Converter control method and system based on characteristic damping and fuzzy neural network

Abstract

This invention discloses a converter control method and system based on characteristic damping and fuzzy neural networks. The method includes: performing eigenvalue analysis on a grid-connected inverter system and training a pre-set fuzzy neural network to obtain a pre-trained fuzzy neural network; sensing real-time operating parameters to obtain a real-time short-circuit ratio estimate and a current per-unit output active power value; inputting the real-time short-circuit ratio estimate and the current per-unit output active power value into the pre-trained fuzzy neural network for online fuzzy neural network inference; and performing nonlinear hysteresis mode switching based on the inference results to achieve operation control of the grid-connected inverter system. This invention can improve the adaptability of the converter across the entire operating range from extremely weak to strong power grids through smooth switching between grid-following and grid-connected modes. As a converter control method and system based on characteristic damping and fuzzy neural networks, this invention can be widely applied in the field of grid-connected inverter system control technology.

Description

Technical Field

[0001] This invention relates to the field of grid-connected inverter system control technology, and in particular to a converter control method and system based on characteristic damping and fuzzy neural networks. Background Technology

[0002] As a high proportion of new energy sources (such as wind power and photovoltaics) are connected to the grid through power electronic converters, the power system is gradually exhibiting typical weak grid characteristics of low inertia and weak damping. The continuous reduction of the grid short-circuit ratio (SCR) poses a severe challenge to the grid connection stability of the converters.

[0003] Currently, converter control strategies are mainly divided into two categories: grid-following (GFL) and grid-forming (GFM). When the grid short-circuit ratio is large (strong grid), grid-following converters exhibit good grid-connected current tracking performance. However, under extremely weak grid conditions (e.g., SCR < 1.5) and high power output, due to the dynamic coupling between the phase-locked loop (PLL) and the weak grid impedance, the system damping ratio of the grid-following converter will decrease sharply or even become negative (exhibiting negative damping collapse), causing severe oscillations in the subsynchronous frequency band or even grid disconnection and failure. Conversely, grid-forming converters have excellent voltage support capabilities under weak grid conditions, but under strong grid conditions, they are prone to low-frequency power fluctuations on the electromechanical time scale. Therefore, real-time switching between grid-following and grid-forming modes based on the grid strength is crucial to ensuring stable operation of the converter over a wide grid impedance range.

[0004] For dual-mode switching of grid-connected inverters, related technologies include dual-mode control methods for LC-type grid-connected inverters based on grid impedance adaptation. These methods propose switching between current source and voltage source modes based on changes in grid impedance, combined with set impedance boundaries and hysteresis. However, they only consider the impact of impedance and short-circuit ratio changes on system stability, without taking into account the impact of renewable energy power factors on stability.

[0005] Another approach proposes a definition and physical meaning of the operating short-circuit ratio, rigorously quantifying the intensity under actual operating conditions using the generalized operating short-circuit ratio and the critical operating short-circuit ratio of equipment. Furthermore, it reveals the specific mechanisms by which actual operating conditions affect intensity under different stability states. However, this approach considers the stability of the new energy system under a unity power factor, without considering the system under non-unity power factor conditions.

[0006] Furthermore, some related technologies propose the concept of Operating Short-Circuit Ratio (OSCR, i.e., the ratio of short-circuit ratio to active power) to maintain the stability of power station oscillations in the low-frequency and subsynchronous frequency bands. Based on this ratio, a single hard threshold is set, serving as the direct basis for switching between grid-connected and grid-connected converter operating modes. However, this method has limitations: forcibly merging the short-circuit ratio and active power into a one-dimensional linear ratio (OSCR) can easily lead to numerical explosions due to small disturbances in the denominator under light load or power fluctuation conditions, causing the converter to experience high-frequency flutter and malfunctions at the critical hard threshold. Summary of the Invention

[0007] To address the aforementioned technical problems, the present invention aims to provide a converter control method and system based on characteristic damping and fuzzy neural networks, which can improve the adaptability of the converter across the entire operating range from extremely weak to strong power grids through smooth switching between grid-following and grid-connecting modes.

[0008] The first technical solution adopted in this invention is: a converter control method based on characteristic damping and fuzzy neural network, comprising the following steps: Eigenvalue analysis is performed on the grid-connected inverter system to construct a three-dimensional characteristic damping dataset, and a pre-trained fuzzy neural network is trained to obtain the pre-trained fuzzy neural network. Initialize the grid-connected inverter system and perform real-time operating parameter sensing on the initialized grid-connected inverter system to obtain the real-time short-circuit ratio estimate and the current per-unit value of the output active power of the grid-connected inverter system. The real-time short-circuit ratio estimate of the grid-connected inverter system and the per-unit value of the current output active power are input into the pre-trained fuzzy neural network for online fuzzy neural network inference, and nonlinear hysteresis mode switching is performed based on the inference results to realize the operation control of the grid-connected inverter system.

[0009] Furthermore, the step of constructing a three-dimensional feature damping dataset by performing eigenvalue analysis on the grid-connected inverter system and training a pre-trained fuzzy neural network to obtain the pre-trained fuzzy neural network specifically includes: Construct small-signal state-space models of grid-connected inverter systems in both grid-following and grid-connected modes; The eigenvalues ​​of the state matrix in the state-space small-signal model are solved to obtain the eigenvalue distribution of the grid-connected inverter system under different short-circuit ratios and active power conditions. Based on the eigenvalue distribution, damping ratio data of grid-connected inverter systems under different short-circuit ratios and active power conditions are extracted to construct a three-dimensional characteristic damping dataset. Based on a three-dimensional feature damping dataset, a pre-set fuzzy neural network is trained to obtain a pre-trained fuzzy neural network.

[0010] Furthermore, the step of constructing the state-space small-signal model of the grid-connected inverter system in both grid-following and grid-connected modes specifically includes: The nonlinear dynamic equations of each structure of the grid converter are determined, and small-signal perturbation linearization is performed at the steady-state operating point to construct a mathematical model of the grid converter. The nonlinear dynamic equations of each structure of the grid converter are determined, and small-signal perturbation linearization is performed at the steady-state operating point to construct a mathematical model of the grid converter. By combining the mathematical models of grid-connected inverters and grid-connected inverters, the small-signal state-space models of the grid-connected inverter system in both grid-connected and grid-connected modes are obtained.

[0011] Furthermore, the step of determining the nonlinear dynamic equations of each structure of the grid converter, performing small-signal perturbation linearization at the steady-state operating point, and constructing the mathematical model of the grid converter specifically includes: Based on the main circuit topology of the grid converter, and combining Kirchhoff's voltage law and Kirchhoff's current law, the nonlinear dynamic equations of the main circuit of the grid converter are constructed. Based on the current inner loop control structure of the grid-connected converter, and combined with the PI and feedforward decoupling principle, the nonlinear dynamic equation of the current inner loop control structure of the grid-connected converter is constructed. Based on the PQ outer loop control structure of the grid converter, and combined with instantaneous power theory, the nonlinear dynamic equation of the PQ outer loop control structure of the grid converter is constructed. Based on the phase-locked loop structure of the grid converter and combined with the principle of PI-type phase-locked loop, the nonlinear dynamic equation of the phase-locked loop structure of the grid converter is constructed. Based on the coordinate transformation equation between the controller coordinate system and the global coordinate system, the nonlinear dynamic equations of the main circuit of the grid-connected converter, the nonlinear dynamic equations of the inner current loop control structure of the grid-connected converter, the nonlinear dynamic equations of the PQ outer loop control structure of the grid-connected converter, and the nonlinear dynamic equations of the phase-locked loop structure of the grid-connected converter are linearized at the operating point, and the corresponding connection relationships are determined to construct the mathematical model of the grid-connected converter.

[0012] Furthermore, the step of determining the nonlinear dynamic equations of each structure of the grid converter and performing small-signal perturbation linearization at the steady-state operating point to construct the mathematical model of the grid converter specifically includes: Based on the main circuit topology of the grid converter, and combining Kirchhoff's voltage law and Kirchhoff's current law, the nonlinear dynamic equations of the main circuit of the grid converter are constructed. Based on the current inner loop control structure of the grid converter, and combined with the PI and feedforward decoupling principle, the nonlinear dynamic equation of the current inner loop control structure of the grid converter is constructed. Based on the voltage and reactive power outer loop control structure of the grid converter, and combined with the cascaded PI control principle, the nonlinear dynamic equation of the voltage and reactive power outer loop control structure of the grid converter is constructed. Based on the power synchronization loop structure of the grid converter, and combined with the rotor motion equation of the synchronous generator, a nonlinear dynamic equation for the power synchronization loop structure of the grid converter is constructed. Based on the coordinate transformation equations between the controller coordinate system and the global coordinate system, the nonlinear dynamic equations of the main circuit of the grid converter, the nonlinear dynamic equations of the inner current loop control structure of the grid converter, the nonlinear dynamic equations of the outer voltage and reactive power loop control structure of the grid converter, and the nonlinear dynamic equations of the power synchronization loop structure of the grid converter are linearized at the operating point, and the corresponding connection relationships are determined to construct a mathematical model of the grid converter.

[0013] Furthermore, the step of initializing the grid-connected inverter system and sensing its real-time operating parameters to obtain the estimated real-time short-circuit ratio and the current per-unit value of the output active power specifically includes: The initial operating mode of the grid-connected inverter system is set to grid-following mode, and an initial state value is assigned to the unit delay element in the nonlinear hysteresis logic. Collect the grid connection point voltage and current signals of the grid-connected inverter system during operation; Based on the grid connection point voltage and current signals, the grid impedance identification algorithm is used to estimate the real-time short-circuit ratio of the grid-connected inverter system and simultaneously obtain the per-unit value of the current output active power of the grid-connected inverter system.

[0014] Furthermore, the step of estimating the real-time short-circuit ratio of the grid-connected inverter system based on the grid connection point voltage and current signals using a grid impedance identification algorithm, and simultaneously obtaining the current per-unit value of the grid-connected inverter system's output active power, specifically includes: Based on the grid-connected inverter system, all key parameters to be identified are pre-set. The key parameters to be identified include the injected non-characteristic harmonic frequency, complex filter parameters, notch filter parameters, and fundamental angular frequency. Based on the grid-connected inverter system after setting and identifying key parameters, a sinusoidal harmonic current of preset amplitude is injected, and the voltage and current signals at the grid connection point are collected simultaneously. Target harmonic components are extracted from the voltage and current signals at the grid connection point to obtain the voltage harmonic components and current harmonic components in the stationary coordinate system. The equivalent impedance parameters of the power grid are calculated based on the voltage harmonic components and the current harmonic components, and the power grid impedance amplitude is obtained. Based on the grid impedance amplitude, determine the estimated real-time short-circuit ratio of the grid-connected inverter system, and simultaneously obtain the per-unit value of the current output active power of the grid-connected inverter system.

[0015] Furthermore, the step of inputting the real-time short-circuit ratio estimate of the grid-connected inverter system and the current per-unit value of the output active power into the pre-trained fuzzy neural network for online fuzzy neural network inference, and performing nonlinear hysteresis mode switching based on the inference results to realize the operation control of the grid-connected inverter system, specifically includes: The real-time short-circuit ratio estimate of the grid-connected inverter system and the current per-unit value of the output active power are input into the pre-trained fuzzy neural network for fuzzification, fuzzy rule reasoning and defuzzification calculation in sequence to obtain the continuous control confidence signal of the grid-connected inverter system under the current operating conditions. Based on the continuous control confidence signal of the grid-connected inverter system under the current operating conditions, nonlinear hysteresis mode switching is performed to realize the operation control of the grid-connected inverter system.

[0016] Furthermore, the step of performing nonlinear hysteresis mode switching based on the continuous control confidence signal of the grid-connected inverter system under the current operating conditions to realize the operation control of the grid-connected inverter system specifically includes: The continuous control confidence signal of the grid-connected inverter system under the current operating condition is input into a nonlinear hysteresis logic with Schmitt triggering characteristics for comparison. If the continuous control confidence signal of the grid-connected inverter system under the current operating conditions is less than the preset grid-connection trigger lower threshold, the operating mode of the grid-connected inverter system will be switched to grid-connection mode. If the continuous control confidence signal of the grid-connected inverter system under the current operating conditions is greater than the preset grid-following trigger upper limit threshold, the operating mode of the grid-connected inverter system will be switched to grid-following mode. If the continuous control confidence signal of the grid-connected inverter system under the current operating condition is between the preset lower threshold for grid connection triggering and the preset upper threshold for grid connection triggering, then the hold logic is triggered to maintain the operating mode of the grid-connected inverter system in the previous control cycle, thereby realizing the operation control of the grid-connected inverter system.

[0017] The second technical solution adopted in this invention is: a converter control system based on characteristic damping and fuzzy neural network, comprising: The first module is used to perform eigenvalue analysis on the grid-connected inverter system to construct a three-dimensional feature damping dataset, and to train a preset fuzzy neural network to obtain a pre-trained fuzzy neural network. The second module is used to initialize the grid-connected inverter system and to sense the real-time operating parameters of the initialized grid-connected inverter system, thereby obtaining the estimated real-time short-circuit ratio and the current per-unit value of the output active power of the grid-connected inverter system. The third module is used to input the real-time short-circuit ratio estimate of the grid-connected inverter system and the current per-unit value of the output active power into the pre-trained fuzzy neural network for online fuzzy neural network inference, and to perform nonlinear hysteresis mode switching based on the inference results to realize the operation control of the grid-connected inverter system.

[0018] The beneficial effects of the method and system of this invention are as follows: This invention constructs a three-dimensional characteristic damping dataset by performing eigenvalue analysis on the grid-connected inverter system and trains a pre-trained fuzzy neural network to obtain a pre-trained fuzzy neural network; further, it initializes the grid-connected inverter system and performs real-time operating parameter sensing on the initialized grid-connected inverter system to obtain the real-time short-circuit ratio estimate and the current per-unit value of the output active power of the grid-connected inverter system; by establishing a fuzzy neural network based on the evolution law of system eigenvalues, the grid short-circuit ratio and output power are used as independent dimensions for two-dimensional deduction to accurately identify the collapse boundary of the system characteristic damping ratio; finally, the real-time short-circuit ratio estimate and the current per-unit value of the output active power of the grid-connected inverter system are input into the pre-trained fuzzy neural network for online fuzzy neural network inference, and nonlinear hysteresis mode switching is performed according to the inference results to realize the operation control of the grid-connected inverter system. To ensure stability during switching, the mode command is issued using a nonlinear hysteresis switching method based on confidence signals, which effectively absorbs measurement noise and power fluctuations and avoids high-frequency malfunctions at critical boundaries. It can improve the adaptability of converters across the entire operating range from extremely weak grids to strong grids by smoothly switching between grid-following and grid-building modes. Attached Figure Description

[0019] Figure 1 This is a flowchart of the steps of the converter control method based on characteristic damping and fuzzy neural network of the present invention; Figure 2 This is a structural block diagram of the converter control system based on characteristic damping and fuzzy neural network of the present invention; Figure 3 This is a schematic diagram of the topology of a grid-connected inverter connected to a weak grid according to a specific embodiment of the present invention; Figure 4 This is a schematic diagram of the VSC small signal interface provided in a specific embodiment of the present invention; Figure 5 This is a schematic diagram of the inner current loop provided in a specific embodiment of the present invention; Figure 6 This is a schematic diagram of the GFL outer ring d-axis control block provided in a specific embodiment of the present invention; Figure 7 This is a schematic diagram of the GFL outer ring q-axis control block provided in a specific embodiment of the present invention; Figure 8 This is a schematic diagram of the GFM outer ring d-axis control block provided in a specific embodiment of the present invention; Figure 9 This is a schematic diagram of the GFM outer ring q-axis control block provided in a specific embodiment of the present invention; Figure 10 This is a schematic diagram of the GFL power synchronization loop control block provided in a specific embodiment of the present invention; Figure 11 This is a schematic diagram of the GFM power synchronization loop control block provided in a specific embodiment of the present invention; Figure 12 This is a schematic diagram of GFL / GFM hysteresis switching provided in a specific embodiment of the present invention; Figure 13 This is a schematic diagram of the converter switching process provided in a specific embodiment of the present invention; Figure 14 This is a schematic diagram of a power grid impedance identification method provided in a specific embodiment of the present invention; Figure 15 This is a schematic diagram of the GFL damping characteristics provided in a specific embodiment of the present invention; Figure 16 This is a schematic diagram of the GFM damping characteristics provided in a specific embodiment of the present invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. The step numbers in the following embodiments are only for ease of explanation and do not limit the order of the steps. The execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.

[0021] First, it should be noted that this invention addresses the issues of nonlinear damping assessment distortion and switching chatter in the operating short-circuit ratio (OSCR) discrimination method, which addresses the limitations of single grid-following or grid-connecting modes when converters operate under different grid impedance and power conditions. It proposes a dual-mode control method for converters based on characteristic damping surfaces and fuzzy neural networks. By establishing a fuzzy neural network based on the evolution of system eigenvalues, the per-unit values ​​of the grid short-circuit ratio and output active power are used as independent dimensions for two-dimensional deduction to accurately identify the collapse boundary of the system's characteristic damping ratio. To ensure stability during switching, the mode command issuance adopts a nonlinear hysteresis switching method based on confidence signals, effectively absorbing measurement noise and power fluctuations, and avoiding high-frequency malfunctions at critical boundaries. This invention improves the adaptability of converters across the entire operating range from extremely weak to strong grids through smooth switching between grid-following and grid-connecting modes.

[0022] Reference Figure 1 This invention provides a converter control method based on characteristic damping and fuzzy neural networks, which includes the following steps: S100. Perform eigenvalue analysis on the grid-connected inverter system to construct a three-dimensional feature damping dataset, and train the preset fuzzy neural network to obtain the pre-trained fuzzy neural network. First, it should be noted that the topology of the LC-type grid-connected converter involved in the embodiments of the present invention is as follows: Figure 3 As shown, the circuit includes a single-phase full-bridge inverter circuit and an LC filter. The two ends of the DC-side filter capacitor Cdc are connected to the two input terminals of the single-phase full-bridge inverter circuit, respectively. The output terminal of the single-phase full-bridge inverter circuit is connected to the input terminal of the LC filter, and the output terminal of the LC filter is connected to a single-phase power grid with grid impedance Zg through a common coupling point PCC.

[0023] In this embodiment, state-space small-signal models of the grid-connected inverter system are established in both grid-following (GFL) and grid-fitting (GFM) modes. The eigenvalue distributions of the system under different short-circuit ratios (SCR) and per-unit active power (P) conditions are obtained through eigenvalue solving. Damping ratio data corresponding to each operating point are extracted to construct a three-dimensional characteristic damping dataset. Based on this dataset, a fuzzy neural network is established and trained to establish a path from two-dimensional input variables (SCR, P) to control confidence output. The nonlinear mapping relationship.

[0024] S110. Construct small-signal state-space models of grid-connected inverter systems in grid-following and grid-connected modes; S111. Determine the nonlinear dynamic equations of each structure of the grid converter, and perform small-signal perturbation linearization at the steady-state operating point to construct the mathematical model of the grid converter. Specifically, based on the main circuit topology of the grid-connected converter, and combining Kirchhoff's voltage law and current law, a nonlinear dynamic equation for the main circuit of the grid-connected converter is constructed; based on the current inner loop control structure of the grid-connected converter, and combining the PI and feedforward decoupling principle, a nonlinear dynamic equation for the current inner loop control structure of the grid-connected converter is constructed; based on the PQ outer loop control structure of the grid-connected converter, and combining instantaneous power theory, a nonlinear dynamic equation for the PQ outer loop control structure of the grid-connected converter is constructed; based on the phase-locked loop structure of the grid-connected converter... Based on the principle of PI-type phase-locked loop, a nonlinear dynamic equation for the phase-locked loop structure of the grid-connected converter is constructed. Based on the coordinate transformation equation between the controller coordinate system and the global coordinate system, the nonlinear dynamic equations of the main circuit of the grid-connected converter, the nonlinear dynamic equations of the inner current loop control structure of the grid-connected converter, the nonlinear dynamic equations of the PQ outer loop control structure of the grid-connected converter, and the nonlinear dynamic equations of the phase-locked loop structure of the grid-connected converter are linearized at the operating point, and the corresponding connection relationships are determined to construct a mathematical model of the grid-connected converter.

[0025] First, it should be noted that the topology diagram of the grid-connected converter connected to the weak current grid in this embodiment is as follows: Figure 3 As shown, the small-signal interface diagram of the converter is as follows: Figure 4 As shown, the current inner loop control block diagram is as follows: Figure 5 As shown, the GFLdq axis control block diagram is as follows: Figure 6 , Figure 7 As shown, the GFMdq axis control block diagram is as follows: Figure 8 , Figure 9 As shown, the control block diagram of the GFL power synchronization loop is as follows: Figure 10 As shown, the control block diagram of the GFM power synchronization loop is as follows: Figure 11 As shown.

[0026] The mathematical model of the grid converter is shown below: According to Kirchhoff's Voltage Law (KVL) and Current Law (KCL), the dynamic equation of an AC circuit can be obtained as follows: ; In the above formula, For transformer inductance, Equivalent conductance of the power grid The equivalent impedance of the power grid. ω is the fundamental angular frequency. , , , , and It is both a state variable and an output variable in the circuit. , , and For input variables.

[0027] The dynamic equation of the inner loop can be expressed as: ; In the above formula, , , , For the inner loop state variables. , For output variables. , , , , , , For input variables. To control the state variable output of the i-th PI circuit, and To control the state variable output of the first PI loop (current inner loop). To control the state variable output of the 8th PI loop (current inner loop). , The first-order inertial element of the inner current loop dq axis is output as the state variable of the inertial element. This is the proportional gain for the first PI stage. The integral coefficient of the first PI stage; This is the proportional coefficient for the 8th PI stage. This is the integral coefficient of the 8th PI stage.

[0028] The dynamic equations of the outer loop of the ground truth network (GFL) can be expressed as follows: ; In the above formula, , For state variables, , For output variables, , , , For input variables. To control the state variable output of the third PI stage (GFL outer loop d-axis). To control the state variable output of the fourth PI stage (GFL outer loop q-axis). Rated output active power, Rated output reactive power, The dynamic equation of a phase-locked loop (PLL) can be expressed as follows: ; In the above formula and These are the state variables of the phase-locked loop. For input variables, , For output variables.

[0029] The coordinate transformation equations between the controller coordinate system and the global coordinate system are as follows: ; superscript in the above formula Indicates the value of the variable in the controller coordinate system, superscript This indicates the value of the variable in the common coordinate system of the power grid. The angle by which the controller coordinate system leads the global coordinate system. For input variables, before coordinate transformation and As input variables, and also after coordinate transformation and For output variables.

[0030] The above dynamic equations can be linearized at the operating point to obtain their respective small-signal models. Based on the input and output variables of each small-signal model, the connection relationships between them can be determined, thus obtaining the overall small-signal model of the grid-fed converter (GFL). The mathematical equations of the GFL converter small-signal model are as follows: ; in For state variables, The state matrix, For the input matrix, The input vector. For the output matrix, To directly transmit the matrix, This is the output vector.

[0031] S112. Determine the nonlinear dynamic equations of each structure of the grid converter, and perform small-signal perturbation linearization at the steady-state operating point to construct the mathematical model of the grid converter. Specifically, based on the main circuit topology of the grid-connected converter, and combining Kirchhoff's voltage law and current law, a nonlinear dynamic equation for the main circuit of the grid-connected converter is constructed; based on the current inner loop control structure of the grid-connected converter, and combining the PI and feedforward decoupling principle, a nonlinear dynamic equation for the current inner loop control structure of the grid-connected converter is constructed; based on the voltage and reactive power outer loop control structure of the grid-connected converter, and combining the cascaded PI control principle, a nonlinear dynamic equation for the voltage and reactive power outer loop control structure of the grid-connected converter is constructed; and based on the power synchronization loop structure of the grid-connected converter... Based on the synchronous generator rotor motion equation, a nonlinear dynamic equation for the power synchronization loop structure of the grid-type converter is constructed. Then, based on the coordinate transformation equation between the controller coordinate system and the global coordinate system, the nonlinear dynamic equations of the main circuit of the grid-type converter, the nonlinear dynamic equations of the inner current loop control structure, the nonlinear dynamic equations of the outer voltage and reactive power loop control structure, and the nonlinear dynamic equations of the power synchronization loop structure are linearized at the operating point, and the corresponding connection relationships are determined, thus constructing a mathematical model of the grid-type converter.

[0032] In this embodiment, the mathematical model of the grid-type converter is established as follows: The dynamic equations for the inner loop of the network can be expressed as follows: ; In the above formula, , , , For the state variables of the inner loop, , For output variables. , , , , , , For input variables; ; The state variables of the outer loop of GFM are , and The output variable is and The input variables are , , and .

[0033] The dynamic equations of the power synchronization loop can be expressed as follows: ; In the above formula, the state variable is and The input variables are , , and The output variable is and .

[0034] Similarly, the mathematical equations for the small-signal model of a grid-type (GFM) converter are as follows: ; in, For state vectors, The state matrix, The input matrix is ​​denoted as . The input vector. For the output matrix, This is for direct matrix transmission. This is the output vector.

[0035] Solving the state matrix , The eigenvalues ​​are the eigenvalues ​​required for the analysis of the small-signal model of the system.

[0036] S113. Combining the mathematical models of the grid-connected inverter and the grid-connected inverter, the state-space small-signal models of the grid-connected inverter system in both grid-connected and grid-connected modes are obtained.

[0037] S120. Solve for the eigenvalues ​​of the state matrix in the state-space small-signal model to obtain the eigenvalue distribution of the grid-connected inverter system under different short-circuit ratios and active power conditions. S130. Based on the eigenvalue distribution, the damping ratio data of the grid-connected inverter system under different short-circuit ratios and active power per unit values ​​are extracted, and a three-dimensional characteristic damping dataset is constructed. In this embodiment, a state-space small-signal model of the grid-connected converter system under grid-following (GFL) and grid-forming (GFM) modes is established. The dominant eigenvalues ​​and corresponding characteristic damping ratios of the system under different short-circuit ratios (SCR) and per-unit active power (P) conditions are obtained by solving the eigenvalues, and a three-dimensional nonlinear characteristic damping dataset is constructed.

[0038] S140. Based on the three-dimensional feature damping dataset, the preset fuzzy neural network is trained to obtain the pre-trained fuzzy neural network.

[0039] In this embodiment, a fuzzy neural network is established and trained based on a three-dimensional feature damping dataset to establish a model from the two-dimensional input variable (SCR, P) to the control confidence signal. The nonlinear mapping relationship, where, It is a continuous scalar that characterizes the stability margin of the system, and its value range is [0, 1].

[0040] S200. Initialize the grid-connected inverter system and perform real-time operating parameter sensing on the initialized grid-connected inverter system to obtain the real-time short-circuit ratio estimate and the current per-unit value of the output active power of the grid-connected inverter system. In this embodiment, during the operation of the inverter, the voltage and current signals at the grid connection point are collected in real time, the real-time short-circuit ratio (SCR) estimate is obtained through the grid impedance identification algorithm, and the per-unit value of the active power output in real time (P) is collected simultaneously.

[0041] S210. Set the initial operating mode of the grid-connected inverter system to grid-following mode, and assign an initial state value to the unit delay element in the nonlinear hysteresis logic. S220: Collect the grid connection point voltage and current signals of the grid-connected inverter system during operation; S230. Based on the grid connection point voltage and current signals, the grid impedance identification algorithm is used to estimate and obtain the real-time short-circuit ratio estimate of the grid-connected inverter system, and simultaneously obtain the current per-unit value of the output active power of the grid-connected inverter system.

[0042] In this embodiment, the initial mode of grid-connected converter operation is set to grid-following control mode, and an initial state value is assigned to the unit delay element in the nonlinear hysteresis logic. After the system starts, the voltage and current signals at the grid connection point are collected in real time, and the real-time short-circuit ratio (SCR) estimate is obtained through the grid impedance identification algorithm. At the same time, the per-unit value (P) of the active power output of the converter is obtained.

[0043] Specifically, based on the grid-connected inverter system, all key identification parameters are pre-set. These key parameters include the injected non-characteristic harmonic frequency, complex filter parameters, notch filter parameters, and fundamental angular frequency. Based on the grid-connected inverter system with the key identification parameters set, a sinusoidal harmonic current of preset amplitude is injected, and the grid connection point voltage and current signals are simultaneously acquired. Target harmonic components are extracted from the grid connection point voltage and current signals to obtain the voltage harmonic components and current harmonic components in the stationary coordinate system. The equivalent impedance parameters of the grid are calculated based on the voltage and current harmonic components to obtain the grid impedance amplitude. Based on the grid impedance amplitude, the real-time short-circuit ratio estimate of the grid-connected inverter system is determined, and the current per-unit output active power of the grid-connected inverter system is simultaneously acquired.

[0044] In this embodiment, the harmonic injection method is used to detect changes in the grid impedance, such as... Figure 14 As shown. Considering the influence of actual power grid background harmonic content, the injected non-characteristic harmonic frequency is selected as 75Hz. Parameters of the complex filter. Take 30 rad / s, Take 140 rad / s. A notch filter is connected to the output of the complex filter. The resonant frequency of the notch filter is... After setting the frequency to -471 rad / s (corresponding to -75 Hz) and extracting the voltage and current harmonic signals, The formula for calculating the magnitude of the power grid impedance in a coordinate system is: ; In the above formula: This represents the grid connection point voltage, in V; Indicates the non-characteristic subharmonic frequency, in Hz; subscript , Representing electrical quantities , Quantity; Represents the imaginary unit; This represents the current injected into the grid connection point by the converter, in A; Represents the equivalent resistance of the power grid, in Ω; Indicates the equivalent inductance of the power grid; Represents the non-characteristic subharmonic angular frequency, in rad / s; Represents the equivalent impedance of the power grid, in Ω; Indicates the fundamental angular frequency. .

[0045] S300: The real-time short-circuit ratio estimate of the grid-connected inverter system and the per-unit value of the current output active power are input into the pre-trained fuzzy neural network for online fuzzy neural network inference, and nonlinear hysteresis mode switching is performed according to the inference results to realize the operation control of the grid-connected inverter system.

[0046] In this embodiment, as Figure 13 As shown, the fuzzy neural network established by the acquired real-time SCR and P input undergoes fuzzification processing, fuzzy rule inference, and defuzzification calculation to output the continuous control confidence signal under the current operating condition. The obtained confidence signal In nonlinear hysteresis logic with Schmitt triggering characteristics: when When the threshold for triggering network construction is less than the preset threshold, switch to network construction mode (GFM); when When the trigger limit is exceeded, switch to the GFL (Gate Follower) control mode; when When the value is between the two thresholds mentioned above, the hold logic is triggered to maintain the operating mode of the previous control cycle.

[0047] S310. Input the estimated real-time short-circuit ratio of the grid-connected inverter system and the per-unit value of the current output active power into the pre-trained fuzzy neural network for fuzzification processing, fuzzy rule reasoning and defuzzification calculation in sequence to obtain the continuous control confidence signal of the grid-connected inverter system under the current operating condition. S320: Based on the continuous control confidence signal of the grid-connected inverter system under the current operating conditions, perform nonlinear hysteresis mode switching to realize the operation control of the grid-connected inverter system.

[0048] Specifically, the continuous control confidence signal of the grid-connected inverter system under the current operating condition is input to a nonlinear hysteresis logic with Schmitt triggering characteristics for comparison. If the continuous control confidence signal of the grid-connected inverter system under the current operating condition is less than the preset lower threshold for grid connection triggering, the operating mode of the grid-connected inverter system is switched to grid connection mode. If the continuous control confidence signal of the grid-connected inverter system under the current operating condition is greater than the preset upper threshold for grid connection triggering, the operating mode of the grid-connected inverter system is switched to grid connection mode. If the continuous control confidence signal of the grid-connected inverter system under the current operating condition is between the preset lower threshold for grid connection triggering and the preset upper threshold for grid connection triggering, the holding logic is triggered to maintain the operating mode of the grid-connected inverter system in the previous control cycle, thereby realizing the operation control of the grid-connected inverter system.

[0049] In this embodiment, the fuzzy neural network established by the acquired real-time SCR and P input undergoes fuzzification processing, fuzzy rule inference, and defuzzification calculation to output a continuously changing control confidence signal in real time. ; Judgment conditions If the condition is met, then the mode switching logic is triggered, such as... Figure 12 As shown, the grid-connected converter will be smoothly switched from grid-following (GFL) mode to grid-forming (GFM) mode; otherwise, the current operating mode will be maintained and the system will return to continue sensing parameters. Judgment conditions If the condition is met, then the reverse switching logic is triggered, such as... Figure 12 As shown, the grid-connected converter is switched from grid-connected mode (GFM) to grid-following mode (GFL); if it is in [ Within the hysteresis buffer, the state preservation logic is triggered to maintain the current network configuration mode.

[0050] Substituting the data from Table 1 into the solution for the system eigenvalues, while keeping other system parameters constant, and only changing the grid equivalent impedance (SCR) and the per-unit value of active power (P), we can observe the changes in the eigenvalues ​​of the dominant mode of the system under GFL / GFM control. The results are as follows: Figure 15 , Figure 16The short-circuit ratio and actual output power per unit value are shown to affect the dominant modes of GFL and GFM control.

[0051] Table 1 System Parameters

[0052] To adapt to the actual operating requirements of the power grid, changes to the original structure should be minimized, and the original operating state should be maintained unless absolutely necessary. The switchable unit should only be switched to GFM control when there is a risk of instability during grid-connected (GFL) operation; otherwise, it should continue to operate in grid-connected mode. Therefore, the switching boundary tuning should fully consider the stability boundary of the GFL control unit. Figure 15 The three-dimensional feature damping dataset is used to establish and train a fuzzy neural network, which establishes a relationship between the two-dimensional input variables (SCR, P) and the control confidence signal (…). The nonlinear mapping relationship of the fuzzy neural network dataset is shown in Table 2. A value of 0 indicates that the system needs to switch to network mode. A value of 0.5 indicates that the system remains in its original state. A value of 1 indicates that the system needs to switch to network mode.

[0053] Table 2 Partial Dataset of Fuzzy Neural Networks

[0054] In summary, this invention first senses the short-circuit ratio (SCR) at the grid connection point and the per-unit value (P) of the converter's active power in real time, and inputs them as independent variables into a pre-trained fuzzy neural network. Then, the fuzzy neural network performs two-dimensional fuzzy inference based on the nonlinear evolution law of the small-signal characteristic damping ratio of the power system, outputting a continuous signal representing the stability confidence of the current operating condition. Finally, this confidence signal is introduced into a Schmitt triggering logic with nonlinear hysteresis characteristics to determine and issue either a grid-following (GFL) or grid-forming (GFM) control command. This invention abandons the traditional one-dimensional hard-boundary switching mode of the operating short-circuit ratio (OSCR), and accurately fits the complex damping collapse boundary through a fuzzy neural network, reducing the number of frequent switching caused by operating condition fluctuations, reducing transient impacts on power devices, and improving the reliability of converter operation under all operating conditions in weak power grids.

[0055] Reference Figure 2 A converter control system based on characteristic damping and fuzzy neural networks includes: The first module 201 is used to perform eigenvalue analysis on the grid-connected inverter system to construct a three-dimensional feature damping dataset, and to train a preset fuzzy neural network to obtain a pre-trained fuzzy neural network. The second module 202 is used to initialize the grid-connected inverter system and to perform real-time operating parameter sensing on the initialized grid-connected inverter system to obtain the real-time short-circuit ratio estimate and the current per-unit value of the output active power of the grid-connected inverter system. The third module 203 is used to input the real-time short-circuit ratio estimate of the grid-connected inverter system and the current per-unit value of the output active power into the pre-trained fuzzy neural network for online fuzzy neural network inference, and to perform nonlinear hysteresis mode switching based on the inference results to realize the operation control of the grid-connected inverter system.

[0056] The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0057] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this is not intended to limit the scope of the embodiments of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. A converter control method based on characteristic damping and fuzzy neural network, characterized in that, Includes the following steps: Eigenvalue analysis is performed on the grid-connected inverter system to construct a three-dimensional characteristic damping dataset, and a pre-trained fuzzy neural network is trained to obtain the pre-trained fuzzy neural network. Initialize the grid-connected inverter system and perform real-time operating parameter sensing on the initialized grid-connected inverter system to obtain the real-time short-circuit ratio estimate and the current per-unit value of the output active power of the grid-connected inverter system. The real-time short-circuit ratio estimate of the grid-connected inverter system and the per-unit value of the current output active power are input into the pre-trained fuzzy neural network for online fuzzy neural network inference, and nonlinear hysteresis mode switching is performed based on the inference results to realize the operation control of the grid-connected inverter system.

2. The converter control method based on characteristic damping and fuzzy neural network according to claim 1, characterized in that, The step of constructing a three-dimensional feature damping dataset by performing eigenvalue analysis on the grid-connected inverter system and training a pre-trained fuzzy neural network to obtain the pre-trained fuzzy neural network specifically includes: Construct small-signal state-space models of grid-connected inverter systems in both grid-following and grid-connected modes; The eigenvalues ​​of the state matrix in the state-space small-signal model are solved to obtain the eigenvalue distribution of the grid-connected inverter system under different short-circuit ratios and active power conditions. Based on the eigenvalue distribution, damping ratio data of grid-connected inverter systems under different short-circuit ratios and active power conditions are extracted to construct a three-dimensional characteristic damping dataset. Based on a three-dimensional feature damping dataset, a pre-set fuzzy neural network is trained to obtain a pre-trained fuzzy neural network.

3. The converter control method based on characteristic damping and fuzzy neural network according to claim 2, characterized in that, The step of constructing the state-space small-signal model of the grid-connected inverter system in both grid-following and grid-connected modes specifically includes: The nonlinear dynamic equations of each structure of the grid converter are determined, and small-signal perturbation linearization is performed at the steady-state operating point to construct a mathematical model of the grid converter. The nonlinear dynamic equations of each structure of the grid converter are determined, and small-signal perturbation linearization is performed at the steady-state operating point to construct a mathematical model of the grid converter. By combining the mathematical models of grid-connected inverters and grid-connected inverters, the small-signal state-space models of the grid-connected inverter system in both grid-connected and grid-connected modes are obtained.

4. The converter control method based on characteristic damping and fuzzy neural network according to claim 3, characterized in that, The step of determining the nonlinear dynamic equations of each structure of the grid converter, performing small-signal perturbation linearization at the steady-state operating point, and constructing the mathematical model of the grid converter specifically includes: Based on the main circuit topology of the grid converter, and combining Kirchhoff's voltage law and Kirchhoff's current law, the nonlinear dynamic equations of the main circuit of the grid converter are constructed. Based on the current inner loop control structure of the grid-connected converter, and combined with the PI and feedforward decoupling principle, the nonlinear dynamic equation of the current inner loop control structure of the grid-connected converter is constructed. Based on the PQ outer loop control structure of the grid converter, and combined with instantaneous power theory, the nonlinear dynamic equation of the PQ outer loop control structure of the grid converter is constructed. Based on the phase-locked loop structure of the grid converter and combined with the principle of PI-type phase-locked loop, the nonlinear dynamic equation of the phase-locked loop structure of the grid converter is constructed. Based on the coordinate transformation equation between the controller coordinate system and the global coordinate system, the nonlinear dynamic equations of the main circuit of the grid-connected converter, the nonlinear dynamic equations of the inner current loop control structure of the grid-connected converter, the nonlinear dynamic equations of the PQ outer loop control structure of the grid-connected converter, and the nonlinear dynamic equations of the phase-locked loop structure of the grid-connected converter are linearized at the operating point, and the corresponding connection relationships are determined to construct the mathematical model of the grid-connected converter.

5. The converter control method based on characteristic damping and fuzzy neural network according to claim 3, characterized in that, The step of determining the nonlinear dynamic equations of each structure of the grid converter, performing small-signal perturbation linearization at the steady-state operating point, and constructing the mathematical model of the grid converter specifically includes: Based on the main circuit topology of the grid converter, and combining Kirchhoff's voltage law and Kirchhoff's current law, the nonlinear dynamic equations of the main circuit of the grid converter are constructed. Based on the current inner loop control structure of the grid converter, and combined with the PI and feedforward decoupling principle, the nonlinear dynamic equation of the current inner loop control structure of the grid converter is constructed. Based on the voltage and reactive power outer loop control structure of the grid converter, and combined with the cascaded PI control principle, the nonlinear dynamic equation of the voltage and reactive power outer loop control structure of the grid converter is constructed. Based on the power synchronization loop structure of the grid converter, and combined with the rotor motion equation of the synchronous generator, a nonlinear dynamic equation for the power synchronization loop structure of the grid converter is constructed. Based on the coordinate transformation equations between the controller coordinate system and the global coordinate system, the nonlinear dynamic equations of the main circuit of the grid converter, the nonlinear dynamic equations of the inner current loop control structure of the grid converter, the nonlinear dynamic equations of the outer voltage and reactive power loop control structure of the grid converter, and the nonlinear dynamic equations of the power synchronization loop structure of the grid converter are linearized at the operating point, and the corresponding connection relationships are determined to construct a mathematical model of the grid converter.

6. The converter control method based on characteristic damping and fuzzy neural network according to claim 5, characterized in that, The step of initializing the grid-connected inverter system and sensing its real-time operating parameters to obtain the estimated real-time short-circuit ratio and the current per-unit value of the output active power specifically includes: The initial operating mode of the grid-connected inverter system is set to grid-following mode, and an initial state value is assigned to the unit delay element in the nonlinear hysteresis logic. Collect grid connection point voltage and current signals of the grid-connected inverter system during operation; Based on the grid connection point voltage and current signals, the grid impedance identification algorithm is used to estimate the real-time short-circuit ratio of the grid-connected inverter system and simultaneously obtain the per-unit value of the current output active power of the grid-connected inverter system.

7. The converter control method based on characteristic damping and fuzzy neural network according to claim 6, characterized in that, The step of estimating the real-time short-circuit ratio of the grid-connected inverter system based on the grid connection point voltage and current signals using a grid impedance identification algorithm, and simultaneously obtaining the current per-unit value of the grid-connected inverter system's output active power, specifically includes: Based on the grid-connected inverter system, all key identification parameters are pre-set, including the injected non-characteristic harmonic frequency, complex filter parameters, notch filter parameters, and fundamental angular frequency; Based on the grid-connected inverter system after setting and identifying key parameters, a sinusoidal harmonic current of preset amplitude is injected, and the voltage and current signals at the grid connection point are collected simultaneously. Target harmonic components are extracted from the voltage and current signals at the grid connection point to obtain the voltage harmonic components and current harmonic components in the stationary coordinate system. The equivalent impedance parameters of the power grid are calculated based on the voltage harmonic components and the current harmonic components, and the power grid impedance amplitude is obtained. Based on the grid impedance amplitude, determine the estimated real-time short-circuit ratio of the grid-connected inverter system, and simultaneously obtain the per-unit value of the current output active power of the grid-connected inverter system.

8. The converter control method based on characteristic damping and fuzzy neural network according to claim 7, characterized in that, The step of inputting the real-time short-circuit ratio estimate of the grid-connected inverter system and the current per-unit value of the output active power into a pre-trained fuzzy neural network for online fuzzy neural network inference, and performing nonlinear hysteresis mode switching based on the inference results to achieve operation control of the grid-connected inverter system, specifically includes: The real-time short-circuit ratio estimate of the grid-connected inverter system and the current per-unit value of the output active power are input into the pre-trained fuzzy neural network for fuzzification, fuzzy rule reasoning and defuzzification calculation in sequence to obtain the continuous control confidence signal of the grid-connected inverter system under the current operating conditions. Based on the continuous control confidence signal of the grid-connected inverter system under the current operating conditions, nonlinear hysteresis mode switching is performed to realize the operation control of the grid-connected inverter system.

9. The converter control method based on characteristic damping and fuzzy neural network according to claim 8, characterized in that, The step of performing nonlinear hysteresis mode switching based on the continuous control confidence signal of the grid-connected inverter system under the current operating conditions to achieve operation control of the grid-connected inverter system specifically includes: The continuous control confidence signal of the grid-connected inverter system under the current operating condition is input into a nonlinear hysteresis logic with Schmitt triggering characteristics for comparison. If the continuous control confidence signal of the grid-connected inverter system under the current operating conditions is less than the preset grid-connection trigger lower threshold, the operating mode of the grid-connected inverter system will be switched to grid-connection mode. If the continuous control confidence signal of the grid-connected inverter system under the current operating conditions is greater than the preset grid-following trigger upper limit threshold, the operating mode of the grid-connected inverter system will be switched to grid-following mode. If the continuous control confidence signal of the grid-connected inverter system under the current operating condition is between the preset lower threshold for grid connection triggering and the preset upper threshold for grid connection triggering, then the hold logic is triggered to maintain the operating mode of the grid-connected inverter system in the previous control cycle, thereby realizing the operation control of the grid-connected inverter system.

10. A converter control system based on characteristic damping and fuzzy neural network, characterized in that, Includes the following modules: The first module is used to perform eigenvalue analysis on the grid-connected inverter system to construct a three-dimensional feature damping dataset, and to train a preset fuzzy neural network to obtain a pre-trained fuzzy neural network. The second module is used to initialize the grid-connected inverter system and to sense the real-time operating parameters of the initialized grid-connected inverter system, thereby obtaining the estimated real-time short-circuit ratio and the current per-unit value of the output active power of the grid-connected inverter system. The third module is used to input the real-time short-circuit ratio estimate of the grid-connected inverter system and the current per-unit value of the output active power into the pre-trained fuzzy neural network for online fuzzy neural network inference, and to perform nonlinear hysteresis mode switching based on the inference results to realize the operation control of the grid-connected inverter system.

Images