Communication-interconnection-free cooperative autonomous synchronous network construction method for multi-network-construction type energy storage converters

Abstract

The invention discloses a cooperative autonomous synchronous networking method for a multi-networking type energy storage converter without communication interconnection. The method comprises the following steps: S1, state perception and estimation based on local measurement: acquiring local electrical parameters in real time by using a sensor, and performing state estimation through an improved Kalman filtering algorithm and a local dynamic model; s2, a cooperation mechanism of the virtual synchronous machine and droop control: combining the virtual synchronous machine with the droop control mechanism, and realizing automatic power adjustment and cooperation work by setting a virtual inertia, a damping coefficient and a droop coefficient; s3, self-adaptive synchronization algorithm: based on the local state estimation and the output of the virtual synchronous machine, adjusting the phase and frequency of the output voltage, and dynamically adjusting synchronization parameters according to the state of the power grid; and S4, detecting a fault by monitoring local electrical parameters and a self state in real time, and automatically adjusting a control strategy and realizing system fault tolerance when the fault occurs.

Description

Technical Field

[0001] The present invention relates to the field of power system energy storage technology, and in particular to a technology for realizing collaborative and autonomous synchronous networking of multi-network type energy storage converters without communication interconnection. Background Art

[0002] In modern power systems, energy storage converters play a key role in maintaining the stable operation of the power grid and improving the quality of power. With the large-scale integration of distributed energy resources and the increasing complexity of power systems, higher requirements are placed on the collaborative operation of multiple energy storage converters. Traditional communication-based collaborative methods have problems such as communication delays and reliability affected by the communication network. In some complex environments or areas with weak communication infrastructure, achieving collaborative operation of multi-network energy storage converters without communication interconnection has become a difficult problem that needs to be solved urgently. Current related technologies have deficiencies in synchronization accuracy, stability, and adaptability in the absence of communication, and cannot meet the requirements for efficient operation of power systems. For example, in microgrids in remote mountainous areas, communication network coverage is not perfect, and traditional collaborative technologies are difficult to apply, which seriously affects the support effect of energy storage systems on local power grids. Summary of the Invention

[0003] In view of this, the present invention proposes a method for collaborative and autonomous synchronous networking of multiple grid-type energy storage converters without communication interconnection. The method aims to enable multiple grid-type energy storage converters to automatically sense each other's status and achieve high-precision synchronous networking without relying on external communication by designing a unique control algorithm and synchronization mechanism, thereby solving the problem of global autonomous collaborative synchronization of multiple energy storage converters without communication interconnection, thereby improving the stability and reliability of the power system and the adaptability of the energy storage system in complex environments.

[0004] A method for collaborative and autonomous synchronous networking of multi-network energy storage converters without communication interconnection, comprising the following steps: S1. State perception and estimation based on local measurement: using sensors to collect local electrical parameters in real time and performing state estimation through an improved Kalman filter algorithm and a local dynamic model; S2. Collaborative mechanism of virtual synchronous machine and droop control: combining the virtual synchronous machine with the droop control mechanism, and realizing automatic power regulation and collaborative work by setting virtual inertia, damping coefficient and droop coefficient; S3. Adaptive synchronization algorithm: adjusting the output voltage phase and frequency based on local state estimation and the output of the virtual synchronous machine, and dynamically adjusting the synchronization parameters according to the grid state.

[0005] Furthermore, the following steps are included: S4, detecting faults by real-time monitoring of local electrical parameters and self-state, and automatically adjusting control strategies and achieving system fault tolerance when a fault occurs.

[0006] Furthermore, the improved Kalman filter algorithm in step S1 adopts a method based on extended Kalman filter, combined with a nonlinear model of the energy storage converter, to iteratively update the state variables to achieve the state estimation.

[0007] Furthermore, the state equation of the improved Kalman filter algorithm described in step S1 is x(k+1)=f(x(k), u(k))+w(k), and the measurement equation is z(k)=h(x(k))+v(k); wherein f(·) is a nonlinear state transfer function, u(k) is an input vector; w(k) is a process noise vector, which satisfies a Gaussian distribution with a mean of zero and a covariance matrix of Q(k); z(k) is a measurement vector, h(·) is a measurement function; v(k) is a measurement noise vector, which satisfies a Gaussian distribution with a mean of zero and a covariance matrix of R(k).

[0008] Furthermore, in the extended Kalman filter method, a prediction step is first performed:

[0009]

[0010] P - (k+1)=F(k)P(k)F(k) T +Q(k),

[0011] in, is the estimated value of the predicted state, P - (k+1) is the predicted covariance matrix, F(k) is the Jacobian matrix of the state transfer function f(·) with respect to the state variable x(k);

[0012] Then do the update steps:

[0013] K(k+1)=P - (k+1)H(k+1) T [H(k+1)P - (k+1)H(k+1) T +R(k+1) T ] -1 ,

[0014] P(k+1)=[IK(k+1)H(k+1)]P - (k+1),

[0015] Where K(k+1) is the Kalman gain, H(k+1) is the Jacobian matrix of the measurement function h(·) with respect to the state variable x(k+1), is the updated state estimate, and P(k+1) is the updated covariance matrix.

[0016] Furthermore, step S1 also includes: establishing the local dynamic model based on the change trend and historical data of local electrical parameters to predict changes in its own state, adjust the control strategy in advance, and improve the response speed and stability of the system; the local dynamic model is based on the long short-term memory network LSTM in machine learning, which can capture the time series characteristics of electrical parameters.

[0017] Furthermore, step S2 includes: using the rated power P of the energy storage converter rated The power output boundaries of the virtual inertia and damping coefficient are constrained, the short-circuit ratio is used to determine the parameter stable feasible domain, and an optimization algorithm is used to determine the Pareto optimal solution of the virtual inertia and damping coefficient within the stable feasible domain. Combined with the droop control strategy, the output power of the energy storage converter is dynamically adjusted according to the locally measured frequency and voltage deviations.

[0018] Furthermore, the optimization algorithm in step S2 includes: using a genetic algorithm or a particle swarm optimization algorithm, with the objective function J being to minimize the weighted square sum of the system frequency deviation Δf and voltage deviation ΔV obj , optimize the virtual inertia and the damping coefficient:

[0019] J obj =α(Δf) 2 +β(ΔV) 2 ;

[0020] Among them, α and β are weight coefficients, which are set according to the grid's emphasis on frequency and voltage stability. During the algorithm execution process, the values ​​of the virtual inertia and the damping coefficient are continuously adjusted to simulate different operating scenarios, calculate the corresponding objective function value, and find the optimal solution for the virtual inertia and the damping coefficient after multiple iterations.

[0021] Furthermore, the dynamic adjustment of the output power of the energy storage converter in step S2 includes:

[0022] Assume the frequency droop coefficient is m f , the voltage droop coefficient is m v When the grid frequency deviation is Δf and the voltage deviation is ΔV, the output power adjustment ΔP and reactive power adjustment ΔQ of the energy storage converter are expressed as follows:

[0023] ΔP=-m f Δf;

[0024] ΔQ=-m v ΔV;

[0025] Different energy storage converters automatically share power changes without communication based on their own droop characteristics, achieving collaborative work; when the grid frequency drops, the energy storage converter automatically increases output power; when the grid frequency rises, the energy storage converter automatically reduces output power to maintain grid frequency stability.

[0026] Furthermore, step S2 also includes differentially configuring the droop coefficient according to the capacity C of the energy storage converter: for a larger capacity energy storage converter, relatively smaller frequency droop coefficient and voltage droop coefficient are set so that it plays the main role in power regulation; and for a smaller capacity energy storage converter, relatively larger frequency droop coefficient and voltage droop coefficient are set so that it plays an auxiliary and fine-tuning role in power regulation.

[0027] Furthermore, step S2 further includes: according to the remaining capacity C of the energy storage converter remain The droop coefficient is dynamically adjusted based on the health status H, including the use of linear or nonlinear functional relationships:

[0028]

[0029]

[0030] in, and are the initial frequency droop coefficient and voltage droop coefficient, k f 、k v 、l f 、l v is the preset adjustment factor.

[0031] Furthermore, step S3 includes: setting the synchronization error to be e=θ-θ ref , where θ, θ ref are the output voltage phase and reference phase of the energy storage converter respectively; a variable structure control strategy is adopted to dynamically adjust the control gain according to the size of the synchronization error. In the early stage of synchronization, when the synchronization error is large, a larger control gain K1 is adopted to speed up the convergence speed; as the synchronization process proceeds, the synchronization error gradually decreases. At this time, the control gain is reduced to improve the synchronization accuracy and avoid over-adjustment.

[0032] Furthermore, the control gain is dynamically adjusted according to the size of the synchronization error, including: setting the thresholds of the synchronization error to E1 and E2, where E1>E2; when the synchronization error is greater than E1, the control gain is set to K1; when the synchronization error is between E1 and E2, the control gain K is linearly adjusted. Decrease, where K2 is a preset smaller control gain; when the synchronization error is less than E2, the control gain is set to K2.

[0033] Furthermore, in step S3, the phase and frequency of the output voltage are dynamically adjusted, wherein:

[0034] The frequency adjustment formula is:

[0035]

[0036] Among them, K p is a preset proportional coefficient, determined according to the system dynamic performance requirements;

[0037] The phase adjustment formula is:

[0038]

[0039] Among them, K i is the integral coefficient, which is used to eliminate steady-state errors.

[0040] Furthermore, step S4 specifically includes: each energy storage converter has a built-in fault detection module, which can quickly detect short circuit, overcurrent, and overvoltage faults by real-time monitoring of local electrical parameters and its own status; once a fault is detected, the fault tolerance mechanism is immediately activated, and the fault information is notified to the operation and maintenance personnel through a local display or alarm device, and the control strategy is automatically adjusted to reduce the impact of the fault on the system; among them, the fault detection adopts a method based on a combination of threshold judgment and wavelet transform, first preliminarily judging whether there is a fault by setting a threshold, and then using wavelet transform to extract and analyze the characteristics of the fault signal, thereby improving the accuracy and reliability of fault detection.

[0041] Furthermore, the fault detection process of step S4 specifically includes: setting the voltage threshold as V th , the current threshold is I th and the frequency threshold is f th , when the measured voltage u(t) satisfies |u(t)|>V th , or the current i(t) satisfies |i(t)|>I th , or the frequency f(t) satisfies |f(t)-f nom |>f th When f nom is the rated frequency, triggering the wavelet transform analysis; the wavelet transform uses the preset wavelet basis function to perform multi-scale decomposition on the fault signal s(t):

[0042]

[0043] Among them, c j,k is the wavelet coefficient, ψ j，k It is a wavelet function, and the fault type and severity are determined by analyzing the variation pattern and energy distribution of the wavelet coefficients.

[0044] Furthermore, step S4 also includes: in the event that some energy storage converters fail, other normal energy storage converters can automatically adjust their working modes, take on more power regulation tasks, maintain the stable operation of the power grid, and achieve the fault tolerance of the system; specifically, it includes: adopting a distributed coordinated control strategy to redistribute power regulation tasks according to the capacity and performance of the remaining normal equipment to ensure the stable operation of the power grid.

[0045] Furthermore, in the distributed coordinated control, let the set of remaining normal energy storage converters be S. For each energy storage converter i∈S, its new power regulation target is Calculated according to the following formula:

[0046]

[0047] in, is the original power regulation target, ΔP total is the power shortage caused by the fault, C i is the capacity of energy storage converter i; each energy storage converter exchanges information with its neighboring energy storage converters locally. Through iterative calculation, the power regulation targets of all energy storage converters are made consistent, thereby achieving reasonable power distribution and stable operation of the system.

[0048] The present invention further proposes a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the aforementioned method for collaborative and autonomous synchronous networking of multi-networked energy storage converters can be implemented.

[0049] The beneficial effects of the technical solution of the present invention are reflected in: the method for collaborative and autonomous synchronous networking of multi-grid energy storage converters without communication interconnection proposed in the present invention, through state perception and estimation based on local measurement, a collaborative mechanism combining virtual synchronous machine and droop control, and an adaptive synchronization algorithm, while designing a unique control algorithm and synchronization mechanism, so that multiple grid-type energy storage converters can automatically sense each other's status and achieve high-precision synchronous networking without relying on external communication, thereby improving the stability and reliability of the power system and enhancing the adaptability of the energy storage system in complex environments. It solves the problem of collaborative synchronization of multi-grid energy storage converters without communication in the existing technology, reduces dependence on communication infrastructure, and has important application value.

[0050] In areas with high penetration of distributed energy, it can significantly improve the absorption capacity of new energy and reduce the phenomenon of wind and solar power curtailment. It also reduces dependence on communication infrastructure, reduces system construction and maintenance costs, and improves the overall economy and practicality of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1It is a flow chart of a method for collaborative and autonomous synchronous networking of multi-networked energy storage converters without communication interconnection according to an embodiment of the present invention.

[0052] Figure 2 This is an architecture diagram of a multi-networked energy storage converter collaborative autonomous synchronous networking system without communication interconnection according to an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The present invention will be further described below with reference to the accompanying drawings, specific implementation methods, and examples. The examples are provided for illustration only and are not intended to limit the scope of protection of the present invention.

[0054] The embodiment of the present invention provides a method for collaborative autonomous synchronization of multi-network energy storage converters without communication interconnection, referring to Figure 1 The method includes the following steps: S1, state perception and estimation based on local measurement: using sensors to collect local electrical parameters in real time and performing state estimation through an improved Kalman filter algorithm and a local dynamic model; S2, collaborative mechanism of virtual synchronous machine and droop control: combining the virtual synchronous machine with the droop control mechanism, and realizing automatic power regulation and collaborative work by setting virtual inertia, damping coefficient and droop coefficient; S3, adaptive synchronization algorithm: adjusting the output voltage phase and frequency based on local state estimation and the output of the virtual synchronous machine, and dynamically adjusting the synchronization parameters according to the power grid state; S4, detecting faults by real-time monitoring of local electrical parameters and its own state, and automatically adjusting the control strategy and realizing system fault tolerance in the event of a fault.

[0055] In this embodiment of the present invention, each grid-type energy storage converter is equipped with high-precision voltage, current, and frequency sensors to collect local electrical parameters in real time. Let the collected voltage be u(t), current be i(t), and frequency be f(t). Based on this local measurement data, advanced signal processing algorithms (an improved Kalman filter algorithm) and state estimation techniques are used to accurately estimate the operating state of the energy storage converter, including key information such as amplitude, phase, and frequency.

[0056] The improved Kalman filter algorithm described in step S1 uses an extended Kalman filter-based approach, combined with a nonlinear model of the energy storage converter, to iteratively update state variables and achieve state estimation. This algorithm effectively reduces the impact of measurement noise (such as white noise collected by sensors) and system uncertainty (such as the nonlinearity of the energy storage converter's switching process), improves estimation accuracy, and accurately estimates state variables (such as voltage amplitude, phase, and frequency), providing reliable data for subsequent control strategies.

[0057] Step S1 is based on the improved Kalman filter algorithm. For the state space model of the energy storage converter, the state variable is set to be x=[x1, x2, ..., x n ]T , where x1 represents the voltage amplitude, x2 represents the voltage phase, and x3 represents the frequency. Other variables can also be included, such as the current amplitude, phase, and frequency, as well as active power, reactive power, and apparent power. The state equation can be expressed as x(k+1)=f(x(k),u(k))+w(k), and the measurement equation is z(k)=h(x(k))+v(k). The prediction and update steps follow the standard EKF (Extended Kalman Filter) algorithm. Here, f(·) is the nonlinear state transfer function, u(k) is the input vector (such as the control input); w(k) is the process noise vector, which has a zero-mean Gaussian distribution and a covariance matrix Q(k); z(k) is the measurement vector (such as the measured electrical parameters such as voltage, current, frequency, and power); h(·) is the measurement function; and v(k) is the measurement noise vector, which has a zero-mean Gaussian distribution and a covariance matrix R(k). In these formulas, k is the discrete time step index, corresponding to the sampling instant of continuous time t (t=k·Ts), which is used to describe the iterative process of the discretized state space model.

[0058] Furthermore, in the Extended Kalman Filter (EKF) method, a prediction step is first performed:

[0059]

[0060] P - (k+1)=F(k)P(k)F(k) T +Q(k),

[0061] in, is the predicted state estimate (prior estimate), P - (k+1) is the predicted covariance matrix, F(k) is the Jacobian matrix of the state transfer function f(·) with respect to the state variable x(k);

[0062] Then do the update steps:

[0063] K(k+1)=P - (k+1)H(k+1) T [H(k+1)P - (k+1)H(k+1) T +R(k+1) T ] -1 ,

[0064]

[0065] P(k+1)=[IK(k+1)H(k+1)]P - (k+1),

[0066] Where K(k+1) is the Kalman gain, H(k+1) is the Jacobian matrix of the measurement function h(·) with respect to the state variable x(k+1), is the updated state estimate (posterior optimization value), and P(k+1) is the updated covariance matrix. In this way, the impact of measurement noise and system uncertainty is effectively reduced, and the estimation accuracy is improved.

[0067] At the same time, based on the changing trends and historical data of local electrical parameters, a local dynamic model is established to predict changes in its own state, adjust the control strategy in advance, and improve the response speed and stability of the system. This local dynamic model is based on the long short-term memory network (LSTM) in machine learning and can capture the time series characteristics of electrical parameters. Suppose the input sequence is {u(tN), u(t-N+1), ..., u(t)} (where u can represent parameters such as voltage, current or frequency, and N is the length of the time series). After processing by the LSTM network, the predicted state value is output The internal calculation process of the LSTM network includes the operations of the input gate, forget gate, and output gate. The calculation formula is as follows:

[0068] Input Gate:

[0069] Forget Gate:

[0070] Output Gate:

[0071] Cell status update:

[0072] Hidden state output: h t =o t ⊙tanh(c t );

[0073] Among them, σ(·) is the sigmoid function, ⊙ represents element-wise multiplication; W and b represent the weight and bias parameters of each gate, respectively. The weight and bias parameters are continuously adjusted through training to optimize the prediction performance. Specifically, W xi 、W hi is the weight of the input gate, which controls the input data x t 、h t-1 The degree of influence on the input gate, b i is the bias parameter of the input gate; for the forget gate, output gate and cell state update, the variable explanation is similar and will not be repeated here.

[0074] Hidden state h t Calculated by the LSTM gating mechanism, the time series characteristics of the electrical parameters are stored, and the predicted state value is obtained after mapping through the output layer Directly used for state estimation and collaborative control in non-communication scenarios.

[0075] The collaborative mechanism of virtual synchronous machine and droop control described in step S2 introduces virtual synchronous machine (VSG) technology to simulate the operating characteristics of synchronous generator, so that the energy storage converter has inertia and damping characteristics. By properly setting the virtual inertia J and damping coefficient D, it can automatically provide frequency and voltage support when the power grid is disturbed, thereby enhancing the stability of the power grid. rated The power output boundary of the virtual inertia and damping coefficient is constrained, the parameter stable feasible region is determined by the short-circuit ratio, and the Pareto optimal solution of the virtual inertia and damping coefficient in the stable feasible region is determined by the optimization algorithm to achieve the optimal grid support effect. Specifically, the inertia power P j It is proportional to the virtual inertia J and the frequency change rate, and P j ≤P rated , so the virtual inertia J is limited by the rated power P rated Similarly, the damping power P d is proportional to the damping coefficient D, and P d ≤P rated The damping coefficient D is also limited by the rated power P rated Therefore, P rated The size of limits the maximum value of virtual inertia and damping coefficient. At the same time, the short circuit ratio is inversely proportional to the frequency change rate. When the short circuit ratio is high, the frequency stability is high, and the frequency change rate is small. When P rated When the short-circuit ratio is constant, the range of the damping inertia J becomes larger; on the contrary, when the short-circuit ratio is low, the frequency change rate becomes larger, which makes the range of the damping inertia J smaller. The specific optimization algorithm can use genetic algorithm or particle swarm optimization algorithm, with the objective function J being to minimize the weighted square sum of the system's frequency deviation Δf and voltage deviation ΔV. obj , optimize the virtual inertia and damping coefficient:

[0076] J obj =α(Δf) 2 +β(ΔV) 2 ;

[0077] Among them, α and β are weight coefficients, which are set according to the grid's emphasis on frequency and voltage stability. During the algorithm execution process, the values ​​of the virtual inertia and the damping coefficient are continuously adjusted to simulate different operating scenarios, calculate the corresponding objective function value, and find the optimal solution for the virtual inertia and the damping coefficient after multiple iterations.

[0078] Combined with the droop control strategy, the output power of the energy storage converter is dynamically adjusted according to the local measured frequency and voltage deviation. Assume that the frequency droop coefficient is m f , the voltage droop coefficient is m vWhen the grid frequency deviation is Δf and the voltage deviation is ΔV, the output power adjustment ΔP and reactive power adjustment ΔQ of the energy storage converter can be expressed as:

[0079] ΔP=-m f Δf;

[0080] ΔQ=-m v ΔV;

[0081] Different energy storage converters automatically share power changes in the absence of communication based on their own droop characteristics to achieve collaborative work; when the grid frequency drops, the energy storage converter automatically increases the output power; when the grid frequency rises, the energy storage converter automatically reduces the output power to maintain the frequency stability of the grid. The droop coefficient is configured differently according to the capacity C and / or performance of the energy storage converter to ensure the reasonable output of each device in collaborative work. In an embodiment of the present invention, for energy storage converters with larger capacity, relatively small frequency droop coefficients and voltage droop coefficients are set so that they bear the main responsibility in power regulation; and for energy storage converters with smaller capacity, relatively large frequency droop coefficients and voltage droop coefficients are set so that they play an auxiliary and fine-tuning role in power regulation.

[0082] In addition, the droop coefficient can also be calculated based on the remaining capacity C of the energy storage converter. remain and health status H are dynamically adjusted, for example, by using a linear or nonlinear function relationship:

[0083]

[0084] in, and are the initial frequency droop coefficient and voltage droop coefficient, k f 、k v 、l f 、l v is a preset adjustment coefficient, which can be determined through experiments or simulations according to actual needs, or selected based on experience. f 、k v When selecting the value, consider balancing the response speed and overshoot (for example, k f Can be adjusted from the upper limit of 10 to about 5), l f 、l v The value can be between 0.3 and 0.7 according to the signal noise level.

[0085] In step S3, an adaptive synchronization algorithm is designed. Based on local state estimation and the output of the virtual synchronizer (such as electric power and voltage / frequency control signals), each energy storage converter continuously adjusts its output voltage phase θ and frequency f to gradually synchronize with other energy storage converters. This algorithm utilizes the principle of phase-locked loop (PLL) technology, but has been improved to enable it to quickly and accurately track the system synchronization signal without communication. Assume that the synchronization error is e = θ - θ ref (where θ ref As the reference phase, a variable structure control strategy is used to dynamically adjust the control gain according to the magnitude of the synchronization error. In the early stages of synchronization, when the synchronization error is large, a larger control gain K1 (e.g., 10) is used to accelerate convergence. As the synchronization process progresses, the synchronization error gradually decreases. At this time, the control gain is reduced to improve synchronization accuracy and avoid over-adjustment. For example, if the synchronization error thresholds are set to E1 and E2, where E1>E2, for example, E1>5° and E2<1°, then:

[0086] When the synchronization error is greater than E1, the control gain is set to K1;

[0087] When the synchronization error is between E1 and E2, the control gain K is linearly proportional to Decrease, where K2 is a preset smaller control gain (e.g. 0.5);

[0088] When the synchronization error is less than E2, the control gain is set to K2.

[0089] In addition, the phase and frequency of the output voltage are dynamically adjusted in step S3, wherein:

[0090] The frequency adjustment formula is:

[0091]

[0092] Among them, K p is a preset proportional coefficient, determined according to the system dynamic performance requirements;

[0093] The phase adjustment formula is:

[0094]

[0095] Among them, K i is the integral coefficient, which is used to eliminate steady-state errors.

[0096] The above-mentioned adaptive synchronization algorithm introduces a dynamic adjustment mechanism, which automatically adjusts synchronization parameters such as synchronization gain and bandwidth according to the operating status and disturbance of the power grid to improve the accuracy and stability of synchronization. For example, when the power grid voltage fluctuates greatly, the synchronization bandwidth BW is appropriately increased to speed up the synchronization speed; when the power grid is operating stably, the synchronization bandwidth is reduced to improve the synchronization accuracy. By establishing a mapping relationship model between the power grid status and the synchronization parameters, the adaptive adjustment of the synchronization parameters is achieved by using intelligent methods such as fuzzy logic control or neural networks. In fuzzy logic control, let the power grid voltage fluctuation amplitude be V fluct , the frequency change rate df / dt, etc. are used as input variables. After fuzzification, rule inference, and defuzzification, the adjusted values ​​of the synchronization bandwidth and gain are output. For example, a triangular membership function is used to fuzzify the input variables and establish a fuzzy rule table. For example, if the voltage fluctuation is large and the frequency change rate is high, the bandwidth and gain are increased. The adjustment values ​​are obtained through fuzzy inference and defuzzification using the center of gravity method. For neural network methods, a multilayer perceptron (MLP) can be used, with the relevant operating parameters of the power grid as the input layer and the adjusted values ​​of the synchronization parameters as the output layer. The network is trained with a large amount of sample data, enabling it to accurately output appropriate synchronization parameter adjustment values ​​based on the input power grid status.

[0097] In step S4, each energy storage converter has a built-in fault detection module, which can quickly detect short circuit, overcurrent, and overvoltage faults by real-time monitoring of local electrical parameters and its own status. Once a fault is detected, the fault tolerance mechanism is immediately activated, and the fault information is notified to the operation and maintenance personnel through a local display or alarm device. At the same time, the control strategy is automatically adjusted to reduce the impact of the fault on the system. Among them, fault detection adopts a method based on a combination of threshold judgment and wavelet transform. First, a threshold is set to preliminarily determine whether there is a fault, and then the wavelet transform is used to extract and analyze the characteristics of the fault signal to improve the accuracy and reliability of fault detection.

[0098] Specifically, let the voltage threshold be V th , the current threshold is I th and the frequency threshold is f th , when the measured voltage u(t) satisfies |u(t)|>V th , or the current i(t) satisfies |i(t)|>I th , or the frequency f(t) satisfies |f(t)-f nom |>f th When f nom is the rated frequency, triggering the wavelet transform analysis; the wavelet transform uses the preset wavelet basis function to perform multi-scale decomposition on the fault signal s(t):

[0099]

[0100] Among them, c j,k is the wavelet coefficient, ψ j，k It is a wavelet function, and the fault type and severity are determined by analyzing the variation pattern and energy distribution of the wavelet coefficients.

[0101] In the event of a failure of some energy storage converters, other normal energy storage converters can automatically adjust their working modes, take on more power regulation tasks, maintain the stable operation of the power grid, and achieve the fault tolerance of the system. For example, a distributed coordinated control strategy is adopted to redistribute power regulation tasks according to the capacity and performance of the remaining normal equipment to ensure the stable operation of the power grid. In distributed coordinated control, let the set of remaining normal energy storage converters be S. For each energy storage converter i∈S, its new power regulation target is Calculated according to the following formula:

[0102]

[0103] in, is the original power regulation target, ΔP total is the power shortage caused by the fault, C i is the capacity of energy storage converter i; each energy storage converter exchanges information with its neighboring energy storage converters locally. Through iterative calculation, the power regulation targets of all energy storage converters are made consistent, thereby achieving reasonable power distribution and stable operation of the system.

[0104] The technical solution of the present invention realizes high-precision collaborative and autonomous synchronous networking of multi-network energy storage converters without communication interconnection, improves the stability and reliability of the power system, and reduces the problem of system performance degradation caused by communication failures or delays. In actual power grid applications, technologies such as virtual synchronous machine control and energy storage frequency regulation can reduce frequency fluctuations by 30%-50% and voltage fluctuations by 20%-30% in simulation and laboratory environments. In actual scenarios such as wind farms and charging stations, through intelligent control strategies and equipment optimization, the frequency fluctuations can be reduced by more than 20% and the voltage fluctuations can be reduced by 32%. The experimental data sources include authoritative simulation platforms, actual engineering tests and international power market reports, and have a high degree of credibility.

[0105] In addition, the adaptability of the energy storage system in complex environments and distributed energy access scenarios has been enhanced, and it can automatically respond to various disturbances and faults in the power grid, effectively improving the power quality and power supply reliability of the power grid. In areas with a high penetration rate of distributed energy, problems such as frequency instability, weak voltage support, and oscillation of multiple converters in parallel are prone to occur. The technical solution of the present invention can reduce frequency fluctuations, provide virtual inertial support, and adaptively adjust power output to ensure reasonable load sharing when multiple units are connected in parallel, significantly improving the region's new energy absorption capacity and reducing the phenomenon of wind and solar power abandonment. It reduces dependence on communication infrastructure, reduces the construction and maintenance costs of the system, and improves the overall economy and practicality of the system. Compared with traditional communication-based collaborative technologies, it can reduce the investment cost of communication equipment by about 50%, while also reducing maintenance costs caused by communication failures.

[0106] The embodiment of the present invention also provides a multi-network energy storage converter collaborative autonomous synchronous network system architecture without communication interconnection, please refer to Figure 2 The system includes multiple energy storage converters 1, 2, ..., n. Each energy storage converter is equipped with a control unit, a virtual synchronous machine module, a droop control module, an adaptive synchronization module, a fault detection module, and several sensors (such as current sensors, voltage sensors, and frequency sensors) for collecting local electrical parameters. The control unit is connected and interacted with the other parts mentioned above. Among them, each sensor transmits the collected electrical parameters to the control unit. The control unit cooperates with the other four modules to jointly implement the above-mentioned multi-network type energy storage converter collaborative autonomous synchronization networking method.

[0107] For the above system, it mainly includes the following three aspects:

[0108] (1) Sensor installation and data collection

[0109] Specifically, high-precision voltage, current, and frequency sensors are installed at the input and output of each energy storage converter to ensure measurement accuracy and stability. These sensors are connected to the converter's control unit and transmit collected data to the control unit in real time for processing. High-precision sensors are selected, and shielded wiring and filtering circuits are employed to minimize the impact of electromagnetic interference on measurement data. For example, voltage sensors with an accuracy of ±0.1%, current sensors with an accuracy of ±0.2%, and frequency sensors with an accuracy of ±0.01Hz are selected. Double-shielded wiring is also used to minimize external electromagnetic interference, and a Butterworth low-pass filter with a cutoff frequency of 5kHz is connected to the sensor output to further filter out high-frequency noise and ensure the accuracy of collected data.

[0110] Data collected by the sensors is preprocessed, including filtering, amplification, and analog-to-digital conversion, to improve data quality and usability. A Butterworth filter is used for filtering, with an appropriate cutoff frequency set to remove high-frequency noise. An amplifier is used to amplify the signal to an appropriate range for subsequent processing. A high-speed analog-to-digital converter is used to convert the analog signal into a digital signal, ensuring real-time data acquisition. For example, a voltage signal is filtered using a Butterworth filter with a cutoff frequency of 3 kHz, then amplified by an amplifier with a gain of 2. Finally, a 12-bit high-speed analog-to-digital converter is used to convert the signal into a digital signal. The conversion time is less than 5 μs, meeting real-time requirements.

[0111] (2) Algorithm implementation of the control unit

[0112] A Kalman filter-based state estimation algorithm is implemented in the control unit of the energy storage converter. This algorithm estimates the converter's operating state in real time based on sensor data. Simultaneously, a local dynamic model is established based on historical data and trends of local electrical parameters, and model parameters are continuously optimized using a machine learning algorithm. The program is written in Python or C, utilizing existing machine learning libraries (such as TensorFlow or PyTorch) to train and update the LSTM model, ensuring the accuracy of state estimation and prediction. When implementing the EKF algorithm, the functional form and parameter values ​​in the state and measurement equations are determined based on the specific circuit parameters and operating characteristics of the energy storage converter. For example, for a specific model of energy storage converter, the specific expressions for the state transfer function f(·) and the measurement function h(·) are determined through circuit analysis and experimental testing. Initial values ​​for the process noise covariance matrix Q(k) and the measurement noise covariance matrix R(k) are determined based on the sensor noise characteristics. These values ​​are then adjusted appropriately during operation based on actual conditions. For the training of the LSTM model, a large amount of operating data of the energy storage converter under different working conditions was collected and divided into training set, validation set and test set. The small batch gradient descent method was used for training, and the learning rate was set to 0.001 and the number of training rounds was 100. The model parameters were adjusted based on indicators such as the loss function value and accuracy of the validation set to ensure that the model has good performance on the test set.

[0113] Implement the virtual synchronous machine and droop control algorithms, appropriately setting parameters such as virtual inertia, damping coefficient, and droop coefficient based on system requirements and the performance of the energy storage converter. Programming implements an adaptive synchronization algorithm, dynamically adjusting synchronization parameters based on the grid's operating status and disturbances to ensure efficient algorithm operation. In actual programming, a modular design is adopted, encapsulating each algorithm module into independent functions for easy debugging and maintenance. For example, the virtual synchronous machine module is encapsulated as a function, taking as input the current state estimate and the set parameters such as virtual inertia and damping coefficient, and outputting the voltage and frequency reference values ​​processed by the virtual synchronous machine. The droop control module is encapsulated as a function, taking as input the frequency and voltage deviations and droop coefficient, and outputting the power adjustment. The adaptive synchronization module is encapsulated as a function, taking as input the current phase and frequency, the synchronization reference signal, and grid operating status parameters, and outputting the updated phase and frequency values. In the main program, these functions are called in a specific order to implement the entire control algorithm flow.

[0114] (3) Fault detection and fault tolerance mechanism implementation

[0115] Write a fault detection program in the control unit and set the thresholds and judgment conditions for faults such as short circuit, overcurrent, and overvoltage. When a fault is detected, the control unit immediately sends an alarm signal and starts the fault tolerance mechanism. The fault tolerance mechanism minimizes the impact of the fault on the system by adjusting the control strategy, such as reducing the output power, switching the working mode, etc. For example, when an overcurrent fault is detected, the output power of the energy storage converter is gradually reduced according to the degree of overcurrent. If the overcurrent situation continues to deteriorate, it switches to the backup working mode or shuts down for protection. Specifically, let the overcurrent multiple be I faulut , when 1.2<I faulut ≤1.5, the output power is gradually reduced by 5%; when I faulut When the current is greater than 1.5, the system switches to the standby working mode. In programming implementation, the current sensor data is monitored in real time to calculate the overcurrent multiple and the corresponding control logic code is written according to the above rules.

[0116] Establish a fault information sharing mechanism between energy storage converters. When some energy storage converters fail, other functioning converters can promptly receive fault information, automatically adjust their operating modes, and take on more power regulation tasks. Simple communication methods based on power line communication (PLC) or wireless ad hoc networking can be used to quickly transmit critical information when a fault occurs, enabling information sharing and coordinated response within a limited scope. For example, using PLC technology, signals of a specific frequency can be transmitted on power lines to convey fault information, including fault type and location. During normal operation, each energy storage converter periodically sends a heartbeat signal. When a converter detects a fault, it immediately stops sending heartbeat signals and transmits fault information. If other converters fail to receive heartbeat signals within a certain period of time, they determine that a fault may have occurred and begin receiving fault information and taking appropriate action.

[0117] The embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the steps of the aforementioned multi-network type energy storage converter collaborative autonomous synchronization networking method. Based on this understanding, the method / algorithm of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, USB flash drive, mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the steps of the method in each implementation scenario of the present invention.

[0118] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. Those skilled in the art will recognize that several equivalent substitutions or obvious variations can be made without departing from the scope of the present invention, and that such variations, with the same performance or use, should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for collaborative and autonomous synchronization of multi-network energy storage converters without communication interconnection, characterized in that: The following steps are involved: S1. State perception and estimation based on local measurements: sensors are used to collect local electrical parameters in real time and state estimation is performed using an improved Kalman filter algorithm and a local dynamic model. S2. Coordination mechanism between virtual synchronous machine and droop control: The virtual synchronous machine is combined with the droop control mechanism to achieve automatic power regulation and coordinated operation by setting virtual inertia, damping coefficient and droop coefficient; S3, Adaptive synchronization algorithm: Based on local state estimation and the output of the virtual synchronous machine, it adjusts the output voltage phase and frequency, and dynamically adjusts the synchronization parameters according to the grid state.

2. The method for collaborative and autonomous synchronization of multi-network energy storage converters according to claim 1, characterized in that: The following steps are also included: S4. Detect faults by real-time monitoring of local electrical parameters and its own status, and automatically adjust control strategies and achieve system fault tolerance when faults occur.

3. The method for collaborative and autonomous synchronization of multi-network energy storage converters according to claim 1, characterized in that: The improved Kalman filter algorithm described in step S1 adopts a method based on extended Kalman filtering, combined with a nonlinear model of the energy storage converter, to iteratively update the state variables to achieve the state estimation; the state equation of the improved Kalman filter algorithm is x(k+1)=f(x(k), u(k))+w(k), and the measurement equation is z(k)=h(x(k))+v(k); wherein, f(·) is a nonlinear state transfer function, u(k) is an input vector; w(k) is a process noise vector, which satisfies a Gaussian distribution with a mean of zero and a covariance matrix of Q(k); z(k) is a measurement vector, h(·) is a measurement function; v(k) is a measurement noise vector, which satisfies a Gaussian distribution with a mean of zero and a covariance matrix of R(k).

4. The method for collaborative and autonomous synchronization of multi-network energy storage converters according to claim 3, characterized in that: In the extended Kalman filter method, the prediction step is first performed: P - (k+1)=F(k)P(k)F(k) T +Q(k), in, is the estimated value of the predicted state, P - (k+1) is the predicted covariance matrix, F(k) is the Jacobian matrix of the state transfer function f(·) with respect to the state variable x(k); Then do the update steps: K(k+1)=P - (k+1)H(k+1) T [H(k+1)P - (k+1)H(k+1) T +R(k+1) T ] -1 , P(k+1)=[I-K(k+1)H(k+1)]P - (k+1), Where K(k+1) is the Kalman gain, H(k+1) is the Jacobian matrix of the measurement function h(·) with respect to the state variable x(k+1), is the updated state estimate, and P(k+1) is the updated covariance matrix.

5. The method for collaborative and autonomous synchronization of multi-network energy storage converters according to claim 1, characterized in that: Step S1 also includes: establishing the local dynamic model based on the change trend and historical data of local electrical parameters to predict changes in its own state, adjust the control strategy in advance, and improve the response speed and stability of the system; the local dynamic model is based on the long short-term memory network LSTM in machine learning, which can capture the time series characteristics of electrical parameters.

6. The method for collaborative and autonomous synchronization of multi-network energy storage converters according to claim 1, characterized in that: Step S2 includes: constraining the power output boundary of the virtual inertia and damping coefficient with the rated power of the energy storage converter, determining the parameter stable feasible region with the short-circuit ratio, using an optimization algorithm to determine the Pareto optimal solution of the virtual inertia and damping coefficient within the stable feasible region, and dynamically adjusting the output power of the energy storage converter according to the locally measured frequency and voltage deviations in combination with a droop control strategy; The optimization algorithm includes: using genetic algorithm or particle swarm optimization algorithm, with the objective function J being to minimize the weighted square sum of the system frequency deviation Δf and voltage deviation ΔV obj , optimize the virtual inertia and the damping coefficient: J obj =α(Δf) 2 +β(ΔV) 2 ; Among them, α and β are weight coefficients, which are set according to the grid's emphasis on frequency and voltage stability. During the algorithm execution process, the values ​​of the virtual inertia and the damping coefficient are continuously adjusted to simulate different operating scenarios, calculate the corresponding objective function value, and find the optimal solution for the virtual inertia and the damping coefficient after multiple iterations.

7. The method for collaborative and autonomous synchronization of multi-network energy storage converters according to claim 6, characterized in that: The step S2 of dynamically adjusting the output power of the energy storage converter includes: Assume the frequency droop coefficient is m f , the voltage droop coefficient is m v When the grid frequency deviation is Δf and the voltage deviation is ΔV, the output power adjustment ΔP and reactive power adjustment ΔQ of the energy storage converter are expressed as follows: ΔP=-m f Δf; ΔQ=-m v ΔV; Different energy storage converters automatically share power changes without communication based on their own droop characteristics, achieving collaborative work; when the grid frequency drops, the energy storage converter automatically increases output power; when the grid frequency rises, the energy storage converter automatically reduces output power to maintain grid frequency stability.

8. The method for collaborative and autonomous synchronization of multi-network energy storage converters according to claim 7, characterized in that: Step S2 also includes differentially configuring the droop coefficient according to the capacity C of the energy storage converter: for a larger capacity energy storage converter, a relatively smaller frequency droop coefficient and a relatively smaller voltage droop coefficient are set, so that the converter plays a major role in power regulation; whereas for a smaller capacity energy storage converter, a relatively larger frequency droop coefficient and a relatively larger voltage droop coefficient are set, so that the converter plays an auxiliary and fine-tuning role in power regulation; Or, according to the remaining capacity C of the energy storage converter remain The droop coefficient is dynamically adjusted based on the health status H, including the use of linear or nonlinear functional relationships: in, and are the initial frequency droop coefficient and voltage droop coefficient, k f 、k v 、l f 、l v is the preset adjustment factor.

9. The method for collaborative and autonomous synchronization of multi-network energy storage converters according to claim 8, characterized in that: Step S3 includes: setting the synchronization error to e = θ - θ ref , where θ, θ ref are the output voltage phase and reference phase of the energy storage converter respectively; a variable structure control strategy is adopted to dynamically adjust the control gain according to the size of the synchronization error. In the early stage of synchronization, when the synchronization error is large, a larger control gain K1 is adopted to speed up the convergence speed; as the synchronization process proceeds, the synchronization error gradually decreases. At this time, the control gain is reduced to improve the synchronization accuracy and avoid over-adjustment.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it can implement the steps of the multi-grid type energy storage converter collaborative autonomous synchronization networking method described in any one of claims 1-9.

Images