A decoupling control method for weak grid inverter system

Abstract

The present invention discloses a decoupling control method for a weak power grid inverter system. In a weak power grid, due to the characteristics of a resistor network, active power and reactive power are highly coupled. Considering the inherent coupling characteristics of the entire distribution network, by changing the controller parameters and the grid strength, the active power and reactive power of the weak distribution network can be effectively decoupled, solving the impact of the high impedance ratio in the weak distribution network. At the same time, a fuzzy neural network FNN is introduced, and the fuzzy neural network is trained using historical data to learn and model the complex nonlinear relationship of the power system. Combining fuzzy logic and fuzzy neural networks, the network can process fuzzy and uncertain information, and has adaptability and generalization capabilities. The controller dynamically adjusts the system parameters according to the real-time collected data and the output of the fuzzy neural network to achieve decoupling control and optimal distribution of power. It provides a more robust and adaptive solution for the next generation of distribution networks with high penetration.

Description

Technical Field

[0001] The present invention relates to the field of inverter control technology, and in particular to a power decoupling control method for a weak power grid inverter system. Background Art

[0002] In the context of global environmental pollution and energy crisis, distributed generation has become an attractive option for solving environmental problems due to its environmentally friendly characteristics. In addition, microgrids (MGs) are a practical solution to utilize renewable and sustainable energy and serve as a connection to the distribution network to reduce the impact of DG on the grid and coordinate the energy distribution in the system. Traditional power grids are rapidly transitioning to a more distributed structure to facilitate the widespread integration of renewable energy. These distributed energy resources (DERs) are usually integrated into the distribution network through inverters, which introduce a large number of power electronic components and greatly reduce the system inertia. The resulting low inertia can greatly reduce the strength of the grid, making it vulnerable to voltage instability and frequency oscillations. The operational challenges and instability issues of inverter-based resources in weak transmission grids have received widespread attention.

[0003] In traditional power systems, certain line losses will occur during power transmission, resulting in energy waste. Power decoupling can optimize the power distribution of the power system, reduce line losses, and improve energy efficiency. If the node voltage is unstable or the power distribution is unreasonable, it is easy to cause the system voltage deviation to be too large, and even cause instability in the power system. Power decoupling can achieve independent control of node voltage and power, improving the stability and reliability of the power system.

[0004] The control schemes (droop control or virtual synchronous generator control) of grid-connected inverters for distributed energy resources (DERs) mainly rely on the active-frequency and reactive-voltage control mechanisms of highly inductive transmission networks. In weak power grids, due to their resistance network characteristics, active power and reactive power are highly coupled. Therefore, these control schemes cannot independently and accurately regulate frequency and voltage, which results in the system dynamic and stability performance being affected when the inverter is subjected to external disturbances.

[0005] Compared with the traditional grid tracking control scheme, the grid shaping control scheme has been proven to enhance damping. By adjusting the output power of the generator, the voltage of each node in the system is kept within a suitable range. Specifically, the grid shaping control will adjust the output power of the generator according to the actual situation of the node voltage in the power system, so that the node voltage approaches the set value and remains within a suitable range. This can effectively maintain the voltage stability of the power system and prevent the system from undervoltage or overvoltage.

[0006] Fuzzy neural network (FNN) is an intelligent control method that combines fuzzy system and artificial neural network. It combines the modeling ability of fuzzy system for fuzzy logic and the learning ability of neural network for complex relationships, and has good nonlinear approximation and adaptive capabilities.

[0007] As inverter penetration levels increase, inverter power decoupling becomes increasingly important to facilitate continued reliable system operation and optimal dynamic performance. Summary of the invention

[0008] Purpose of the invention: In order to solve the problems existing in the prior art, the present invention discloses a method for decoupling control of a weak power grid inverter system. The control method combining grid forming control with fuzzy neural network can achieve accurate prediction and intelligent control, and improve the stability, reliability and economy of the power system.

[0009] Technical solution: The present invention discloses a decoupling control method for a weak grid inverter system, comprising the following steps:

[0010] Step 1: Based on the basic principle of droop control of grid-connected inverter, a basic model of droop control active power and reactive power is established;

[0011] Step 2: Based on the power model obtained in step 1, the impedance and resistance-inductance characteristics in the transmission network are considered to simplify the power model, and the inverter droop control active-frequency (Pf) and reactive-voltage (QV) models are obtained;

[0012] Step 3: Design a grid power decoupling controller to decouple the active power and reactive power in the system according to the characteristics and requirements of the power system, and realize independent grid frequency and voltage control to achieve stable and optimized scheduling of the system;

[0013] Step 4: Based on the grid power decoupling controller proposed in step 3, an enhanced control mechanism is proposed and a small signal model of the controller is derived;

[0014] Step 5: Based on the power decoupling controller proposed in step 4, design the fuzzy neural network FNN control rule, use historical data to train the fuzzy neural network, and improve the accuracy and robustness of the model through repeated training and optimization;

[0015] Step 6: Combine the fuzzy neural network FNN with the grid power decoupling controller as the core part of the control system. Use the fuzzy neural network to learn and predict the state of the power system, and adjust the parameters in the power decoupling controller according to the prediction results to achieve intelligent control of the system.

[0016] Furthermore, in step 1, P and Q transmitted from the inverter to the grid are represented by the voltage at the inverter terminal and the grid connection point PCC and the impedance between the two points. The basic model of active power and reactive power based on droop control is:

[0017]

[0018]

[0019] Among them, P and Q are the active power and reactive power output by distributed energy DER respectively; E and V are the inverter terminal voltage and PCC point voltage respectively. It is expressed as the phase angle difference between the inverter terminal voltage and the PCC voltage, and Z∠θ represents the output impedance of the distributed energy DER system, including the impedance of the filter and the output line.

[0020] Furthermore, in step 2, in the high voltage transmission network, the transmission line has a high impedance ratio, θ≈90°, and the phase angle difference Ignore, so there is The above formula is simplified to:

[0021]

[0022]

[0023] When the PCC voltage and line impedance are fixed, it is observed that P and The linear relationship between them and between Q and E is expressed as:

[0024] f * =f0-m(P mea -P0) (5)

[0025] E * =E0-n(Q mea -Q0) (6)

[0026] In the formula, f * 、E * is the reference value of the inverter terminal voltage frequency and amplitude; f0 and E0 are their rated values; P mea , Q mea are the output active power and reactive power; P0 and Q0 are the rated active power and reactive power; m and n are the active-frequency and reactive-voltage droop coefficients.

[0027] Furthermore, the grid forming power decoupling controller designed in step 3 is:

[0028] The weak grid inverter system includes a weak distribution network, a DER filter and an export distribution line. The equivalent circuit of the system consists of two time-varying elements: the time-varying impedance Z s ∠θ s and a time-varying voltage source Z s ∠θ s It represents the total system impedance obtained by the inverter, which is the sum of the system-side filter impedance, the output distribution line impedance, and the equivalent impedance of the weak distribution network. The system-side filter impedance and the output distribution line impedance are determined according to the parameters of the DER inverter and the distribution line. The equivalent impedance of the weak distribution network is identified online by monitoring the voltage and current changes of the PCC during load change events:

[0029]

[0030] in, and respectively Represents the pre-operational measurements of the voltage and current phases, and Indicates the measured value after operation, by applying KVL, Derivation of equivalent grid voltage and Z g ∠θ g :

[0031]

[0032] According to the equivalent time-varying system circuit, the combined system impedance of the output impedance and the grid impedance is determined by the above formula:

[0033] Z s ∠θ s =Z l ∠θ l +Z g ∠θ g (9)

[0034] in,

[0035]

[0036]

[0037] Due to its resistive nature, the impedance Z s ∠θ s The assumption that sinθ≈1, cosθ≈0 cannot be used to simplify. In the distribution network, the inverter terminal voltage Equivalent voltage to the grid The phase difference between Very small, so The assumption is valid, therefore, the outputs P and Q of DER in the weak distribution network are expressed as:

[0038]

[0039]

[0040] Where P s , Q s are the active power and reactive power of DER, V g is the voltage amplitude of the equivalent voltage source of the power grid;

[0041] The system coupling characteristics are expressed by matrix using the above formula: To decouple P s , Q s , multiply the matrix by its inverse From this we get:

[0042]

[0043] Among them, η -1 is the decoupling matrix, P d and Q d represent the decoupled P and Q respectively; it is worth noting that P d and There is a linear relationship, while Q d It is linearly related to E, which is consistent with the characteristics of the droop law; droop control is achieved by utilizing P d and Q d To effectively implement, decoupled P and Q;

[0044] The grid power decoupling controller is composed of a decoupler and a traditional droop controller. The decoupler is based on a decoupling matrix η. -1 The input of the proposed grid power decoupling controller is the measured P and Q, i.e., the P generated by the P and Q calculation blocks. mea and Q mea , the output is the frequency and voltage reference generated by the inner loop voltage and current controller, from which the active-frequency and reactive-voltage control decoupler is obtained:

[0045] f * =f0-mD1(P0-P mea )-mD3(Q0-Q mea ) (15)

[0046] E * =E0-nD2(P0-P mea )-nD4(Q0-Q mea ) (16)

[0047] Decoupler parameters D1~D4 are calculated based on the decoupling matrix η -1 Sure:

[0048]

[0049] Among them, θ s Represents the system impedance angle.

[0050] Furthermore, in step 4, the power transfer function is linearized to construct a small signal model, and the proposed grid power decoupling controller and the model of the low-pass filter for power measurement are combined. The final representation of the small signal model is:

[0051]

[0052]

[0053] Secondly, the droop control with low-pass filter is modeled as a first-order inertia link and Replace with Then the small signal model is expressed as:

[0054]

[0055]

[0056] Among them, w f Represents the cutoff frequency of the low-pass filter. The characteristic equation derived from the above formula is a third-order expression:

[0057] a3s 3 +a2s 2 +a1s 1 +a0=0 (22)

[0058] in,

[0059]

[0060] Furthermore, the fuzzy neural network FNN in step 5 is designed as follows:

[0061] (1) The first layer is the input layer, which takes the input variable e i (i=1,...,n) is passed to the next layer, and the output of this layer can be expressed as:

[0062] x i =f i (e i )=e i (twenty four)

[0063] Among them, x i Represents the output of the i-th input node, and the input node adopts reactive deviation e1=e q and its derivatives That is, n = 2;

[0064] (2) The membership layer implements the fuzzification operation, and the activation function of each node is a Gaussian function to be defined:

[0065]

[0066] in Respectively represent the mean and standard deviation of the Gaussian function at the jth node of the ith input variable; m i is the number of membership functions in each cluster of the input variables, all parameters They are uniformly defined by the mean vector μ and the standard deviation vector σ as and in Indicates the number of all member functions;

[0067] (3) The third layer is the rule layer, which represents the fuzzy reasoning mechanism. The output of the node in this layer is the product of the input signal and its weight, expressed as:

[0068]

[0069] Among them l k (k=1,...,p r ) is the kth output of the regular layer, p r Indicates the total number of rules; represents the weight between the membership layer and the rule layer. Assuming the weight is uniform, all outputs of the rule layer can be collected as vector

[0070] (4) The output layer connects the rule layer with the corresponding weights, and the sum of their products is used as the output control law E of the FNN. f The vector form is expressed as:

[0071] y0=ωl=E f (27)

[0072] where ω=[ω1 ω2 ... ω pr ]∈R 1×p , is the vector of weights between the rules and the output layer.

[0073] Furthermore, the optimal active power-frequency control law update rule f of the grid power decoupling controller is obtained by repeated training and optimization of the fuzzy neural network FNN. * It is expressed as:

[0074] E c_s =f * +ε=ω * l * +ε (28)

[0075] Where ε represents the minimum reconstruction error vector, and the superscript “*” represents the optimal value of the corresponding parameter. The output control law is designed as:

[0076]

[0077] The superscript “^” indicates the estimated value of the corresponding parameter and is updated online.

[0078] Subtracting equations (28) and (29) yields the approximate error:

[0079]

[0080] Among them, the superscript “~” indicates the approximate error of the corresponding parameter; By using Taylor linearization technique, Transformed into a partially linear form:

[0081]

[0082] in, O h Represents the higher-order terms of a vector;

[0083] From the above formula, we can get:

[0084]

[0085] in The reactive-voltage update rules are basically the same.

[0086] Beneficial effects:

[0087] 1. The present invention adopts a control method combining grid forming control with fuzzy neural network, and proposes an innovative decoupled grid forming control scheme, which takes into account the inherent coupling characteristics of the entire distribution network. By changing the controller parameters and grid strength, it can effectively decouple the active power and reactive power of the weak distribution network, solve the impact caused by the high impedance ratio in the weak distribution network, and enable DERs to provide independent, accurate and autonomous frequency and voltage regulation, thereby improving the stability and dynamics of the power grid. At the same time, a fuzzy neural network FNN is introduced to learn and model the complex nonlinear relationships of the power system. Combining fuzzy logic and neural networks, the network can process fuzzy and uncertain information, and has adaptability and generalization capabilities. The controller dynamically adjusts the system parameters according to the real-time collected data and the output of the fuzzy neural network to achieve decoupled control and optimal distribution of power, providing a more robust and adaptive solution for the next generation of distribution networks with high penetration.

[0088] 2. The control strategy proposed in the present invention, FNN can be used to learn and model complex nonlinear relationships in the power system, such as the relationship between node voltage, line power flow and load information. Combined with grid forming control, the output power of the generator can be adjusted in real time according to the prediction results to maintain the voltage and frequency stability of the system.

[0089] 3. The present invention can optimize the output power of the dispatching generator to minimize the node voltage deviation and power loss, reduce the line loss and harmonic pollution in the power system, save energy and improve energy efficiency.

[0090] 4. The controller proposed in this invention is designed from a system-wide perspective, taking into account the coupling characteristics of the entire power grid. The proposed control strategy enables DERs to decouple active and reactive power and independently regulate voltage and frequency. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 It is the structure diagram of the power decoupling control system based on weak power grid FNN grid shaping control;

[0092] Figure 2 Decouple the control block diagram for the system;

[0093] Figure 3 It is a four-layer FNN network structure. DETAILED DESCRIPTION

[0094] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0095] The present invention discloses a decoupling control method for a weak power grid inverter system, and the specific steps are as follows:

[0096] Step 1: According to the basic principle of droop control of grid-connected inverter, a basic model of droop control active power and reactive power is established.

[0097] The P and Q transmitted from the inverter to the grid can be expressed by the voltage at the inverter terminal and the grid connection point (PCC) and the impedance between the two points. The basic model of active power and reactive power based on droop control is:

[0098]

[0099]

[0100] Where P and Q are the active power and reactive power output by DER respectively; E and V are the inverter terminal voltage and PCC point voltage respectively. It is expressed as the phase angle difference between the inverter terminal voltage and the PCC voltage. Z∠θ represents the output impedance of the DER system, including the impedance of the filter and the output line.

[0101] Step 2: Based on the power model obtained in step 1, the power model is simplified by considering the impedance and resistance-inductance characteristics in the transmission network. At the same time, the inverter droop control active-frequency (Pf) and reactive-voltage (QV) models are obtained.

[0102] In the high-voltage transmission network, due to the high impedance ratio of the transmission line, θ≈90°, and the phase angle difference Ignore, so there is The above formula can be simplified to:

[0103]

[0104]

[0105] When the PCC voltage and line impedance are fixed, it can be observed that P and The linear relationship between , and between Q and E. This linear relationship is conducive to the applicability of the droop characteristics in the control of grid-connected inverters within the transmission system. Specifically, these relationships can be expressed as:

[0106] f * =f0-m(P mea -P0) (5)

[0107] E * =E0-n(Q mea -Q0) (6)

[0108] Where f * 、E * is the reference value of the inverter terminal voltage frequency and amplitude; f0 and E0 are their rated values; P mea , Q mea are the output active power and reactive power; P0 and Q0 are the rated active power and reactive power; m and n are the active-frequency and reactive-voltage droop coefficients.

[0109] The inherent assumptions of droop control are limited to highly inductive transmission networks. These assumptions cannot hold in medium and low voltage distribution networks that are dominated by resistive and weak current. For distribution lines, due to their low X / R ratio, their power flow equations cannot be simplified, and as a result, the inverter outputs P and Q become coupled and associated with f and V. Weak distribution networks usually have very high grid impedance, so they cannot be regarded as ideal voltage sources. In addition, due to the coupling of P and Q, the resistive dominant characteristics of high grid impedance cause f and V control to be interdependent. Therefore, in the design of inverter control, weak current distribution networks and their related effects cannot be ignored. This makes it impossible for traditional droop-based control to independently adjust f and V by adjusting P and Q respectively.

[0110] Step 3: Based on the droop control model established in step 1 and step 2, a grid power decoupling controller is designed to decouple the active power and reactive power in the system according to the characteristics and requirements of the power system. Independent grid frequency and voltage control is achieved to achieve stable and optimized scheduling of the system. The design process of the grid power decoupling controller is as follows:

[0111] The proposed system decoupling grid formation control aims to decouple the entire resistance-dominated system, including weak distribution grids, DER filters, and export distribution lines. The coupling characteristics of the entire grid are continuously monitored. The designed system decoupling control diagram is shown in Figure 1 As shown, its equivalent circuit consists of two time-varying elements: time-varying impedance Z s ∠θ s and a time-varying voltage source These components reflect the online dynamics of the weak system under study and can be determined online based on real-time measurements during system operation. s ∠θ s Represents the total system impedance obtained by the inverter. It is the sum of the system side filter impedance, the output distribution line impedance, and the equivalent impedance of the weak distribution network. The first two parts are easy to determine based on the parameters of the DER inverter and the distribution line. The equivalent impedance of the weak distribution network can be identified online by monitoring the voltage and current changes of the PCC during events such as load changes:

[0112]

[0113] in, and respectively Represents the pre-operational measurements of the voltage and current phases, and Indicates the measured value after operation. By applying KVL, Derivation of equivalent grid voltage and Z g ∠θ g :

[0114]

[0115] According to the equivalent time-varying system circuit, the combined system impedance of the output impedance and the grid impedance is determined by the above formula:

[0116] Z s ∠θ s =Z l ∠θ l +Z g ∠θ g (9)

[0117] in,

[0118]

[0119]

[0120] Due to its resistive nature, the impedance Z s ∠θ s It cannot be simplified by the assumption that sinθ≈1, cosθ≈0. However, in the distribution network, the inverter terminal voltage Equivalent voltage to the grid The phase difference between Usually small, so The assumption is still valid. Therefore, according to formulas (1) and (2), the outputs P and Q of DER in the weak distribution network are expressed as:

[0121]

[0122]

[0123] Where P s , Q s are the active power and reactive power of DER, V g is the voltage amplitude of the equivalent voltage source of the power grid.

[0124] The system coupling characteristics can be expressed by matrix from the above formula: To decouple P s , Q s , multiply the matrix by its inverse From this we get:

[0125]

[0126] where η -1 is the decoupling matrix, P d and Q d They represent the decoupled P and Q respectively. It is worth noting that P d and There is a linear relationship, while Q dIt is linearly related to E, which is consistent with the characteristics of the droop law. Therefore, droop control can be achieved by utilizing P d and Q d to be efficiently implemented, they represent the decoupled P and Q. Therefore, this insight guides the design of system decoupled grid forming controllers, such as Figure 3 As shown in Figure 1, it consists of a decoupler and a traditional droop controller. Specifically, the decoupler is based on the decoupling matrix η -1 The proposed system is decoupled into a network controller whose input is the measured P and Q, i.e., the P generated by the P and Q calculation blocks. mea and Q mea The output of this controller is the frequency and voltage reference generated for the inner loop voltage and current controllers. The active-frequency, reactive-voltage control decoupler is obtained as:

[0127] f * =f0-mD1(P0-P mea )-mD3(Q0-Q mea ) (15)

[0128] E * =E0-nD2(P0-P mea )-nD4(Q0-Q mea ) (16)

[0129] The decoupler parameters (D1~D4) are calculated according to the decoupling matrix η -1 Sure:

[0130]

[0131] It means that the decoupling parameters mainly depend on the system impedance angle θ s .

[0132] Step 4: Based on the control scheme proposed in step 3, in order to enable DERs to dynamically adapt to the time-varying conditions of the power grid and ensure continuous stable and efficient operation, an enhanced control mechanism is proposed and the small signal model of the proposed controller is derived to ensure the stable operation of the system.

[0133] In order to enable DERs to autonomously decouple P and Q of the entire power grid and achieve independent f and V control, DERs can dynamically adapt to the time-varying conditions of the power grid and ensure continuous stable and efficient operation. The enhanced control mechanism helps to achieve independent and accurate frequency and voltage regulation, and plays a key role in strengthening the reliability and resilience of modern distribution systems. The small signal model of the proposed controller is derived. The model is used for small signal stability analysis and determination of control parameters.

[0134] The small signal model is constructed by linearizing the power transfer function. In addition, it also combines the model of the proposed system decoupled grid forming controller and the low-pass filter for power measurement. The final representation of the small signal model is:

[0135]

[0136]

[0137] Secondly, the droop control with low-pass filter is modeled as a first-order inertia link and Replace with Then the small signal model of the system under study can be expressed as:

[0138]

[0139]

[0140] where w f Represents the cutoff frequency of the low-pass filter. From the above formula, the characteristic equation can be derived as a third-order expression:

[0141] a3s 3 +a2s 2 +a1s 1 +a0=0 (22)

[0142] in,

[0143]

[0144] Step 5: Based on the power decoupling controller proposed in step 4, design the fuzzy neural network (FNN) control rules. Enable it to accurately predict the operating status and trend of the power system. Improve the accuracy and robustness of the model through repeated training and optimization.

[0145] (1) The first layer is the input layer, which takes the input variable e i (i=1,...,n) is passed to the next layer. The output of this layer can be expressed as:

[0146] x i =f i (e i )=e i (twenty four)

[0147] where x i Represents the output of the ith input node. The input node adopts reactive power deviation e1=e q and its derivatives That is, n=2.

[0148] (2) The membership layer implements the fuzzification operation, and the activation function of each node is a Gaussian function to be defined:

[0149]

[0150] in Respectively represent the mean and standard deviation of the Gaussian function at the jth node of the ith input variable; m i is the number of membership functions in each cluster of the input variables. Therefore, all parameters They are uniformly defined by the mean vector (μ) and the standard deviation vector (σ) as and in Indicates the number of all member functions.

[0151] (3) The third layer is the rule layer, which represents the fuzzy reasoning mechanism. The output of the node in this layer is the product of the input signal and its weight, which can be expressed as:

[0152]

[0153] Among them l k (k=1,...,p r ) is the kth output of the regular layer, p r Indicates the total number of rules; represents the weight between the membership layer and the rule layer, assuming that the weight is uniform. All outputs of the rule layer can be collected as the vector

[0154] (4) The output layer connects the rule layer with the corresponding weights, and the sum of their products is used as the output control law of the FNN (E f ) can be expressed in vector form as:

[0155] y0=ωl=E f (27)

[0156] where ω=[ω1 ω2 … ω pr ]∈R 1×p , is the vector of weights between the rules and the output layer.

[0157] By adjusting the network parameters online, the FNN framework can learn the control law of the decoupling controller, reducing the requirements for detailed model information of the system. In addition, the online tuning algorithm of FNN is derived through the Lyapunov stability theorem and the projection algorithm to ensure the convergence and stable control performance of the network. Due to the powerful approximation ability of the FNN framework, the optimal control law frequency update rule f of the decoupling strategy * It can be expressed as:

[0158] E c_s =f* +ε=ω * l * +ε (28)

[0159] Where ε represents the minimum reconstruction error vector. The superscript “*” represents the optimal value of the corresponding parameter. Therefore, the output control law is designed as:

[0160]

[0161] The superscript “^” indicates the estimated value of the corresponding parameter and is updated online.

[0162] Subtract the above formula to get the approximate error:

[0163]

[0164] The superscript “~” indicates the approximate error of the corresponding parameter; By using the Taylor linearization technique, Transformed into a partially linear form:

[0165]

[0166] in O h Represents higher-order terms of a vector.

[0167] From the above formula, we can get:

[0168]

[0169] in The voltage update rules are basically the same and will not be further described here.

[0170] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A decoupling control method for a weak grid inverter system, characterized in that: The steps include: Step 1: According to the basic principle of droop control of grid-connected inverter, a basic model of droop control of active power and reactive power is established; The P and Q transmitted from the inverter to the grid are represented by the voltage at the inverter terminal and the grid connection point PCC and the impedance between the two points. The basic model of active power and reactive power based on droop control is: Among them, P and Q are the active power and reactive power output by distributed energy DER respectively; E and V are the inverter terminal voltage and PCC point voltage respectively. It is expressed as the phase angle difference between the inverter terminal voltage and the PCC voltage, and Z∠θ represents the output impedance of the distributed energy DER system, including the impedance of the filter and the output line; Step 2: Based on the power model obtained in step 1, the impedance and resistance-inductance characteristics in the transmission network are considered to simplify the power model, and the inverter droop control active-frequency Pf model and reactive-voltage QV model are obtained at the same time; Step 3: Design a grid power decoupling controller to decouple the active power and reactive power in the system according to the characteristics and requirements of the power system, and realize independent grid frequency and voltage control to achieve stable and optimized scheduling of the system; The weak grid inverter system includes a weak distribution network, a DER filter and an export distribution line. The equivalent circuit of the system consists of two time-varying elements: the time-varying impedance Z s ∠θ s and a time-varying voltage source Z s ∠θ s It represents the total system impedance obtained by the inverter, which is the sum of the system-side filter impedance, the output distribution line impedance, and the equivalent impedance of the weak distribution network. The system-side filter impedance and the output distribution line impedance are determined according to the parameters of the DER inverter and the distribution line. The equivalent impedance of the weak distribution network is identified online by monitoring the voltage and current changes of the PCC during load change events: in, and respectively Represents the pre-operational measurements of the voltage and current phases, and Indicates the measured value after operation, by applying KVL, Derivation of equivalent grid voltage and Z g ∠θ g : According to the equivalent time-varying system circuit, the combined system impedance of the output impedance and the grid impedance is determined by the above formula: WITH s ∠θ s =Z l ∠θ l +Z g ∠θ g (9) in, Due to its resistive nature, the impedance Z s ∠θ s The assumption that sinθ≈1, cosθ≈0 cannot be used to simplify. In the distribution network, the inverter terminal voltage Equivalent voltage to the grid The phase difference between Very small, so The assumption is valid, therefore, the outputs P and Q of DER in the weak distribution network are expressed as: Where P s , Q s are the active power and reactive power of DER, V g is the voltage amplitude of the equivalent voltage source of the power grid; The system coupling characteristics are expressed by matrix using the above formula: To decouple P s , Q s , multiply the matrix by its inverse From this we get: Among them, η -1 is the decoupling matrix, P d and Q d represent the decoupled P and Q respectively; it is worth noting that P d and There is a linear relationship, while Q d It is linearly related to E, which is consistent with the characteristics of the droop law; droop control is achieved by utilizing P d and Q d To effectively implement, decoupled P and Q; The grid power decoupling controller is composed of a decoupler and a traditional droop controller. The decoupler is based on a decoupling matrix η. -1 The input of the proposed grid power decoupling controller is the measured P and Q, i.e., the P generated by the P and Q calculation blocks. mea and Q mea , the output is the frequency and voltage reference generated by the inner loop voltage and current controller, from which the active-frequency and reactive-voltage control decoupler is obtained: f * =f0-mD1(P0-P mea )-mD3(Q0-Q mea ) (15) E * =E0-nD2(P0-P mea )-nD4(Q0-Q mea ) (16) Decoupler parameters D1~D4 are calculated based on the decoupling matrix η -1 Sure: Among them, θ s represents the system impedance angle; Step 4: Based on the grid power decoupling controller proposed in step 3, an enhanced control mechanism is proposed and a small signal model of the controller is derived; By linearizing the power transfer function and constructing a small signal model, the proposed grid power decoupling controller and the model of the low-pass filter for power measurement are also combined. The final representation of the small signal model is: Secondly, the droop control with low-pass filter is modeled as a first-order inertia link and Replace with Then the small signal model is expressed as: Among them, w f Represents the cutoff frequency of the low-pass filter. The characteristic equation derived from the above formula is a third-order expression: a3s 3 +a2s 2 +a1s 1 +a0=0 (22) in, Step 5: Based on the power decoupling controller proposed in step 4, design the fuzzy neural network FNN control rule, use historical data to train the fuzzy neural network, and improve the accuracy and robustness of the model through repeated training and optimization; Step 6: Combine the fuzzy neural network FNN with the grid power decoupling controller as the core part of the control system. Use the fuzzy neural network to learn and predict the state of the power system, and adjust the parameters in the power decoupling controller according to the prediction results to achieve intelligent control of the system.

2. A decoupling control method for a weak grid inverter system according to claim 1, characterized in that: In step 2, in the high-voltage transmission network, the transmission line has a high impedance ratio, θ≈90°, and the phase angle difference Ignore, so sinθ≈1,cosθ≈0, The above formula is simplified to: When the PCC voltage and line impedance are fixed, it is observed that P and The linear relationship between them and between Q and E is expressed as: f * =f0-m(P mea -P0) (5) E * =E0-n(Q mea -Q0) (6) In the formula, f * 、E * is the reference value of the inverter terminal voltage frequency and amplitude; f0 and E0 are their rated values; P mea , Q mea are the output active power and reactive power; P0 and Q0 are the rated active power and reactive power; m and n are the active-frequency and reactive-voltage droop coefficients.

3. A decoupling control method for a weak grid inverter system according to claim 1, characterized in that: The fuzzy neural network FNN in step 5 is designed as follows: (1) The first layer is the input layer, which takes the input variable e i (i=1,…,n) is passed to the next layer, and the output of this layer can be expressed as: x i =f i (e i )=e i (24) Among them, x i Represents the output of the i-th input node, and the input node adopts reactive deviation e1=e q and its derivatives That is, n = 2; (2) The membership layer implements the fuzzification operation, and the activation function of each node is a Gaussian function to be defined: in Respectively represent the mean and standard deviation of the Gaussian function at the jth node of the ith input variable; m i is the number of membership functions in each cluster of the input variables, all parameters They are uniformly defined by the mean vector μ and the standard deviation vector σ as and in Indicates the number of all member functions; (3) The third layer is the rule layer, which represents the fuzzy reasoning mechanism. The output of the node in this layer is the product of the input signal and its weight, expressed as: Among them l k (k=1,…,p r ) is the kth output of the regular layer, p r Indicates the total number of rules; represents the weight between the membership layer and the rule layer. Assuming the weight is uniform, all outputs of the rule layer can be collected as vector (4) The output layer connects the rule layer with the corresponding weights, and the sum of their products is used as the output control law E of the FNN. f The vector form is expressed as: y0=ωl=E f (27) where ω=[ω1ω2…ω pr ]∈R 1×p , is the vector of weights between the rules and the output layer.

4. A decoupling control method for a weak grid inverter system according to claim 3, characterized in that: The optimal active power-frequency control law update rule f of the grid power decoupling controller is obtained by repeated training and optimization of the fuzzy neural network FNN. * It is expressed as: E c_s =f * +e=oh * l * +e (28) Where ε represents the minimum reconstruction error vector, and the superscript "*" represents the optimal value of the corresponding parameter; the output control law is designed as: The superscript "^" indicates the estimated value of the corresponding parameter and is updated online. Subtracting equations (28) and (29) yields the approximate error: Among them, the superscript "~" indicates the approximate error of the corresponding parameter; By using Taylor linearization technique, Transformed into a partially linear form: in, O h Represents the higher-order terms of a vector; From the above formula, we can get: in The reactive-voltage update rules are the same.

Images