3, characterized in that: The radial basis function neural network module is used to calculate the dynamic virtual inductance L of the i-th inverter via The specific steps are: Establish a radial basis function neural network, including input layer, hidden layer and output layer; the input layer input is the reactive power distribution error ΔQ of the i-th inverter Qi and the derivative of reactive power distribution error of the i-th inverter The output of the input layer is x1 = ΔQ Qi , The input of the hidden layer is: The output of the hidden layer is: in, is the center vector of the nth neuron in the hidden layer, b n is the width of the nth neuron in the hidden layer; The input y of the output layer out for: Among them, w n is the weight of the nth neuron in the hidden layer; The output of the output layer is the dynamic virtual impedance L of the i-th inverter via , the specific formula is: Among them, u L is the upper limit of dynamic virtual impedance; Since the inverter output power will produce transient oscillation and then gradually adjust to a new steady state after the line equivalent impedance changes, in order to reduce the impact of the oscillation process on the neural network fitting effect, the performance index function of the neural network is selected as: in: is the reactive power allocation error ΔQ of the i-th inverter Qi The sliding average of the reactive power distribution error of the i-th inverter at the t-th moment is The specific calculation formula is: Where t represents the current sampling time, ΔQ Qi,t is the reactive power allocation error of the i-th inverter at the t-th moment, β is the sliding average weight, and the value range of β is [0,1).

5. The improved droop control method based on adaptive virtual impedance according to claim 1, characterized in that: The steps of inputting an adaptive virtual impedance value into the inverter droop controller, calculating the voltage drop on the virtual impedance, adding the voltage drop to the original output voltage reference value, and obtaining the output voltage reference value considering the virtual impedance are specifically as follows: First, calculate the voltage drop V on the adaptive virtual impedance νdi 、V νqi : Among them, V νdi is the voltage drop of the virtual impedance of the i-th inverter on the d-axis, V νqi is the voltage drop of the virtual impedance of the i-th inverter on the q axis, I di is the d-axis component of the ABC phase current output by the i-th inverter after Park transformation, I qi is the q-axis component obtained after Park transformation of the ABC phase current output by the i-th inverter; Then update the output voltage reference value considering the voltage drop on the adaptive virtual impedance: Among them, V odi and V oqi are the d-axis and q-axis voltages input to the voltage outer loop of the voltage-current dual closed-loop control by the i-th inverter; V di and V qi They are respectively the d-axis and q-axis components of the output voltage reference values ​​obtained by power calculation in the droop controller of the i-th inverter when the virtual impedance is ignored.

6. An improved droop control system based on adaptive virtual impedance, characterized in that: include: A droop controller building module is used to build a parallel operation model of multiple inverters and build an inverter droop controller in a microgrid system where the inverters are connected to the AC bus to supply power to the load. The adaptive virtual impedance calculation module collects the reactive power output by each inverter, and obtains the adaptive virtual impedance value for adjusting the line equivalent impedance through a radial basis function neural network based on the difference in reactive power output between each inverter and the adjacent inverter communicating with it; an output voltage reference value calculation module, inputting an adaptive virtual impedance value into an inverter droop controller, calculating a voltage drop on the virtual impedance, adding the voltage drop to an original output voltage reference value, and obtaining an output voltage reference value taking the virtual impedance into account; The drive signal adjustment module updates the virtual impedance value according to the reactive power output difference of each inverter in each sampling period, generates the switch tube drive signal according to the output voltage reference value considering the virtual impedance, and adjusts the reactive power distribution of each inverter so that the output port voltage of each inverter tends to be equal.

7. The improved droop control system based on adaptive virtual impedance according to claim 6, characterized in that: The step of collecting the reactive power output by each inverter is specifically as follows: Consider an inverter as a node and form the communication adjacency matrix U between node i and its communication node j. G : Where n is the number of inverters, and the element u in the matrix n，n The value is 0 or 1; Calculate the reactive power distribution error ΔQ between node i and its adjacent node j Qi Among them, N i is the set of nodes connected to node i, n i is the reactive power-voltage droop coefficient of the i-th inverter, Q i Output reactive power for the i-th inverter.

8. The improved droop control system based on adaptive virtual impedance according to claim 6, characterized in that: The specific steps of obtaining the adaptive virtual impedance value for adjusting the line equivalent impedance through the radial basis function neural network are as follows: Estimating line resistance from line length Perform static virtual impedance value setting, where the virtual static resistance Used to reduce line resistance, the calculation formula is: in, is the estimated line resistance; l is the line length, ρ line is the line resistance parameter per unit length; Static virtual inductance To increase the line inductance, the calculation formula is: oh n =2πf n Among them, ω n is the power system angular frequency, f n is the power system frequency. Then the radial basis function neural network module is used to calculate the dynamic virtual inductance L of the i-th inverter via and dynamic virtual resistance R via , the reactive power distribution error ΔQ Qi and reactive power sharing error derivative As the input information of the neural network module, the radial basis function neural network module is used to fit this nonlinear relationship. The specific calculation formula is as follows: Dynamic virtual inductance L of the i-th inverter via : Dynamic virtual resistance R of the i-th inverter via : R via =k QR L via In the formula, k QR is the dynamic virtual resistance gain coefficient; Using the dynamic virtual inductance L of the i-th inverter via and dynamic virtual resistance R via , update the adaptive virtual inductor L used to adjust the line equivalent impedance vi and adaptive virtual resistor R vi , the specific calculation formula is as follows: The adaptive virtual inductor L vi and adaptive virtual resistor R vi The value that constitutes the adaptive virtual impedance used to adjust the line equivalent impedance.

9. The improved droop control system based on adaptive virtual impedance according to claim 8, characterized in that: The radial basis function neural network module is used to calculate the dynamic virtual inductance L of the i-th inverter via The specific steps are: Establish a radial basis function neural network, including input layer, hidden layer and output layer; the input layer input is the reactive power distribution error ΔQ of the i-th inverter Qi and the derivative of reactive power distribution error of the i-th inverter The output of the input layer is x1 = ΔQ Qi , The input of the hidden layer is: The output of the hidden layer is: in, is the center vector of the nth neuron in the hidden layer, b n is the width of the nth neuron in the hidden layer; The input y of the output layer out for: Among them, w n is the weight of the nth neuron in the hidden layer; The output of the output layer is the dynamic virtual impedance L of the i-th inverter via , the specific formula is: Among them, u L is the upper limit of dynamic virtual impedance; Since the inverter output power will produce transient oscillation and then gradually adjust to a new steady state after the line equivalent impedance changes, in order to reduce the impact of the oscillation process on the neural network fitting effect, the performance index function of the neural network is selected as: in: is the reactive power allocation error ΔQ of the i-th inverter Qi The sliding average of the reactive power distribution error of the i-th inverter at the t-th moment is The specific calculation formula is: Where t represents the current sampling time, ΔQ Qi,t is the reactive power allocation error of the i-th inverter at the t-th moment, β is the sliding average weight, and the value range of β is [0,1).

10. The improved droop control system based on adaptive virtual impedance according to claim 6, characterized in that: The steps of inputting an adaptive virtual impedance value into the inverter droop controller, calculating the voltage drop on the virtual impedance, adding the voltage drop to the original output voltage reference value, and obtaining the output voltage reference value considering the virtual impedance are specifically as follows: First, calculate the voltage drop V on the adaptive virtual impedance νdi 、V νqi : Among them, V νdi is the voltage drop of the virtual impedance of the i-th inverter on the d-axis, V νqi is the voltage drop of the virtual impedance of the i-th inverter on the q axis, I di is the d-axis component of the ABC phase current output by the i-th inverter after Park transformation, I qi is the q-axis component obtained after Park transformation of the ABC phase current output by the i-th inverter; Then update the output voltage reference value considering the voltage drop on the adaptive virtual impedance: Among them, V odi and V oqi are the d-axis and q-axis voltages input to the voltage outer loop of the voltage-current dual closed-loop control by the i-th inverter; V di and V qi They are respectively the d-axis and q-axis components of the output voltage reference values ​​obtained by power calculation in the droop controller of the i-th inverter when the virtual impedance is ignored.

Images