Low-voltage area multi-point voltage inertia support method and device

Abstract

The application discloses a low-voltage area multi-point voltage inertia support method and device, which comprises the following steps: S1: each node collects power grid operation parameters; S2: each node combines line impedance characteristic parameters, virtual inertia parameters and virtual damping parameters to form a multi-dimensional parameter vector, performs multi-parameter synchronous identification on the multi-dimensional parameter vector through a recursive identification algorithm, and outputs multi-parameter joint identification results of each node; S3: virtual inertia parameters in the multi-parameter joint identification results are used to calculate inertia representation of each node; S4: inertia support of each node and voltage support of each node are calculated, and after verification in combination with energy storage states of each node, the inertia support and the voltage support are sent to controllers of each node for execution. The application can realize low-voltage area multi-point voltage inertia support by means of multi-parameter joint identification and quantitative driving of inertia and voltage two-dimensional support distribution based on identification results.

Description

Technical Field

[0001] This invention relates to the field of power system inertia support technology, specifically to a method and apparatus for multi-point voltage inertia support in low-voltage distribution areas. Background Technology

[0002] With the continuous increase in photovoltaic penetration in low-voltage distribution substations and the rapid increase in power electronic interface equipment, the proportion of rotating inertia provided by synchronous generators is declining. Under scenarios of sudden load changes and photovoltaic output fluctuations, the frequency variation rate and voltage deviation amplitude in these substations are continuously increasing, making the need for online sensing and quantitative support of nodal inertia increasingly urgent. Especially in low-voltage, high-resistivity line environments, where the resistive component of line impedance is significant, traditional inertia identification and voltage support methods based on purely inductive line assumptions are difficult to apply. Substation operators and control systems need timely and reliable inertia characterization results and voltage support parameters to support stability control, fault analysis, and operational mode adjustments. Therefore, developing a more accurate and physically consistent joint inertia and voltage identification and quantitative support technology tailored to the characteristics of high-resistivity lines in low-voltage substations has become a practical requirement in the field of distribution network monitoring and control.

[0003] Traditional virtual inertia support typically employs a fixed-parameter virtual synchronous generator control method, where the virtual inertia constant and damping coefficient are determined during the design phase and remain unchanged. This method is relatively intuitive and effective in scenarios where the operating conditions of the photovoltaic distribution area do not change significantly. However, it often has significant limitations in actual high-penetration photovoltaic distribution areas experiencing disturbances. First, fixed-parameter schemes cannot perceive the actual inertia state of the distribution area, and there is no quantitative correspondence between the virtual inertia support and the actual inertia gap, making it impossible to achieve quantitative and accurate support for frequency fluctuations and voltage deviations. Second, although some adaptive virtual synchronous generator control schemes can adjust control parameters based on local signals such as frequency deviation or state of charge, they do not include an online identification process. The parameter adjustment is based on local signals rather than the overall inertia state of the distribution area, and there is still a lack of quantitative correlation between the support and actual needs. Third, online inertia identification schemes based on pendulum equations estimate the system inertia online through the sliding window least squares method. However, such schemes are mainly for the transmission network level, adopt the assumption of purely inductive lines, and do not make system deviation corrections for the high resistivity ratio characteristics of the resistive component in the line impedance of low-voltage distribution areas. Furthermore, the lack of multi-node collaborative identification and quantitative support allocation mechanisms makes it difficult to directly transfer to low-voltage distribution area scenarios.

[0004] In recent years, research on virtual synchronous generator control, dynamic frequency identification, and online inertia estimation has gradually shifted from fixed-parameter control to adaptive and identification-driven control. One type of research improves the adaptive adjustment capability under dynamic conditions by improving the virtual synchronous generator control structure and introducing frequency deviation or state-of-charge feedback. Another type of research improves the robustness of inertia identification results under noise and short-term disturbance conditions by optimizing sliding windows, filtering strategies, or noise reduction processes. A third type of research attempts to combine frequency identification results with power disturbance information for online estimation of system inertia levels. These studies have provided useful ideas for virtual inertia support in low-voltage distribution areas, but overall they are still mostly focused on improving control structures, optimizing local filtering, or improving the back-end identification process. There is a lack of a systematic solution that integrates line impedance characteristic parameters into the identification model, online evaluation and active compensation of identification excitation sufficiency, and quantitative driving of inertia and voltage dual-dimensional support allocation based on identification results into the same closed-loop framework. This results in line impedance deviations affecting identification accuracy through sequential autoregression, and the continuous identification excitation conditions during the stable operation period of the distribution area cannot be guaranteed. Inertia and voltage support also lack a quantitative closed loop driven by identification quantities.

[0005] Therefore, it is urgent to develop a new method and device for multi-point voltage inertia support in low-voltage distribution areas, and to establish a closed loop for quantitative allocation of inertia and voltage dual gaps, so as to provide a more reliable foundation for inertia support and voltage support for low-voltage distribution areas with a high proportion of photovoltaic access. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, one of the objectives of this invention is to provide a method for multi-point voltage inertia support in low-voltage transformer areas, which can achieve multi-point voltage inertia support in low-voltage transformer areas by identifying multiple parameters and quantitatively driving the allocation of inertia and voltage dual-dimensional support based on the identification results.

[0007] The second objective of this invention is to provide a multi-point voltage inertia support device for low-voltage distribution areas, which can achieve multi-point voltage inertia support by identifying multiple parameters and quantitatively driving the distribution of inertia and voltage support based on the identification results.

[0008] To achieve one of the objectives of this invention, the following solution is adopted: A method for multi-point voltage inertia support in low-voltage transformer areas includes the following steps: S1: Each node collects grid operating parameters, obtains the active power, reactive power, terminal voltage amplitude and angular frequency of each node based on the grid operating parameters, and calculates the angular acceleration based on the angular frequency; An identification regression vector is constructed based on the active power and the angular frequency, with the angular acceleration used as the output of the regression equation; S2: Using the regression vector, each node combines the line impedance characteristic parameters, virtual inertia parameters and virtual damping parameters to form a multi-dimensional parameter vector. The multi-parameter synchronous identification of the multi-dimensional parameter vector is performed by the recursive identification algorithm, and the multi-parameter joint identification results of each node are output. Each node exchanges data with the others, performs collaborative correction based on network topology, and updates the multi-parameter joint identification results of each node; Each node determines whether the current identification stimulus is sufficient based on the identification accuracy metric. If the stimulus is insufficient, it actively injects a perturbation signal to enhance the stimulus; otherwise, it enters a passive identification waiting mode. S3: Use the virtual inertia parameters in the multi-parameter joint identification results to calculate the inertia characterization of each node, and perform voltage support parameter identification to output the inertia characterization and voltage support characteristic parameters of each node. S4: Calculate the inertia gap and voltage gap based on the inertia characterization and voltage support characteristic parameters of each node; The inertia support amount of each node is calculated based on the inertia gap, and the voltage support amount of each node is calculated based on the voltage gap and the voltage support characteristic parameters. After verification in conjunction with the energy storage status of each node, the inertia support amount and voltage support amount are sent to the controller of each node for execution, thereby realizing multi-point voltage inertia support in the low voltage distribution area.

[0009] Further, step S1 includes the following sub-steps: S101: Extract the frequency of the i-th node of the three-phase power grid from the three-phase voltage and current signals collected by the inverters at each node using a phase-locked loop method. The phase angle is then used to calculate the active power at each node via the Park transformation. and reactive power And read the terminal voltage amplitude. Simultaneously, the angular frequency sequence of each node is calculated from the frequency. This enables the synchronous acquisition and extraction of power grid parameters at each node. Specifically, the formula for calculating the communication cycle is as follows: In the formula, The communication cycle is defined as the data exchange cycle between nodes. The sampling time interval of the ADC analog-to-digital converter; The window length of the Savitzky-Golay adaptive filter; S102: Transform the angular frequency sequence of each node Angular acceleration was calculated using the Savitzky-Golay polynomial smoothing filter method. And construct the augmented regression vector required for AE-JRRLS identification to obtain parameterized data for each node; Specifically, the augmented regression vector The calculation formula is as follows: In the formula, Each component corresponds to the mechanical power reference value, measured active power, and angular frequency deviation; The VSG power reference value is input to the pendulum equation; The three-phase active power is calculated by the Park transformation on the inverter output side. This is the difference between the current angular frequency and the rated angular frequency.

[0010] Further, step S2 includes the following sub-steps: S201: Measure the data at time t at each node: angular acceleration Mechanical power reference value Measure active power Angular frequency deviation By linearizing the augmented pendulum equation into three parameters, we define the augmented parameter vector, the relationship between its components, and the mathematical formula for multi-period joint identification. The calculation formula for the augmented pendulum equation is as follows: In the formula, Parameters characterizing the rotational inertia of a simulated synchronous generator in an energy storage inverter; This is the differential of the angular frequency with respect to time. The parameter to be identified is the ratio of the inductive component to the total impedance in the low-voltage distribution area's line impedance. ; Parameters characterizing the damping properties of a virtual synchronous generator; This is a reference value for mechanical power. To measure active power; This refers to the angular frequency deviation. Divide both sides of the calculation formula for the augmented pendulum equation by J, and let... , , This allows the original nonlinearly coupled three parameters to be... , , Transform into about The linear regression form is used to achieve three-parameter linearization; the calculation formula for the augmented parameter vector is as follows: In the formula, Let be the augmented parameter vector at time t, defined as a column vector containing three parameters to be identified; for , is defined as the reciprocal of the virtual inertia; for The ratio is defined as the ratio of the line impedance inductance to the virtual inertia. for The ratio is defined as the ratio of the virtual damping coefficient to the virtual inertia. in, Represented as the algorithm's response to the physical truth. The recursive estimate; subscript Indicates the first An independent estimator instance for each energy storage inverter node; subscript Indicates the first The recursive timing of each communication cycle; S202: The nodes of each... Periodic measurement data angular acceleration Augmented regression vector and the parameter estimate of the previous time step The identification residual at the current time is calculated using the joint regression residual formula; then the weighting coefficient is calculated using the Huber weighting function; finally, the three parameters are synchronously updated using the JRRLS gain matrix to obtain the node. Parameter estimates at current time ; Specifically, the formula for calculating the joint regression residuals is as follows: In the formula, For nodes No. The joint regression residuals of the period represent the identified residuals at the current time step, and are coupled with , , Three parameters; For nodes The joint estimation results of the three parameters in the previous communication cycle; The formula for calculating the Huber weight function is as follows: In the formula, For nodes No. The periodic Huber weight function value represents the dimensionless weight for reducing the weight of larger residuals; The Huber robustness threshold is a parameter that controls the degree of weight reduction. The formula for calculating the JRRLS gain vector is as follows: In the formula, For nodes No. The periodic JRRLS gain vector determines the step size for each update of the three parameters; For nodes The covariance matrix at the previous time step is a matrix that reflects the uncertainty of parameter estimation; Forgetting factor, a parameter that controls the decay of historical data weights; The calculation formula for the synchronous update of the three parameters is as follows: In the formula, For nodes No. The joint estimate of the three parameters of the period is the result of the three parameters being updated synchronously in the same formula.

[0011] Furthermore, step S2 also includes the following sub-steps: S203: Move the node No. Periodic JRRLS gain vector Augmented regression vector and nodes The covariance matrix of the previous time step Updated using the recursive formula for the covariance matrix This yields the identification accuracy measurement matrix for the current moment; Specifically, the recursive calculation formula for the covariance matrix is ​​as follows: In the formula, For nodes No. The periodic covariance matrix is ​​a three-dimensional symmetric positive definite matrix, reflecting the current accuracy status of the three-parameter identification; It is a three-dimensional identity matrix.

[0012] Furthermore, step S2 also includes the following sub-steps: S204: After the single-node recursive update is completed, each node exchanges data with the others, performs a topology Laplace regularization correction, and improves the overall identification accuracy by utilizing the spatial continuity of parameters of adjacent nodes to realize the line impedance-inductance ratio parameter of each node. Cooperative convergence; Specifically, the calculation formula for the topological Laplace regularization correction is as follows: In the formula, The gradient step size is defined as a parameter that controls the magnitude of the regularization correction. For the current cycle Extracted estimated value of the nodal line impedance-to-inductance ratio; These are the topological regularization coefficients; Topological weights are defined as nodes. With nodes The correlation weight between them; For nodes The set of neighboring nodes is defined as the set of nodes with the node The set of directly connected nodes in the network topology of the distribution area; the correction results are written back to each node. The corresponding component is used as the initial value for the next communication cycle; this correction is implemented in a distributed manner, with each node only needing to exchange with its neighboring nodes. Scalar values ​​do not require a central node for coordination.

[0013] Furthermore, the expression for the joint optimization objective is as follows: In the formula, For the first The true values ​​of the three parameters corresponding to each node; A robust loss function for applying Huber weighting to the residuals; is the topological regularity coefficient, which represents the parameter that controls the strength of the constraint on the consistency of parameters between adjacent nodes; The edge set of the network topology of the transformer area is defined as a set that describes the connection relationships between nodes; For nodes With nodes The correlation weights between nodes are learned adaptively online from the cross-correlation between nodes, without the need for a prior network admittance matrix; For nodes The line impedance-inductance ratio obtained from the identification is defined as... Extracted node-level parameters to be identified; the joint optimization objective is the mathematical basis for the multi-cycle iteration of steps S202 to S204: the first term of the AE-JRRLS recursive minimization in step S202 and the second term of the topological regularization minimization in step S204 are repeatedly iterated until convergence, which is to solve the expression of the joint optimization objective.

[0014] Furthermore, step S2 also includes the following sub-steps: S205: Move the node No. Periodic covariance matrix The inverse of the covariance matrix is ​​calculated through eigenvalue decomposition. The smallest eigenvalue and the corresponding minimum eigenvector This yields the weakest direction of the current identification accuracy in the three-dimensional parameter space; Specifically, the calculation formula for finding the minimum eigenvalue is as follows: In the formula, The smallest eigenvalue of the inverse covariance matrix is ​​defined as a real-time index reflecting the sufficiency of the identification algorithm for the directional excitation of each parameter. The corresponding minimum eigenvector is defined as a unit vector indicating the direction of the weakest parameter of the excitation; each node undergoes multiple iterations from steps S201 to S204 for parameter estimation. Gradually converged to That is, the expression of the joint optimization objective is used to jointly identify the solution of the optimization problem; S206: with For the execution cycle, find the smallest eigenvalue of the inverse covariance matrix. With the threshold of continuous incentive sufficiency By comparing the results, it is determined whether the current identification meets the continuous excitation condition of the three-parameter joint identification, and the excitation sufficiency judgment result is obtained. Specifically, the criteria for judging the sufficiency of incentives are as follows: If the above formula is true, then the active injection mode is triggered; in the formula, The threshold for sustained incentive sufficiency is defined as a threshold parameter for judging whether the current incentive is sufficient, and is determined based on the statistical lower bound of the historical distribution of the inverse eigenvalues ​​of the covariance matrix. S207: In active injection mode, the minimum feature vector By constrained optimization, the injected power signal is calculated to ensure that the injected signal is within the range of... Maximize the power component in the direction to obtain and execute the injected power command; Specifically, the calculation formula for the injection signal optimization is as follows: The above formula represents that in Find the condition that makes Take the maximum value In the formula, The amount of power injected for active excitation; For nodes The rated apparent power of the energy storage inverter is used to constrain the upper limit of single-node injection; The change in the augmented regression vector caused by power injection is defined as follows: ; S208: When judging When the system enters a passive identification and waiting mode, it continues to perform parameter updates in steps S201 to S204 by utilizing the changes in active power, reactive power, and terminal voltage triggered by natural disturbance events in the transformer area. No active injection is required, and the system waits for the next communication cycle to repeat steps S205 to S206 for judgment.

[0015] Furthermore, step S3 includes the following sub-steps: S301: After the joint identification from steps S201 to S204 converges through multiple iterations, take the nodes... First component Independent estimates of the virtual inertia of each node are obtained. ; to each node Based on the rated power of the energy storage inverter at this node, the inertia constant of each node is independently calculated. This yields the inertia identification results for each node in the current transformer area. Specifically, the formula for converting the nodal inertia constant is as follows: In the formula, For nodes The virtual inertia obtained from identification; For nodes The rated apparent power of the energy storage inverter is used as the reference for the per-unit conversion of the node's inertia constant. This is the system's rated angular frequency constant; S302: Each node executes the voltage identification module in parallel, utilizing the terminal voltage uploaded by each node's S1. and reactive power The data is used to identify the voltage sensitivity coefficient of each node through the voltage-reactive power sensitivity equation, thereby obtaining the voltage support parameters of each node in the current transformer area. ; Specifically, the calculation formula for the voltage-reactive power sensitivity is as follows: In the formula, For nodes The voltage-reactive power sensitivity coefficient is defined as a parameter that describes the magnitude of the terminal voltage change caused by the change in reactive power at a node. Voltage deviation is defined as node The difference between the terminal voltage and the reference voltage; The change in reactive power is defined as the node. The increase in reactive power.

[0016] Furthermore, step S4 includes the following sub-steps: S401: Set the inertia constant of each node and voltage support parameters of each node The inertia gap of each node is calculated independently, and the voltage gap is calculated simultaneously to obtain the amount of inertia and voltage support that the current transformer area needs to supplement. Specifically, the calculation formulas for the nodal inertia gap and voltage gap are as follows: In the formula, The overall target inertia constant of the distribution area is set for scheduling, which is the overall inertia level that the distribution area is expected to achieve. For nodes The target inertia constant is the value allocated from the total target of the distribution area to the nodes according to the rated capacity ratio. Target value; For nodes Inertia constant; For nodes The inertia gap is the difference between the target inertia and the identified inertia at that node. The rated terminal voltage is defined as the system's rated operating voltage. The average voltage measured in the current transformer area is the average voltage value of the transformer area collected in real time. Voltage gap is defined as the difference between the rated voltage and the measured voltage. S402: Perform a safety check on the energy storage status of each node: check the state of charge of each node. Does it meet the requirements? If node of If the node does not participate in inertia support during this communication cycle, its assigned value will be set accordingly. , The remaining nodes that meet the conditions proceed to S403 to calculate their respective support parameters. S403: Move the node Inertia gap The virtual inertia and damping coefficient support parameters are directly calculated based on the node's own rated power. Specifically, the calculation formula for the conversion of the supporting parameters is as follows: In the formula, For nodes The virtual inertia allocation value is the inertia support amount calculated from the node's own inertia gap. For nodes The rated apparent power of the energy storage inverter; This is the system's rated angular frequency constant; is the damping ratio tuning factor, and is the design parameter that controls the relationship between the damping ratio and the inertia ratio; For nodes The damping coefficient allocation value is the damping support parameter obtained by converting the inertia of the node according to the damping ratio. S404: Virtual inertia of each node parameter and damping coefficient The command is sent to the virtual synchronous generator controller of the energy storage inverter at each node for execution, and the quantitative command update of the inertia support parameter is completed for this communication cycle.

[0017] To achieve the second objective of this invention, the following solution is adopted: A multi-point voltage inertia support device for low-voltage transformer areas, used to perform a multi-point voltage inertia support method for low-voltage transformer areas as described in one of the objectives of this invention, comprising: The node controller is deployed on the nodes of the low-voltage distribution area and includes a data acquisition module, an AE-JRRLS three-parameter joint identification module, an inertia conversion and voltage identification module, an inertia gap calculation and support parameter conversion module, a virtual synchronous generator VSG controller, and an inverter power tube drive controller. The system coordination module is used to receive the total target inertia constant of the distribution area issued by the scheduling system, and broadcast the total target inertia constant to each node controller through the communication network module, without performing inertia identification or supporting parameter calculation. The communication network module is used for bidirectional measurement data exchange between each node controller and exchange of estimated values ​​of the line impedance-inductance ratio of adjacent nodes required for topology Laplace regularization correction, as well as communication by the system coordination module to each node controller to send parameters of the total target inertia constant of the broadcast station area. The data acquisition module is used to acquire three-phase voltage and current signals, extract grid frequency and phase angle through phase-locked loop, calculate active power, reactive power and terminal voltage amplitude through Park transform, calculate angular acceleration through Savitzky-Golay filtering, and construct augmented regression vector. The AE-JRRLS three-parameter joint identification module is used to combine the line impedance inductance ratio, virtual moment of inertia and virtual damping coefficient to form a three-dimensional augmented parameter vector. The three parameters are identified synchronously by recursive least squares weighted by Huber weight function. The minimum eigenvalue of the covariance matrix inverse is used to judge the excitation sufficiency and generate an active injection signal when the excitation is insufficient. In the multi-node scenario, the estimated line impedance inductance ratio is exchanged with the adjacent nodes to perform topological Laplace regularization correction. The inertia conversion and voltage identification module is used to convert the virtual moment of inertia and inertia constant of this node from the first component of the identified parameter vector, and to perform voltage-reactive sensitivity identification in parallel and output the voltage-reactive sensitivity coefficient of this node. The inertia gap calculation and support parameter conversion module is used to independently calculate the target inertia constant of this node based on the total target inertia constant of the station area broadcast by the communication network and the rated capacity ratio of this node, and then calculate the inertia gap of this node. After verifying the charging state limit of this node, the virtual moment of inertia allocation value and damping coefficient allocation value are directly converted from the inertia gap of this node itself. The virtual synchronous generator (VSG) controller is used to receive support parameters, execute the rotor motion equation of the virtual synchronous generator, calculate the active power and reactive power reference values, and generate a three-phase voltage reference signal through voltage-current dual closed-loop control. The inverter power transistor drive controller is used to generate six PWM switching signals through space vector pulse width modulation (SVPWM) based on the three-phase voltage reference signal output by the VSG controller. These signals are then used to directly control the on / off state of the power switching transistors of each bridge arm of the energy storage inverter via an isolation drive circuit, thereby converting the DC bus voltage into three-phase AC power output.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention combines line impedance characteristic parameters, virtual inertia parameters, and virtual damping parameters to form a multi-dimensional parameter vector. It achieves synchronous identification of multiple parameters through a recursive identification algorithm, avoiding the propagation of line impedance deviation to the virtual inertia estimation result through sequential autoregression. All parameters converge synchronously and correct each other in a unified recursive framework, thereby improving the accuracy of inertia identification in low-voltage, high-resistivity transformer areas.

[0019] 2. This invention enables data exchange between nodes to perform network topology-based collaborative correction. It utilizes the spatial continuity of parameters of adjacent nodes to correct the identification results of each node, reducing the overall spatial identification variance without the need for a central node for coordination. Simultaneously, it uses an identification accuracy metric to determine in real time whether the current identification stimulus is sufficient. If the stimulus is insufficient, it actively injects a disturbance signal to enhance the stimulus, ensuring that the three-parameter joint identification meets the continuous stimulus conditions during the stable operation period of the distribution area, thus eliminating the blind spot in inertia state identification.

[0020] 3. This invention achieves a complete quantitative closed loop for the allocation of inertia and voltage support parameters by having each node perform inertia parameter identification and voltage support parameter identification in parallel, and simultaneously calculates inertia gap and voltage gap. The inertia support is directly and quantitatively converted from the identified inertia gap, and the voltage support is quantitatively converted from the voltage gap and the identified voltage support characteristic parameters. This ensures that the frequency stability and voltage deviation of the transformer area meet the standard requirements. At the same time, each node calculates and executes independently without the need for a central node for coordination, which facilitates distributed deployment and expansion.

[0021] 4. This invention adopts a distributed architecture, in which each node independently collects data and independently performs identification and support calculations. Collaborative correction is achieved only through data exchange between nodes, eliminating the need for a central node for unified coordination. This avoids the risk of single point of failure, reduces communication bandwidth requirements and system construction costs, and allows for flexible addition or removal of nodes in the distribution area according to actual needs, exhibiting good scalability.

[0022] 5. This invention independently acquires local three-phase voltage and current signals at each node, eliminating the need for a global synchronization phasor measurement device for the entire distribution area. This reduces system hardware deployment costs and dependence on high-precision synchronization clocks, facilitating engineering upgrades and widespread application in existing low-voltage distribution areas.

[0023] 6. By integrating identification and support allocation into the same framework, the identification results directly drive the quantitative calculation of support quantity, avoiding the information loss and delay accumulation caused by the separation of multiple links of "identification-decision-execution" in traditional schemes. This invention achieves a rapid closed-loop response from parameter perception to control command generation, improving the dynamic support timeliness of the transformer area under disturbance scenarios.

[0024] 7. This invention continuously updates the inertia characterization parameters and voltage support characteristic parameters of each node through online recursive identification, enabling the support parameters to adaptively adjust with changes in the operating status of the transformer area (such as changes in photovoltaic power output, load fluctuations, network topology adjustments, etc.), overcoming the insufficient adaptability of fixed parameter control schemes under complex operating conditions, and realizing real-time tracking and dynamic matching of transformer area inertia and voltage status.

[0025] 8. This invention actively creates identification excitation conditions during the stable operation period of the transformer area by actively injecting disturbance signals, enabling the recursive identification algorithm to continuously converge. This avoids the drawback of traditional passive identification schemes that cannot update parameters during periods without disturbance, and ensures the timeliness and availability of inertia identification results in various operating scenarios. Attached Figure Description

[0026] Figure 1 This is a flowchart of the multi-point voltage inertia support method for low-voltage substations in an embodiment of the present invention; Figure 2 This is a schematic diagram of the installation of the multi-point voltage inertia support device in the low-voltage substation area in an embodiment of the present invention; Figure 3 This is a schematic diagram of the internal structure of the multi-point voltage inertia support device in the low-voltage substation area in an embodiment of the present invention. Detailed Implementation

[0027] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0028] Example 1 like Figure 1 As shown, this embodiment of the invention provides a method for multi-point voltage inertia support in low-voltage distribution areas, comprising the following steps: S1: Each node collects grid operating parameters, obtains the active power, reactive power, terminal voltage amplitude and angular frequency of each node based on the grid operating parameters, and calculates the angular acceleration based on the angular frequency; constructs an identification regression vector based on the active power and the angular frequency, and uses the angular acceleration as the output of the regression equation.

[0029] S2: Using the regression vector, each node combines the line impedance characteristic parameters, virtual inertia parameters, and virtual damping parameters to form a multi-dimensional parameter vector. The multi-parameter synchronous identification is performed on the multi-dimensional parameter vector through a recursive identification algorithm, and the multi-parameter joint identification results of each node are output. Each node exchanges data with each other, performs collaborative correction based on network topology, and updates the multi-parameter joint identification results of each node. Each node judges whether the current identification excitation is sufficient according to the identification accuracy metric. If the excitation is insufficient, it actively injects a disturbance signal to enhance the excitation; otherwise, it enters a passive identification waiting mode.

[0030] S3: Use the virtual inertia parameters in the multi-parameter joint identification results to calculate the inertia characterization of each node, and perform voltage support parameter identification to output the inertia characterization and voltage support characteristic parameters of each node.

[0031] S4: Based on the inertia characterization and voltage support characteristic parameters of each node, calculate the inertia gap and voltage gap respectively; calculate the inertia support of each node according to the inertia gap, calculate the voltage support of each node according to the voltage gap and the voltage support characteristic parameters, and verify the inertia support and voltage support of each node in conjunction with the energy storage status of each node. Then, send the inertia support and voltage support to the controller of each node for execution to realize multi-point voltage inertia support in the low voltage distribution area.

[0032] The method for multi-point voltage inertia support in low-voltage distribution areas according to embodiments of the present invention will be further described in detail below.

[0033] The low-voltage transformer substation multi-point voltage inertia support method of this invention includes the following: Each node's energy storage inverter collects three-phase voltage and current signals, and uses a phase-locked loop method to extract the grid frequency and phase angle. Then, the active power, reactive power, and terminal voltage amplitude of each node are calculated through Park transform, and the angular acceleration is calculated using the Savitzky-Golay adaptive filtering method on the original angular frequency signal. Finally, an augmented regression vector of the pendulum equation is constructed.

[0034] Using multi-point measurement data collected by each node, each node combines the line impedance-to-inductance ratio, the equivalent virtual moment of inertia of the transformer area, and the virtual damping coefficient to form a three-dimensional augmented parameter vector. The three parameters are then independently identified synchronously using the AE-JRRLS (Augmented Excitation – Jointly Robust Recursive Least Squares) algorithm. After the single-node AE-JRRLS algorithm identification and update is completed, each node exchanges data and performs a topological Laplace regularization correction to improve the identification accuracy of each node. Each node updates the minimum eigenvalue of its own covariance matrix inverse. If the minimum eigenvalue is lower than the continuous excitation sufficiency threshold, the node autonomously generates an active injection signal with the maximum power component in the weakest excitation direction. After injection, the covariance matrix is ​​updated to form a closed feedback within the algorithm. Otherwise, the node enters a passive identification waiting mode and continues to update the identification parameters using the natural disturbances of the transformer area.

[0035] The identified moment of inertia is used to calculate the inertia constant of each node, and the voltage-reactive power sensitivity identification is performed in parallel to output the inertia constant and voltage-reactive power sensitivity coefficient of each node.

[0036] Based on the obtained inertia identification results and voltage-reactive power sensitivity coefficient of each node, the inertia gap and voltage gap are calculated independently for each node. After verifying the under-limit protection conditions of each node under the charged state, the virtual moment of inertia and damping coefficient support parameters are directly converted from the inertia gap of each node. The command is sent to the virtual synchronous generator controller of each node for execution, so as to realize the multi-point voltage inertia support of the low voltage distribution area.

[0037] Further, step S1 includes the following sub-steps: S101: Extract the frequency of the i-th node of the three-phase power grid from the three-phase voltage and current signals collected by the inverters at each node using a phase-locked loop method. The phase angle is then used to calculate the active power at each node via the Park transformation. and reactive power And read the terminal voltage amplitude. Simultaneously, the angular frequency sequence of each node is calculated from the frequency. This enables the synchronous acquisition and extraction of power grid parameters at each node.

[0038] Specifically, the formula for calculating the communication cycle is as follows: In the formula, The communication cycle is defined as the data exchange cycle between nodes. The sampling time interval of the ADC analog-to-digital converter; This is the window length of the Savitzky-Golay adaptive filter.

[0039] S102: Convert the node angular frequency sequence obtained in S101. Angular acceleration was calculated using the Savitzky-Golay polynomial smoothing filter method. We construct AE-JRRLS (Augmented Excitation – Jointly RobustRecursive Least Squares) to identify the required augmented regression vectors and obtain parameterized data for each node.

[0040] Specifically, the augmented regression vector The calculation formula is as follows: In the formula, Each component corresponds to the mechanical power reference value, measured active power, and angular frequency deviation; The VSG power reference value is input to the pendulum equation; The three-phase active power is calculated by the Park transformation on the inverter output side. This is the difference between the current angular frequency and the rated angular frequency.

[0041] Further, step S2 includes the following sub-steps: S201: Measure the angular acceleration at time t at each node. Mechanical power reference value Measure active power Angular frequency deviation By linearizing the augmented pendulum equation into three parameters, we define the augmented parameter vector, the relationship between its components, and the mathematical formula for multi-period joint identification.

[0042] The calculation formula for the augmented pendulum equation is as follows: In the formula, Parameters characterizing the rotational inertia of a simulated synchronous generator in an energy storage inverter; This is the differential of the angular frequency with respect to time. The parameter to be identified is the ratio of the inductive component to the total impedance in the low-voltage distribution area's line impedance. ; Parameters characterizing the damping properties of a virtual synchronous generator; This is a reference value for mechanical power. To measure active power; This represents the angular frequency deviation.

[0043] Divide both sides of formula (3) by J, and let , , This allows the original nonlinearly coupled three parameters to be... , , Transform into about The linear regression form is used to achieve three-parameter linearization; the calculation formula for the augmented parameter vector is as follows: In the formula, Let be the augmented parameter vector at time t, defined as a column vector containing three parameters to be identified; for , is defined as the reciprocal of the virtual inertia; for The ratio is defined as the ratio of the line impedance inductance to the virtual inertia. for The ratio is defined as the ratio of the virtual damping coefficient to the virtual inertia.

[0044] in, Represented as the algorithm's response to the physical truth. The recursive estimate; subscript Indicates the first An independent estimator instance for each energy storage inverter node; subscript Indicates the first The recursive timing of each communication cycle.

[0045] The expression for the joint optimization objective is as follows: In the formula, For the first The true values ​​of the three parameters corresponding to each node; A robust loss function for applying Huber weighting to the residuals; For nodes No. Periodic angular acceleration; For nodes No. The periodic augmented regression vector; is the topological regularity coefficient, which represents the parameter that controls the strength of the constraint on the consistency of parameters between adjacent nodes; The edge set of the network topology of the transformer area is defined as a set that describes the connection relationships between nodes; For nodes With nodes The correlation weights between nodes are learned adaptively online from the cross-correlation between nodes, without the need for a prior network admittance matrix; For nodes The line impedance-inductance ratio obtained from the identification is defined as... Extracted node-level parameters to be identified; the joint optimization objective defined by formula (11) is the mathematical basis for subsequent multi-cycle iterations from S202 to S204: the first term of the AE-JRRLS recursive minimization of S202 and the second term of the topological regularization minimization of S204 are repeatedly iterated until convergence, i.e., formula (11) is solved.

[0046] S202: The nodes of each... Period measurement data (angular acceleration) Augmented regression vector ) and the parameter estimate of the previous time step The identification residual at the current time is calculated using the joint regression residual formula; then the weighting coefficient is calculated using the Huber weighting function; finally, the three parameters are synchronously updated using the JRRLS gain matrix to obtain the node. Parameter estimates at current time .

[0047] Specifically, the formula for calculating the joint regression residuals is as follows: In the formula, For nodes No. The joint regression residuals of the period represent the identified residuals at the current time step, and are coupled with , , Three parameters; For nodes The joint estimation results of the three parameters in the previous communication cycle.

[0048] The formula for calculating the Huber weight function is as follows: In the formula, For nodes No. The periodic Huber weight function value represents the dimensionless weight for reducing the weight of larger residuals; is the Huber robust threshold, a parameter that controls the degree of weighting reduction.

[0049] The formula for calculating the JRRLS gain matrix is ​​as follows: In the formula, For nodes No. The periodic JRRLS gain vector determines the step size for each update of the three parameters; For nodes The covariance matrix at the previous time step is a matrix that reflects the uncertainty of parameter estimation; The forgetting factor is a parameter that controls the decay of historical data weights.

[0050] The calculation formula for the synchronous update of the three parameters is as follows: In the formula, For nodes No. The joint estimate of the three parameters of the period is the result of the three parameters being updated synchronously in the same formula.

[0051] S203: Obtain the node from S202 Gain matrix Augmented regression vector covariance matrix at the previous time step Updated using the recursive formula for the covariance matrix This yields the identification accuracy measurement matrix for the current moment.

[0052] Specifically, the recursive calculation formula for the covariance matrix is ​​as follows: In the formula, For nodes No. The periodic covariance matrix is ​​a three-dimensional symmetric positive definite matrix, reflecting the current accuracy status of the three-parameter identification; It is a three-dimensional identity matrix.

[0053] S204: After the single-node AE-JRRLS recursive update of S203 is completed, each node exchanges data and performs a topology Laplace regularization correction. This utilizes the spatial continuity of parameters between adjacent nodes to improve overall identification accuracy and realize the line impedance-to-inductance ratio parameter of each node. Cooperative convergence.

[0054] Specifically, the calculation formula for the topological Laplace regularization correction is as follows: In the formula, The gradient step size is defined as a parameter that controls the magnitude of the regularization correction. For the current cycle Extracted estimated value of the nodal line impedance-to-inductance ratio; These are the topological regularization coefficients; Topological weights are defined as nodes. With nodes The correlation weight between them; For nodes The set of neighboring nodes is defined as the set of nodes with the node The set of directly connected nodes in the network topology of the distribution area; the correction results are written back to each node. The corresponding component is used as the initial value for the next communication cycle S202 recursion; this correction is implemented in a distributed manner, and each node only needs to exchange with its neighboring nodes. Scalar values ​​do not require a central node for coordination.

[0055] S205: Obtain the node from S203 covariance matrix The inverse of the covariance matrix is ​​calculated through eigenvalue decomposition. The smallest eigenvalue and the corresponding minimum eigenvector This yields the weakest direction in the three-dimensional parameter space for the current identification accuracy.

[0056] Specifically, the calculation formula for finding the minimum eigenvalue is as follows: In the formula, The smallest eigenvalue of the inverse covariance matrix is ​​defined as a real-time index reflecting the sufficiency of the identification algorithm for the directional excitation of each parameter. The corresponding minimum eigenvector is defined as a unit vector indicating the direction of the weakest parameter of the excitation. Each node undergoes multiple iterations from S201 to S204 for parameter estimation. Gradually converged to That is, formula (11) is used to jointly identify the solution to the optimization problem.

[0057] S206: with For the execution cycle, the result obtained in S205 With respect to the threshold of sustained incentive sufficiency By comparing the results, it is determined whether the current identification meets the continuous excitation condition of the three-parameter joint identification, and the excitation sufficiency judgment result is obtained.

[0058] Specifically, the criteria for judging the sufficiency of incentives are as follows: If the above formula is true, then the active injection mode is triggered; in the formula, The continuous incentive sufficiency threshold is defined as a threshold parameter for judging whether the current incentive is sufficient, and is determined based on the statistical lower bound of the historical distribution of the inverse eigenvalues ​​of the covariance matrix.

[0059] S207: In active injection mode, the minimum feature vector obtained in S205 is... By constrained optimization, the injected power signal is calculated to ensure that the injected signal is within the range of... Maximize the power component in the direction to obtain and execute the injected power command.

[0060] Specifically, the calculation formula for the injection signal optimization is as follows: The above formula represents that in Find the condition that makes Take the maximum value In the formula, The amount of power injected for active excitation; For nodes The rated apparent power of the energy storage inverter is used to constrain the upper limit of single-node injection; The change in the augmented regression vector caused by power injection is defined as follows: .

[0061] S208: When S206 makes a judgment When the system enters a passive identification and waiting mode, it continues to perform parameter updates from S201 to S204 by utilizing the changes in active power, reactive power, and terminal voltage triggered by natural disturbance events in the transformer area. No active injection is required, and the system waits for the next communication cycle to repeat the judgment from S205 to S206.

[0062] Furthermore, step S3 includes the following sub-steps: S301: After the joint identification from S201 to S204 converges through multiple cycles of iteration, each node is selected. First component Independent estimates of the virtual inertia of each node are obtained. . Each node Based on the rated power of the energy storage inverter at this node, the inertia constant of each node is independently calculated. This yields the inertia identification results for each node in the current transformer area.

[0063] Specifically, the formula for converting the nodal inertia constant is as follows: In the formula, For nodes The virtual inertia obtained from identification; For nodes The rated apparent power of the energy storage inverter is used as the reference for the per-unit conversion of the node's inertia constant. This is the system's rated angular frequency constant.

[0064] S302: Each node executes the voltage identification module in parallel, utilizing the terminal voltage uploaded by each node's S1. and reactive power The data is used to identify the voltage sensitivity coefficient of each node through the voltage-reactive power sensitivity equation, thereby obtaining the voltage support parameters of each node in the current transformer area. .

[0065] Specifically, the calculation formula for the voltage-reactive power sensitivity is as follows: In the formula, For nodes The voltage-reactive power sensitivity coefficient is defined as a parameter that describes the magnitude of the terminal voltage change caused by the change in reactive power at a node. Voltage deviation is defined as node The difference between the terminal voltage and the reference voltage; The change in reactive power is defined as the node. The increase in reactive power.

[0066] Furthermore, step S4 includes the following sub-steps: S401: The node inertia identification results obtained from S301 Voltage identification results obtained from S302 The inertia gap of each node is calculated independently, and the voltage gap is calculated simultaneously to obtain the amount of inertia and voltage support that the current transformer area needs to supplement.

[0067] Specifically, the calculation formulas for the nodal inertia gap and voltage gap are as follows: In the formula, The overall target inertia constant of the distribution area is set for scheduling, which is the overall inertia level that the distribution area is expected to achieve. For nodes The target inertia constant is the value allocated from the total target of the distribution area to the nodes according to the rated capacity ratio. Target value; The nodes identified for S301 Inertia constant; For nodes The inertia gap is the difference between the target inertia and the identified inertia at that node. The rated terminal voltage is defined as the system's rated operating voltage. The average voltage measured in the current transformer area is the average voltage value of the transformer area collected in real time. The voltage gap is defined as the difference between the rated voltage and the measured voltage.

[0068] S402: Perform a safety check on the energy storage status of each node: check the state of charge of each node. Does it meet the requirements? If node of If the node does not participate in inertia support during this communication cycle, its assigned value will be set accordingly. , The remaining nodes that meet the conditions will proceed to S403 to calculate their respective support parameters.

[0069] S403: The inertia gaps at each node calculated in S401. The virtual inertia and damping coefficient support parameters are directly calculated based on the node's own rated power.

[0070] Specifically, the calculation formula for the conversion of the supporting parameters is as follows: In the formula, For nodes The virtual inertia allocation value is the inertia support amount calculated from the node's own inertia gap. The nodes calculated for S401 Inertia gap; For nodes The rated apparent power of the energy storage inverter; This is the system's rated angular frequency constant; is the damping ratio tuning factor, and is the design parameter that controls the relationship between the damping ratio and the inertia ratio; For nodes The damping coefficient allocation value is the damping support parameter obtained by converting the inertia of the node according to the damping ratio.

[0071] S404: Calculate the node parameters (virtual inertia) obtained from S403. Damping coefficient The command is sent to the virtual synchronous generator controller of the energy storage inverter at each node for execution, thus completing the quantitative command update of the inertia support parameters for this communication cycle.

[0072] In the low-voltage distribution area multi-point voltage inertia support method of this invention, the energy storage inverter at each node collects three-phase voltage and current, and extracts frequency, power, and terminal voltage amplitude. Each node internally generates the line impedance-inductance ratio as the parameter to be identified, and achieves synchronous identification of the three parameters through the AE-JRRLS (Augmented Excitation–Jointly Robust Recursive Least Squares) algorithm, outputting the voltage-reactive power sensitivity coefficient of each node in parallel. The sufficiency of identification excitation is judged by the minimum eigenvalue of the inverse covariance matrix; if insufficient, an active injection signal is sent to form a closed feedback within the algorithm. The inertia gap and voltage gap are calculated in parallel, and the virtual inertia and reactive power of each node are quantitatively allocated according to the state-of-charge weighted ratio. Commands are sent to the virtual synchronous generator controller of each node for execution. The low-voltage distribution area multi-point voltage inertia support method of this invention eliminates the blind zone of inertia identification during the stable operation period, and realizes a complete closed-loop control that quantitatively drives the allocation of inertia and voltage dual-dimensional support based on online identification results.

[0073] The low-voltage transformer substation multi-point voltage inertia support method of this invention has the following advantages: 1. This invention improves the accuracy of inertia identification in low-voltage, high-resistivity transformer areas by intrinsically incorporating the line impedance-inductance ratio as the parameter to be identified, thus eliminating sequential autoregressive bias. Specifically, the AE-JRRLS algorithm of this invention combines the line impedance-inductance ratio with the virtual inertia and damping coefficient to form a three-dimensional augmented parameter vector. The Huber weighting mechanism simultaneously applies to the three-parameter update through the same gain matrix. Line impedance deviation correction and robust noise reduction are mathematically inseparable in the same regression residual. The line impedance-inductance ratio deviation no longer propagates to the virtual inertia estimation result through sequential autoregression, and the three parameters converge synchronously and correct each other. Simultaneously, the identification byproduct, the line impedance-inductance ratio, directly outputs the line impedance-inductance ratio, achieving online estimation of both inertia and line parameters in a single identification process.

[0074] 2. This invention introduces topological Laplace regularization constraints and closed-loop feedback based on the minimum eigenvalue of the covariance matrix to eliminate spatial identification errors, ensure continuous identification excitation around the clock, and eliminate inertia state blind spots. Specifically, the topological Laplace regularization term introduces the network topology spatial constraints into the kernel of the identification objective function, constraining the line impedance-inductance ratio parameters of adjacent nodes to maintain spatial continuity physically, reducing the overall spatial identification variance; the topology weights are adaptively learned online by the cross-correlation of measurement data between nodes, without requiring a priori line admittance matrices. The active excitation driven by the minimum eigenvalue of the covariance matrix ensures that the joint identification of the three parameters also meets the continuous excitation condition during the stable operation period of the distribution area, and the system inertia state is effective in real time, eliminating identification blind spots.

[0075] 3. This invention achieves a complete closed loop by quantitatively driving the allocation of inertia and voltage gaps based on the identification results, thus realizing a dual-dimensional support allocation driven by the identification results. Specifically, in this invention, each node runs inertia and voltage identification in parallel, synchronously calculating the inertia and voltage gaps. The virtual inertia allocation of each node is directly determined by the inertia constant obtained from the identification, which can verify whether the total injected inertia of the transformer area fills the inertia gap. The reactive power allocation of each node is quantitatively tuned by the voltage-reactive power sensitivity coefficient obtained from the identification, which can ensure that the voltage deviation at the end of the transformer area meets the low-voltage power supply voltage deviation standard requirements. The two quantitative closed loops share the measurement dataset, requiring no additional computational overhead and no need to set up a central node for coordination.

[0076] Example 2 like Figure 2 and Figure 3 As shown, this embodiment of the invention also provides a multi-point voltage inertia support device for low-voltage distribution areas, used to perform the multi-point voltage inertia support method for low-voltage distribution areas as described in Embodiment 1, including: The node controller is deployed on the nodes of the low-voltage distribution area and includes a data acquisition module, an AE-JRRLS three-parameter joint identification module, an inertia conversion and voltage identification module, an inertia gap calculation and support parameter conversion module, a virtual synchronous generator (VSG) controller, and an inverter power transistor drive controller.

[0077] The system coordination module receives the total target inertia constant of the distribution area from the scheduling system, broadcasts the total target inertia constant to each node controller through the communication network module, and does not perform inertia identification or support parameter calculation.

[0078] The communication network module is used for bidirectional measurement data exchange between each node controller and exchange of estimated values ​​of the line impedance-inductance ratio of adjacent nodes required for topology Laplace canonical correction, as well as for the system coordination module to send the parameter of the total target inertia constant of the broadcast area to each node controller.

[0079] The data acquisition module is used to acquire three-phase voltage and current signals, extract grid frequency and phase angle through phase-locked loop, calculate active power, reactive power and terminal voltage amplitude through Park transform, calculate angular acceleration through Savitzky-Golay filtering, and construct augmented regression vector.

[0080] The AE-JRRLS three-parameter joint identification module is used to combine the line impedance-inductance ratio, virtual moment of inertia, and virtual damping coefficient to form a three-dimensional augmented parameter vector. The three parameters are identified synchronously through recursive least squares weighted by the Huber weight function. The minimum eigenvalue of the covariance matrix inverse is used to determine the excitation sufficiency and generate an active injection signal when the excitation is insufficient. In multi-node scenarios, the estimated line impedance-inductance ratio is exchanged with adjacent nodes to perform topological Laplace regularization correction.

[0081] The inertia conversion and voltage identification module is used to convert the virtual moment of inertia and inertia constant of this node from the first component of the identified parameter vector, and to perform voltage-reactive power sensitivity identification in parallel and output the voltage-reactive power sensitivity coefficient of this node.

[0082] The inertia gap calculation and support parameter conversion module is used to independently calculate the target inertia constant of the node based on the total target inertia constant of the station area broadcast by the communication network and the rated capacity ratio of the node, and then calculate the inertia gap of the node. After verifying the charging state limit of the node, the virtual moment of inertia allocation value and damping coefficient allocation value are directly converted from the inertia gap of the node itself.

[0083] The virtual synchronous generator (VSG) controller is used to receive support parameters, execute the rotor motion equation of the virtual synchronous generator, calculate the active power and reactive power reference values, and generate a three-phase voltage reference signal through voltage-current dual closed-loop control.

[0084] The inverter power transistor drive controller is used to generate six PWM switching signals through space vector pulse width modulation (SVPWM) based on the three-phase voltage reference signal output by the VSG controller. These signals are then used to directly control the on / off state of the power switching transistors of each bridge arm of the energy storage inverter via an isolation drive circuit, thereby converting the DC bus voltage into three-phase AC power output.

[0085] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for multi-point voltage inertia support in a low-voltage distribution area, characterized in that, Includes the following steps: S1: Each node collects grid operating parameters, obtains the active power, reactive power, terminal voltage amplitude and angular frequency of each node based on the grid operating parameters, and calculates the angular acceleration based on the angular frequency; An identification regression vector is constructed based on the active power and the angular frequency, with the angular acceleration used as the output of the regression equation; S2: Using the regression vector, each node combines the line impedance characteristic parameters, virtual inertia parameters and virtual damping parameters to form a multi-dimensional parameter vector. The multi-parameter synchronous identification of the multi-dimensional parameter vector is performed by the recursive identification algorithm, and the multi-parameter joint identification results of each node are output. Each node exchanges data with the others, performs collaborative correction based on network topology, and updates the multi-parameter joint identification results of each node; Each node determines whether the current identification stimulus is sufficient based on the identification accuracy metric. If the stimulus is insufficient, it actively injects a perturbation signal to enhance the stimulus; otherwise, it enters a passive identification waiting mode. S3: Use the virtual inertia parameters in the multi-parameter joint identification results to calculate the inertia characterization of each node, and perform voltage support parameter identification to output the inertia characterization and voltage support characteristic parameters of each node. S4: Calculate the inertia gap and voltage gap based on the inertia characterization and voltage support characteristic parameters of each node; The inertia support amount of each node is calculated based on the inertia gap, and the voltage support amount of each node is calculated based on the voltage gap and the voltage support characteristic parameters. After verification in conjunction with the energy storage status of each node, the inertia support amount and voltage support amount are sent to the controller of each node for execution, thereby realizing multi-point voltage inertia support in the low voltage distribution area.

2. The method for multi-point voltage inertia support in low-voltage distribution areas according to claim 1, characterized in that, Step S1 includes the following sub-steps: S101: Extract the frequency of the i-th node of the three-phase power grid from the three-phase voltage and current signals collected by the inverters at each node using a phase-locked loop method. The phase angle is then used to calculate the active power at each node via the Park transformation. and reactive power And read the terminal voltage amplitude. Simultaneously, the angular frequency sequence of each node is calculated from the frequency. This enables the synchronous acquisition and extraction of power grid parameters at each node. Specifically, the formula for calculating the communication cycle is as follows: In the formula, The communication cycle is defined as the data exchange cycle between nodes. The sampling time interval of the ADC analog-to-digital converter; The window length of the Savitzky-Golay adaptive filter; S102: Transform the angular frequency sequence of each node Angular acceleration was calculated using the Savitzky-Golay polynomial smoothing filter method. And construct the augmented regression vector required for AE-JRRLS identification to obtain parameterized data for each node; Specifically, the augmented regression vector The calculation formula is as follows: In the formula, Each component corresponds to the mechanical power reference value, measured active power, and angular frequency deviation; The VSG power reference value is input to the pendulum equation; The three-phase active power is calculated by the Park transformation on the inverter output side. This is the difference between the current angular frequency and the rated angular frequency.

3. The method for multi-point voltage inertia support in low-voltage distribution areas according to claim 2, characterized in that, Step S2 includes the following sub-steps: S201: Measure the data at time t at each node: angular acceleration Mechanical power reference value Measure active power Angular frequency deviation By linearizing the augmented pendulum equation into three parameters, we define the augmented parameter vector, the relationship between its components, and the mathematical formula for multi-period joint identification. The calculation formula for the augmented pendulum equation is as follows: In the formula, Parameters characterizing the rotational inertia of a simulated synchronous generator in an energy storage inverter; This is the differential of the angular frequency with respect to time. The parameter to be identified is the ratio of the inductive component to the total impedance in the low-voltage distribution area's line impedance. ; Parameters characterizing the damping properties of a virtual synchronous generator; This is a reference value for mechanical power. To measure active power; This refers to the angular frequency deviation. Divide both sides of the calculation formula for the augmented pendulum equation by J, and let... , , This allows the original nonlinearly coupled three parameters to be... , , Transform into about The linear regression form is used to achieve three-parameter linearization; the calculation formula for the augmented parameter vector is as follows: In the formula, Let be the augmented parameter vector at time t, defined as a column vector containing three parameters to be identified; for , is defined as the reciprocal of the virtual inertia; for The ratio is defined as the ratio of the line impedance inductance to the virtual inertia. for The ratio is defined as the ratio of the virtual damping coefficient to the virtual inertia. in, Represented as the algorithm's response to the physical truth. The recursive estimate; subscript Indicates the first An independent estimator instance for each energy storage inverter node; subscript Indicates the first The recursive timing of each communication cycle; S202: The nodes of each... Periodic measurement data angular acceleration Augmented regression vector and the parameter estimate of the previous time step The identification residual at the current time is calculated using the joint regression residual formula; then the weighting coefficient is calculated using the Huber weighting function; finally, the three parameters are synchronously updated using the JRRLS gain matrix to obtain the node. Parameter estimates at current time ; Specifically, the formula for calculating the joint regression residuals is as follows: In the formula, For nodes No. The joint regression residuals of the period represent the identified residuals at the current time step, and are coupled with , , Three parameters; For nodes The joint estimation results of the three parameters in the previous communication cycle; The formula for calculating the Huber weight function is as follows: In the formula, For nodes No. The periodic Huber weight function value represents the dimensionless weight for reducing the weight of larger residuals; The Huber robustness threshold is a parameter that controls the degree of weight reduction. The formula for calculating the JRRLS gain vector is as follows: In the formula, For nodes No. The periodic JRRLS gain vector determines the step size for each update of the three parameters; For nodes The covariance matrix at the previous time step is a matrix that reflects the uncertainty of parameter estimation; Forgetting factor, a parameter that controls the decay of historical data weights; The calculation formula for the synchronous update of the three parameters is as follows: In the formula, For nodes No. The joint estimate of the three parameters of the period is the result of the three parameters being updated synchronously in the same formula.

4. The multi-point voltage inertia support method for low-voltage distribution areas according to claim 3, characterized in that, Step S2 also includes the following sub-steps: S203: Move the node No. Periodic JRRLS gain vector Augmented regression vector and nodes The covariance matrix of the previous time step Updated using the recursive formula for the covariance matrix This yields the identification accuracy measurement matrix for the current moment; Specifically, the recursive calculation formula for the covariance matrix is ​​as follows: In the formula, For nodes No. The periodic covariance matrix is ​​a three-dimensional symmetric positive definite matrix, reflecting the current accuracy status of the three-parameter identification; It is a three-dimensional identity matrix.

5. The method for multi-point voltage inertia support in low-voltage distribution areas according to claim 4, characterized in that, Step S2 also includes the following sub-steps: S204: After the single-node recursive update is completed, each node exchanges data with the others, performs a topology Laplace regularization correction, and improves the overall identification accuracy by utilizing the spatial continuity of parameters of adjacent nodes to realize the line impedance-inductance ratio parameter of each node. Cooperative convergence; Specifically, the calculation formula for the topological Laplace regularization correction is as follows: In the formula, The gradient step size is defined as a parameter that controls the magnitude of the regularization correction. For the current cycle Extracted estimated value of the nodal line impedance-to-inductance ratio; These are the topological regularization coefficients; Topological weights are defined as nodes. With nodes The correlation weight between them; For nodes The set of neighboring nodes is defined as the set of nodes with respect to each other. The set of nodes that are directly connected in the network topology of the transformer substation; The correction results are written back to each node. The corresponding component is used as the initial value for the next communication cycle; this correction is implemented in a distributed manner, with each node only needing to exchange with its neighboring nodes. Scalar values ​​do not require a central node for coordination.

6. The method for multi-point voltage inertia support in low-voltage distribution areas according to claim 5, characterized in that, The expression for the joint optimization objective is as follows: In the formula, For the first The true values ​​of the three parameters corresponding to each node; A robust loss function for applying Huber weighting to the residuals; is the topological regularity coefficient, which represents the parameter that controls the strength of the constraint on the consistency of parameters between adjacent nodes; The edge set of the network topology of the transformer area is defined as a set that describes the connection relationships between nodes; For nodes With nodes The correlation weights between nodes are learned adaptively online from the cross-correlation between nodes, without the need for a prior network admittance matrix; For nodes The line impedance-inductance ratio obtained from the identification is defined as... Extracted node-level parameters to be identified; the joint optimization objective is the mathematical basis for the multi-cycle iteration of steps S202 to S204: the first term of the AE-JRRLS recursive minimization in step S202 and the second term of the topological regularization minimization in step S204 are repeatedly iterated until convergence, which is to solve the expression of the joint optimization objective.

7. The method for multi-point voltage inertia support in low-voltage distribution areas according to claim 6, characterized in that, Step S2 also includes the following sub-steps: S205: Move the node No. Periodic covariance matrix The inverse of the covariance matrix is ​​calculated through eigenvalue decomposition. The smallest eigenvalue and the corresponding minimum eigenvector This yields the weakest direction of the current identification accuracy in the three-dimensional parameter space; Specifically, the calculation formula for finding the minimum eigenvalue is as follows: In the formula, The smallest eigenvalue of the inverse covariance matrix is ​​defined as a real-time index reflecting the sufficiency of the identification algorithm for the directional excitation of each parameter. The corresponding minimum eigenvector is defined as a unit vector indicating the direction of the weakest parameter of the excitation; each node undergoes multiple iterations from steps S201 to S204 for parameter estimation. Gradually converged to That is, the expression of the joint optimization objective is used to jointly identify the solution of the optimization problem; S206: with For the execution cycle, find the smallest eigenvalue of the inverse covariance matrix. With the threshold of continuous incentive sufficiency By comparing the results, it is determined whether the current identification meets the continuous excitation condition of the three-parameter joint identification, and the excitation sufficiency judgment result is obtained. Specifically, the criteria for judging the sufficiency of incentives are as follows: If the above formula is true, then the active injection mode is triggered; in the formula, The threshold for sustained incentive sufficiency is defined as a threshold parameter for judging whether the current incentive is sufficient, and is determined based on the statistical lower bound of the historical distribution of the inverse eigenvalues ​​of the covariance matrix. S207: In active injection mode, the minimum feature vector By constrained optimization, the injected power signal is calculated to ensure that the injected signal is within the range of... Maximize the power component in the direction to obtain and execute the injected power command; Specifically, the calculation formula for the injection signal optimization is as follows: The above formula represents that in Find the condition that makes Take the maximum value In the formula, The amount of power injected for active excitation; For nodes The rated apparent power of the energy storage inverter is used to constrain the upper limit of single-node injection. The change in the augmented regression vector caused by power injection is defined as follows: ; S208: When judging When the system enters a passive identification and waiting mode, it continues to perform parameter updates in steps S201 to S204 by utilizing the changes in active power, reactive power, and terminal voltage triggered by natural disturbance events in the transformer area. No active injection is required, and the system waits for the next communication cycle to repeat steps S205 to S206 for judgment.

8. The method for multi-point voltage inertia support in low-voltage distribution areas according to claim 7, characterized in that, Step S3 includes the following sub-steps: S301: After the joint identification from steps S201 to S204 converges through multiple iterations, take the nodes... First component Independent estimates of the virtual inertia of each node are obtained. ; to each node Based on the rated power of the energy storage inverter at this node, the inertia constant of each node is independently calculated. This yields the inertia identification results for each node in the current transformer area. Specifically, the formula for converting the nodal inertia constant is as follows: In the formula, For nodes The identified virtual inertia; For nodes The rated apparent power of the energy storage inverter is used as the reference for the per-unit conversion of the node's inertia constant. This is the system's rated angular frequency constant; S302: Each node executes the voltage identification module in parallel, utilizing the terminal voltage uploaded by each node's S1. and reactive power The data is used to identify the voltage sensitivity coefficient of each node through the voltage-reactive power sensitivity equation, thereby obtaining the voltage support parameters of each node in the current transformer area. ; Specifically, the calculation formula for the voltage-reactive power sensitivity is as follows: In the formula, For nodes The voltage-reactive power sensitivity coefficient is defined as a parameter that describes the magnitude of the terminal voltage change caused by the change in reactive power at a node. Voltage deviation is defined as node The difference between the terminal voltage and the reference voltage; The change in reactive power is defined as the node. The increase in reactive power.

9. The method for multi-point voltage inertia support in low-voltage distribution areas according to claim 8, characterized in that, Step S4 includes the following sub-steps: S401: Set the inertia constant of each node and voltage support parameters of each node The inertia gap of each node is calculated independently, and the voltage gap is calculated simultaneously to obtain the amount of inertia and voltage support that the current transformer area needs to supplement. Specifically, the calculation formulas for the nodal inertia gap and voltage gap are as follows: In the formula, The overall target inertia constant of the distribution area is set for scheduling, which is the overall inertia level that the distribution area is expected to achieve. For nodes The target inertia constant is the value allocated from the total target of the distribution area to the nodes according to the rated capacity ratio. The target value; For nodes Inertia constant; For nodes The inertia gap is the difference between the target inertia and the identified inertia at that node. The rated terminal voltage is defined as the system's rated operating voltage. The average voltage measured in the current transformer area is the average voltage value of the transformer area collected in real time. Voltage gap is defined as the difference between the rated voltage and the measured voltage. S402: Perform a safety check on the energy storage status of each node: check the state of charge of each node. Does it meet the requirements? If node of If the node does not participate in inertia support during this communication cycle, its assigned value will be set accordingly. , The remaining nodes that meet the conditions proceed to S403 to calculate their respective support parameters. S403: Move the node Inertia gap The virtual inertia and damping coefficient support parameters are directly calculated based on the node's own rated power. Specifically, the calculation formula for the conversion of the supporting parameters is as follows: In the formula, For nodes The virtual inertia allocation value is the inertia support amount calculated from the node's own inertia gap. For nodes The rated apparent power of the energy storage inverter; This is the system's rated angular frequency constant; is the damping ratio tuning factor, and is the design parameter that controls the relationship between the damping ratio and the inertia ratio; For nodes The damping coefficient allocation value is the damping support parameter obtained by converting the inertia of the node according to the damping ratio. S404: Virtual inertia of each node parameter and damping coefficient The command is sent to the virtual synchronous generator controller of the energy storage inverter at each node for execution, and the quantitative command update of the inertia support parameter is completed for this communication cycle.

10. A multi-point voltage inertia support device for low-voltage distribution areas, used to perform the multi-point voltage inertia support method for low-voltage distribution areas as described in any one of claims 1-9, characterized in that, include: The node controller is deployed on the nodes of the low-voltage distribution area and includes a data acquisition module, an AE-JRRLS three-parameter joint identification module, an inertia conversion and voltage identification module, an inertia gap calculation and support parameter conversion module, a virtual synchronous generator VSG controller, and an inverter power tube drive controller. The system coordination module is used to receive the total target inertia constant of the distribution area issued by the scheduling system, and broadcast the total target inertia constant to each node controller through the communication network module, without performing inertia identification or supporting parameter calculation. The communication network module is used for bidirectional measurement data exchange between each node controller and exchange of estimated values ​​of the line impedance-inductance ratio of adjacent nodes required for topology Laplace regularization correction, as well as communication by the system coordination module to each node controller to send parameters of the total target inertia constant of the broadcast station area. The data acquisition module is used to acquire three-phase voltage and current signals, extract grid frequency and phase angle through phase-locked loop, calculate active power, reactive power and terminal voltage amplitude through Park transform, calculate angular acceleration through Savitzky-Golay filtering, and construct augmented regression vector. The AE-JRRLS three-parameter joint identification module is used to combine the line impedance inductance ratio, virtual moment of inertia and virtual damping coefficient to form a three-dimensional augmented parameter vector. The three parameters are identified synchronously by recursive least squares weighted by Huber weight function. The minimum eigenvalue of the covariance matrix inverse is used to judge the excitation sufficiency and generate an active injection signal when the excitation is insufficient. In the multi-node scenario, the estimated line impedance inductance ratio is exchanged with the adjacent nodes to perform topological Laplace regularization correction. The inertia conversion and voltage identification module is used to convert the virtual moment of inertia and inertia constant of this node from the first component of the identified parameter vector, and to perform voltage-reactive sensitivity identification in parallel and output the voltage-reactive sensitivity coefficient of this node. The inertia gap calculation and support parameter conversion module is used to independently calculate the target inertia constant of this node based on the total target inertia constant of the station area broadcast by the communication network and the rated capacity ratio of this node, and then calculate the inertia gap of this node. After verifying the charging state limit of this node, the virtual moment of inertia allocation value and damping coefficient allocation value are directly converted from the inertia gap of this node itself. The virtual synchronous generator (VSG) controller is used to receive support parameters, execute the rotor motion equation of the virtual synchronous generator, calculate the active power and reactive power reference values, and generate a three-phase voltage reference signal through voltage-current dual closed-loop control. The inverter power transistor drive controller is used to generate six PWM switching signals through space vector pulse width modulation (SVPWM) based on the three-phase voltage reference signal output by the VSG controller. These signals are then used to directly control the on / off state of the power switching transistors of each bridge arm of the energy storage inverter via an isolation drive circuit, thereby converting the DC bus voltage into three-phase AC power output.

Images