Source network load storage integrated n-1 scene voltage safety analysis method

Abstract

The invention relates to the technical field of power system safety analysis, in particular to a source-grid-load-storage integrated n-1 scene voltage safety analysis method, which comprises the following steps of: acquiring and preprocessing n-1 scene data; constructing a static voltage safety analysis model based on the gradient boosting regression tree; constructing a dynamic voltage safety analysis model based on a long short-term memory network; analyzing safety collaboration of the static voltage and the dynamic voltage; and verifying and optimizing the static voltage safety analysis model and the dynamic voltage safety analysis model. According to the method, the static voltage stability margin and the dynamic response characteristic are fused, so that the comprehensive risk identification accuracy is improved; a gradient boosting regression tree and a long-short-term memory network are adopted to capture nonlinear characteristics, and the voltage prediction error of an n-1 scene is remarkably reduced; characteristics and weights are customized for different source network load storage integrated project types such as industrial and incremental power distribution networks, and the pertinence and engineering practicability of analysis are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system security analysis, in particular to a source-grid-load-storage integrated n-1 scenario voltage security analysis method. BACKGROUND

[0002] The source-grid-load-storage integrated project (such as industrial enterprises, incremental distribution networks, and whole village development) significantly changes the voltage characteristics of the public power grid by integrating distributed power sources (photovoltaic, wind power), loads, energy storage, and internal networks; the n-1 scenario (such as line tripping and transformer withdrawal) as a typical fault scenario for safe operation of the power grid will cause redistribution of power flow, which may cause local voltage out-of-limit or instability.

[0003] The prior art has the following deficiencies in n-1 scenario voltage security analysis: (1) static and dynamic analysis are separated, traditional methods mostly evaluate static voltage stability margin or dynamic voltage recovery process separately, without considering the correlation between the two (such as low static margin may exacerbate the risk of dynamic voltage collapse); (2) the model precision is insufficient, pure mechanism models are difficult to capture the nonlinear dynamic response of source-grid-load-storage equipment (such as photovoltaic inverters and energy storage converters), and pure data-driven models lack physical constraints and have poor generalization ability under extreme scenarios; (3) poor scene adaptability, the voltage response characteristics of different types of source-grid-load-storage projects (such as industrial enterprises containing a large number of inductive loads and incremental distribution networks containing a high proportion of distributed power sources) differ significantly, and existing methods lack targeted analysis strategies.

[0004] Therefore, there is an urgent need for an n-1 scenario voltage security analysis method that integrates static and dynamic characteristics and adapts to different project types. SUMMARY

[0005] The purpose of the present application is to provide a source-grid-load-storage integrated n-1 scenario voltage security analysis method to solve the problems raised in the background.

[0006] To solve the above technical problems, the present application provides a source-grid-load-storage integrated n-1 scenario voltage security analysis method, comprising the following steps: S1, collecting device parameters, operation data and n-1 scenario data of the source-grid-load-storage integrated project and the public power grid, and preprocessing; S2, constructing a static voltage security model using gradient boosting regression trees to predict the static voltage stability margin under the n-1 scenario ; S3, constructing a dynamic voltage security model using a long short-term memory network to predict the voltage drop depth , voltage recovery time , and voltage oscillation amplitude under the n-1 scenario S4, fuse static and dynamic indicators through a synergistic triggering mechanism to calculate a comprehensive risk index , divide voltage safety levels and output prevention and control suggestions; S5, verify the model based on measured and predicted data, dynamically optimize the model parameters, and adjust the model adaptability for different types of source-grid-load-storage integrated projects.

[0007] As a further improvement of the technical solution, in S1, the device parameters, operation data and n-1 scenario data of the source-grid-load-storage integrated project and the public power grid are collected and preprocessed, including the following steps: S1.1, collect the basic data of the source-grid-load-storage integrated project and the public power grid; S1.2, clean, standardize and time-align the data to generate a sample set of n-1 event characteristics-voltage safety indicators.

[0008] As a further improvement of the technical solution, in S1.1, the basic data of the source-grid-load-storage integrated project and the public power grid are collected, specifically including: Device parameters: capacity of distributed power supply, control strategy; charge and discharge characteristics of energy storage system, SOC constraint, impedance of internal line / transformer, rated capacity; Operation data: time series data of voltage, active / reactive power at the gateway; output / load data of each unit inside the source-grid-load-storage; n-1 scenario data: historical n-1 event records and corresponding voltage response curves.

[0009] These two settings can comprehensively cover the core parameters and operation scenarios of the source-grid-load-storage system and the public power grid by clearly defining the specific categories and ranges of basic data, avoiding incomplete model input due to missing key information; data cleaning, standardization and time alignment processing can effectively eliminate interference information such as outliers and noise, unify data dimensions and eliminate time sequence misalignment problems, and the finally generated n-1 event characteristics-voltage safety indicator sample set can ensure the high quality and consistency of the data, providing reliable data support for the training of subsequent static voltage safety model and dynamic voltage safety model, and ensuring the accuracy of the analysis results from the data source.

[0010] As a further improvement of the technical solution, in S2, a gradient boosting regression tree is used to construct a static voltage safety model to predict the static voltage stability margin under n-1 scenarios , including the following steps: S2.1, generate multiple input samples of n-1 scenarios through Monte Carlo sampling, including distributed power output of the source-grid-load-storage system, active and reactive power of various loads, charge and discharge power of the energy storage system, type and location of possible elements that may be taken out of operation; S2.2 For each sample, power flow calculations under the n-1 scenario are performed using a detailed source-grid-load-storage system simulation model to obtain the steady-state voltage at the grid inlet. and the corresponding injected active power reactive power To form a dataset ,in , The number of samples; S2.3. A static voltage response model is constructed using a gradient boosting regression tree to fit a nonlinear mapping relationship. During model training, the parameters are optimized by minimizing the mean square error between the predicted voltage and the actual voltage. S2.4. Based on the above static voltage security analysis model, the static voltage stability margin is calculated using the continuous power flow method. .

[0011] In this setup, Monte Carlo sampling comprehensively covers the combination of multiple variables such as distributed power output, load fluctuations, energy storage charging and discharging, and component decommissioning locations under the n-1 scenario, avoiding scenario omissions. Combined with power flow calculations from the source-grid-load-storage system simulation model, it ensures that the dataset conforms to the physical laws of the power system and is reasonable. Gradient boosting regression trees can accurately capture the nonlinear relationship between voltage and power, and parameter optimization through mean square error can further reduce prediction bias. Finally, by using the continuous power flow method to calculate the static voltage stability margin, the system's ability to resist voltage collapse can be accurately quantified, significantly improving the comprehensiveness and accuracy of static voltage safety assessment.

[0012] As a further improvement to this technical solution, in step S3, a long short-term memory network is used to construct a dynamic voltage safety model to predict the voltage drop depth in the n-1 scenario. Voltage recovery time and voltage oscillation amplitude It includes the following steps: S3.1. An LSTM model is constructed using a Long Short-Term Memory (LSTM) network to capture the temporal dependencies of voltage. The training data for the LSTM model comes from detailed electromagnetic transient simulations: for each possible n-1 scenario, waveform data of voltage changes over time after a fault are generated through simulation. Synchronously record the system state variables before the fault. The dataset format is ,in , The number of samples is dynamic; S3.2, The LSTM model structure includes an input layer, hidden layers, and an output layer: Input layer: Receives system state variables and fault type identifier; Hidden layer: Employs 3 layers of LSTM units, each containing 64 neurons, and controls the flow of information through a gating mechanism; Output layer: Outputs predicted voltage values ​​for the next 5000 time points. ; S3.3 After obtaining the voltage dynamic response curve through the LSTM model, the dynamic voltage safety status under the n-1 scenario is determined by calculating the dynamic voltage stability evaluation index.

[0013] As a further improvement to this technical solution, in S3.2, the loss function for training the LSTM model is the root mean square error, and the model parameters are iteratively updated through the Adam optimizer until the loss function converges.

[0014] As a further improvement to this technical solution, in S3.3, the dynamic voltage safety assessment index includes voltage sag depth. Voltage recovery time and voltage oscillation amplitude ; Voltage sag depth Minimum voltage after the fault With rated voltage The ratio; Voltage recovery time For the voltage to drop and then recover to the rated voltage 90% of the required time; Voltage oscillation amplitude The peak-to-peak value of the voltage fluctuation after the fault versus the rated voltage The ratio of .

[0015] Of these three settings, the LSTM network can effectively handle the long-term time-series dependence of voltage dynamic response. Combined with training data generated by detailed electromagnetic transient simulation, it can realistically reproduce the voltage change process after a fault. The LSTM structure with 3 hidden layers and 64 neurons and the gating mechanism can stably and efficiently process time-series data and accurately output voltage prediction values ​​at 5000 time points. The clearly defined calculation methods for voltage drop depth, voltage recovery time, and voltage oscillation amplitude can transform the abstract dynamic voltage response into quantifiable indicators, providing a clear basis for judging the safety status of dynamic voltage and comprehensively improving the accuracy and operability of dynamic voltage analysis.

[0016] As a further improvement to this technical solution, in step S4, a comprehensive risk index is calculated by fusing static and dynamic indicators through a collaborative triggering mechanism. The process of classifying voltage safety levels and providing prevention and control recommendations includes the following steps: S4.1, Set the critical value for static voltage stability margin. When the static voltage stability margin... When the voltage is less than the critical value of the static voltage stability margin, the dynamic voltage model is automatically activated to assess the risk of dynamic instability. S4.2. Integrating static and dynamic characteristics to define a comprehensive risk index. ; S4.3, Based on the comprehensive risk index The voltage safety status is divided into three states: safe, warning, and dangerous, and corresponding prevention and control suggestions are output.

[0017] This feature integrates a comprehensive risk index that combines static and dynamic characteristics, breaking the limitations of traditional fragmented analysis. It can fully integrate risk information from both types of indicators; by classifying safety levels according to the index and outputting prevention and control recommendations, it can transform complex analysis results into intuitive status judgments and actionable countermeasures, thereby improving the pertinence and decision-making efficiency of voltage safety management.

[0018] As a further improvement to this technical solution, in step S5, the model is validated based on measured and predicted data, the model parameters are dynamically optimized, and the model adaptability is adjusted for different types of integrated source-grid-load-storage projects, including the following steps: S5.1 Compare the model predictions with the actual measured data to ensure the static voltage stability margin. The errors of both the dynamic voltage safety assessment index and the dynamic voltage safety assessment index meet the preset accuracy requirements; S5.2. Incorporate new n-1 event data every quarter and retrain the model; for scenarios where the error exceeds the standard, optimize the model through parameter identification. S5.3 Adapt the model to different types of projects.

[0019] As a further improvement to this technical solution, in S5.3, the models for different types of projects are adapted and adjusted, specifically including the following items: Industrial enterprise projects: The dynamic voltage safety analysis model adds a motor load ratio parameter, and the LSTM input layer of the dynamic voltage safety analysis model incorporates the motor starting current characteristics; Incremental distribution network projects: The gradient boosting regression tree of the static voltage security analysis model is used to add parameters related to the output of distributed power sources. The dynamic voltage security analysis model considers the impact of the inverter's low voltage ride-through strategy and improves the model's adaptability by introducing virtual synchronous machine control parameters. For family workshops and whole-village development projects: the static voltage safety analysis model uses a time-varying parameter gradient boosting regression tree, and the training samples for the dynamic voltage safety analysis model need to cover fault scenarios during peak residential electricity consumption. For manufacturing enterprise projects: The dynamic voltage safety analysis model introduces a detailed model that considers electromechanical transient processes. The state variables of the rotating motor are added as inputs to the Long Short-Term Memory (LSTM) network. At the same time, the model structure is optimized to accurately simulate the dynamic interaction process between the rotating motor and the system, and to quantify the impact of electromechanical transient coupling effects on dynamic voltage safety.

[0020] These two settings, by comparing measured and predicted data and setting error requirements, can verify the model's accuracy and ensure that the results meet engineering requirements; quarterly data updates and model retraining, as well as optimization of parameters for scenarios exceeding standards, can adapt to system changes and avoid accuracy degradation; customized adjustments for different projects can match the voltage characteristics of various scenarios, solve the problem of poor adaptability of traditional models, and improve engineering practicality.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: This method for voltage safety analysis in the integrated generation-grid-load-storage (N-1) scenario combines static voltage stability margin with dynamic response characteristics to address the fragmented assessment problem of traditional methods and improve the accuracy of comprehensive risk identification. It employs gradient boosting regression trees and long short-term memory networks to capture nonlinear characteristics, combined with mechanistic constraints to ensure physical rationality, significantly reducing voltage prediction errors in the N-1 scenario. Furthermore, it customizes features and weights for different types of integrated generation-grid-load-storage projects, such as industrial and incremental distribution networks, enhancing the relevance and engineering applicability of the analysis. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0024] like Figure 1 As shown in the figure, this embodiment provides a voltage security analysis method for an integrated source-grid-load-storage n-1 scenario, including the following steps: S1. Collect equipment parameters, operating data, and n-1 scenario data from integrated power generation, grid, load, and storage projects and public power grids, and perform preprocessing. S2. A gradient boosting regression tree is used to construct a static voltage safety model, capturing the nonlinear mapping between static voltage and power, and predicting the static voltage stability margin in scenario n-1. ; S3. A dynamic voltage safety model is constructed using a long short-term memory network to effectively handle the temporal dependencies of dynamic voltage and predict the voltage drop depth in the n-1 scenario. Voltage recovery time and voltage oscillation amplitude ; S4. Calculate the comprehensive risk index by integrating static and dynamic indicators through a collaborative triggering mechanism. It classifies voltage safety levels and outputs prevention and control suggestions, taking into account both physical mechanism constraints and data-driven advantages, and significantly reduces voltage prediction errors. S5. Verify the model based on measured and predicted data, dynamically optimize model parameters, and adjust the model adaptability for different types of integrated source-grid-load-storage projects.

[0025] In this embodiment, in S1, the equipment parameters, operating data, and n-1 scenario data of the integrated source-grid-load-storage project and the public power grid are collected and preprocessed, including the following steps: S1.1 Collect basic data on integrated power generation, grid, load and storage projects and public power grids; S1.2. Clean the data (remove outliers), standardize (per-unit value processing), and align the time sequence to effectively filter noise and interference information, avoid the impact of missing key information or inconsistent data on subsequent model training, and generate a sample set of n-1 event features - voltage safety indicators. This provides high-quality and highly consistent input data support for static voltage safety models and dynamic voltage safety models, ensuring the accuracy and reliability of subsequent analysis results from the data source.

[0026] Specifically, in S1.1, basic data for integrated power generation, grid, load, and storage projects and public power grids are collected, including: Equipment parameters: capacity and control strategy of distributed power source; charging and discharging characteristics, SOC constraints, impedance of internal lines / transformers, and rated capacity of energy storage system, to ensure that the model can accurately simulate the hardware characteristics and operating limitations of the system and avoid voltage safety assessment deviations caused by misjudgment of equipment capabilities; Operational data: Time-series data of voltage and active / reactive power at the control point; output / load data of each unit within the source-grid-load-storage system (such as photovoltaic output and industrial load fluctuations), which reflect the dynamic operating status of the system in real time, provide the model with input that closely matches the actual operating conditions, and ensure that the analysis results are consistent with the on-site operating conditions; n-1 scenario data: Historical n-1 event records (such as fault component type, occurrence time, and duration) and corresponding voltage response curves (static stability margin, dynamic recovery trajectory) directly provide fault scenario samples, enabling the model to learn based on real fault patterns and avoid the problem of poor generalization ability in extreme cases due to lack of fault scenario data.

[0027] In this embodiment, in S2, a gradient boosting regression tree is used to construct a static voltage safety model to predict the static voltage stability margin in scenario n-1. The static voltage security analysis model is used to evaluate the voltage distribution and stability margin of the system after reaching a new steady state in the n-1 scenario. Its core is to establish an integrated static voltage-power response characteristic model of source-grid-load-storage and gateway equivalent, and to incorporate it into the network topology and parameters of the public power grid. The model includes the following steps: The static voltage security analysis model of the integrated power generation, grid, load and storage gateway takes the static relationship between node injected power and voltage as its core. In the n-1 scenario, the power flow transfer caused by component exit will change the power injection at the gateway, thereby affecting the voltage level. When using the data-driven modeling method, it is necessary to construct the mapping relationship between the voltage at the gateway and the injected active and reactive power based on historical operating data and simulation data. S2.1. Generate multivariate input samples for the n-1 scenario through Monte Carlo sampling, including the output of distributed power sources in the source-grid-load-storage system (e.g., photovoltaic power adopts Beta distribution, wind power adopts Weibull distribution), the active and reactive power of various loads (e.g., following normal distribution), the charging and discharging power of the energy storage system (dynamically adjusted according to the state of charge SOC), and the types and locations of components that may be taken out of operation (e.g., lines, transformers, etc.). This comprehensively covers the combination of multivariate variables such as distributed power output fluctuations, load changes, energy storage charging and discharging status, and component exit locations, avoiding static risk omissions due to scenario omissions. S2.2 For each sample, power flow calculations under the n-1 scenario are performed using a detailed source-grid-load-storage system simulation model to obtain the steady-state voltage at the grid inlet. and the corresponding injected active power reactive power To form a dataset ,in , The number of samples; S2.3, Using gradient boosting regression trees ( A static voltage response model is constructed, and a nonlinear mapping relationship is fitted. This model can accurately capture the complex mapping relationship between the cut-off voltage and the injected power. The corresponding expression is:

[0028] in, For n-1 scene identifier variables (representing the type and location of the exiting element). This is a gradient boosting regression tree model; during model training, the mean square error between the predicted voltage and the actual voltage is minimized. Optimize parameters to further reduce static voltage prediction error; the corresponding expression is: ; in, Mean square error, For the first The model predicts the voltage for each sample. For the first The actual voltage obtained from simulation or measurement of a single sample; S2.4 Static voltage stability margin is a key indicator for measuring the system's ability to resist voltage collapse in the n-1 scenario. It is defined as the distance from the current operating point to the voltage stability limit point. Based on the above static voltage security analysis model, the continuous power flow method is used to calculate the static voltage stability margin. Its core is solving the extended form of the power flow equations: ; in, This is the load growth factor. For nodes steady-state voltage Load growth coefficient The active power below, For nodes steady-state voltage Load growth coefficient The reactive power below, , These are the initial active and reactive power, respectively. , These are the active and reactive load growth coefficients, respectively. Reference node voltage; when When the system reaches its voltage stability limit, the static voltage stability margin is: ; in, This is the maximum load growth factor. This is the load growth factor for the current operating point (usually taken as 0); In the n-1 scenario, the exit parameters of different components are calculated using the gate equivalence model. ,like This indicates that the system has a risk of static voltage instability, which solves the problem of the lack of physical constraints in the pure data-driven model and improves the rationality of the data.

[0029] In this embodiment, in S3, a long short-term memory network is used to construct a dynamic voltage safety model to predict the voltage drop depth in the n-1 scenario. Voltage recovery time and voltage oscillation amplitude Dynamic voltage safety analysis focuses on the transient process after component deactivation in the n-1 scenario, evaluating the dynamic response characteristics of voltage after disturbance. Power electronic devices (such as photovoltaic inverters and energy storage converters) in integrated power generation, grid, load, and energy storage systems have fast response characteristics, and their dynamic behavior significantly affects the voltage recovery process. Therefore, a high-precision dynamic response model needs to be constructed using a data-driven approach, including the following steps: The input to the dynamic voltage safety analysis model is the system state at the time of the n-1 fault occurrence (such as the voltage, power, energy storage, etc. of each node), and the output is the voltage dynamic response curve within 0-5 seconds after the fault. S3.1. An LSTM model is constructed using a Long Short-Term Memory (LSTM) network to capture the temporal dependencies of voltage. This is because it effectively handles long-term dependencies in time-series data and can effectively address long-term temporal correlations in the dynamic voltage response after a fault (such as the hysteresis effect in the voltage recovery process). This overcomes the limitation of traditional time-series models in handling long-period dependencies. The training data for the LSTM model comes from detailed electromagnetic transient simulations: for each possible n-1 scenario (such as line tripping, transformer disconnection, etc.), waveform data of voltage changes over time after a fault are generated through simulation. (Sampling frequency is 1kHz), synchronously record the system state variables before the fault. (Including gate-injected power, initial voltage, distributed power output, etc.), the dataset format is as follows: ,in , With a dynamic sample size, detailed electromagnetic transient simulation can realistically reproduce the voltage change waveform over time after a fault in different n-1 scenarios (such as drop, oscillation, recovery), and synchronously record the system state variables before the fault. S3.2 The LSTM model structure includes an input layer, hidden layers (containing LSTM units), and an output layer: Input layer: Receives system state variables and fault type identifier; Hidden layer: Employs 3 layers of LSTM units, each containing 64 neurons, and controls the flow of information through a gating mechanism; Output layer: Outputs predicted voltage values ​​for the next 5000 time points (typically 5 seconds). ; S3.3 After obtaining the voltage dynamic response curve through the LSTM model, the dynamic voltage safety status under the n-1 scenario is determined by calculating the dynamic voltage stability evaluation index.

[0030] Specifically, in S3.2, the loss function for training the LSTM model is the root mean square error (RMSE). This model is more sensitive to abnormal fluctuations in voltage dynamic response (such as sudden drops and oscillation peaks), and can more accurately measure the deviation between model predictions and actual values. It is especially suitable for non-stationary dynamic processes of voltage after a fault, avoiding underfitting of key dynamic features by the model due to the insensitivity of the loss function. The corresponding expression is: ; in, The root mean square error, The number of time steps for each sample For the first Each sample simulation generates waveform data of voltage changes over time after a fault. For the first The voltage prediction value output by the LSTM model for each sample; The model parameters are iteratively updated using the Adam optimizer until the loss function converges. The Adam optimizer, which combines momentum gradient descent with adaptive learning rate, can effectively accelerate the model convergence speed.

[0031] Furthermore, dynamic voltage safety assessment metrics include voltage sag depth. Voltage recovery time and voltage oscillation amplitude Voltage drop depth The severity of a voltage drop can be quantified, providing a basis for determining whether equipment protection actions (such as motor stall or inverter disconnection) should be triggered, avoiding misjudgments of drop risk due to a lack of quantitative standards, and improving voltage recovery time. It can assess the system's voltage recovery capability, which is directly related to the reliability of load power supply (such as the recovery time requirements of sensitive loads), and provides a quantitative reference for formulating voltage recovery control strategies (such as energy storage reactive power support), as well as voltage oscillation amplitude. It can assess the stability of voltage after a fault, and avoid equipment fatigue damage or secondary system instability caused by excessive oscillation; Voltage sag depth Minimum voltage after the fault With rated voltage The ratio, corresponding to: ; like (Adjusted according to system voltage level), then it is judged as a severe voltage drop; Voltage recovery time For the voltage to drop and then recover to the rated voltage 90% of the required time, requirements Less than 2 seconds; Voltage oscillation amplitude The peak-to-peak value of the voltage fluctuation after the fault versus the rated voltage The ratio, corresponding to the expression: ; in, This represents the peak value of the voltage fluctuation after the fault. This represents the valley value of the voltage fluctuation after the fault.

[0032] In this embodiment, in step S4, a comprehensive risk index is calculated by fusing static and dynamic indicators through a collaborative triggering mechanism. The process of classifying voltage safety levels and providing prevention and control recommendations includes the following steps: S4.1. Set the critical value for static voltage stability margin. If the critical value for static voltage stability margin is equal to 0.05, when the static voltage stability margin... When it is less than the critical value of the static voltage stability margin, that is When the dynamic voltage model is automatically activated to assess the risk of dynamic instability, redundant dynamic analysis of static safety scenarios can be avoided, significantly reducing computational load and improving overall analysis efficiency; at the same time, it ensures that only during voltage stability margin... When there are deficiencies (potential risks), conduct in-depth assessments of dynamic instability risks to avoid underestimating risks due to neglecting the correlation between static and dynamic states; S4.2 Integrate static and dynamic characteristics to comprehensively consolidate voltage safety risk information, ensuring the completeness and objectivity of risk assessment, and define a comprehensive risk index. The corresponding expression is: ; ; in, , All are weights, default. , Adjustments will be made dynamically based on the project type. S4.3, Based on the comprehensive risk index Voltage safety status is divided into three states: safe, warning, and dangerous, and corresponding prevention and control suggestions are output (such as increasing reactive power generation through energy storage, and cutting off part of the load). For example, the safe state corresponds to... Warning status correspondence Response to dangerous situations .

[0033] In this embodiment, step S5 involves validating the model based on measured and predicted data, dynamically optimizing the model parameters, and adjusting the model's adaptability for different types of integrated source-grid-load-storage projects. This includes the following steps: S5.1 Compare the model predictions with the actual measured data to ensure the static voltage stability margin. The errors of both the dynamic voltage safety assessment indicators and the static voltage stability margin meet the preset accuracy requirements. The error is ≤5%, and the error of the dynamic voltage safety assessment index is ≤10%. S5.2. Incorporate new n-1 event data each quarter and retrain the model; for scenarios where the error exceeds the standard, identify the parameters (e.g., adjust the gradient to boost the regression tree). The tree depth and the learning rate of the Long Short-Term Memory (LSTM) network are optimized to make the model continuously adapt to the dynamic changes of the source-grid-load-storage system (such as the addition of distributed power sources or changes in load characteristics), avoiding the decline in model accuracy due to system upgrades or changes in operating conditions, and ensuring the long-term effectiveness of the model. S5.3 Adapt the model to different types of projects.

[0034] Specifically, in S5.3, the models for different types of projects are adapted for adaptation, including the following projects: Industrial enterprise projects: Industrial enterprises typically contain large asynchronous motors. During an n-1 fault, motor stall may cause a voltage sag. The dynamic voltage safety analysis model adds a motor load ratio parameter and incorporates motor starting current characteristics into the LSTM input layer of the dynamic voltage safety analysis model to avoid dynamic voltage prediction deviations caused by ignoring the dynamic effects of the motor, thereby improving the prediction accuracy of voltage sags. Incremental distribution network projects: Incremental distribution networks have a large number of distributed power sources connected to them. In the n-1 scenario, multiple power sources may coordinate their responses. The gradient boosting regression tree of the static voltage safety analysis model adds parameters related to the output of distributed power sources. The dynamic voltage safety analysis model considers the impact of the inverter's low voltage ride-through strategy. The model's adaptability is improved by introducing virtual synchronous machine (VSG) control parameters, thereby improving the model's fitting accuracy to the inverter's dynamics. For family workshops and whole-village development projects: These projects have dispersed and highly volatile loads. In the n-1 scenario, the voltage response is random. The static voltage safety analysis model uses a time-varying parameter gradient boosting regression tree (parameters are dynamically updated over time). The training samples for the dynamic voltage safety analysis model need to cover fault scenarios during peak residential electricity consumption (such as 6-10 pm) to adapt to the temporal variation characteristics of the load and avoid static voltage prediction errors and incomplete coverage of dynamic scenarios caused by load randomness. For manufacturing projects: If rotating motors such as those used for waste heat power generation are included, the electromechanical transient coupling effect is significant in the n-1 scenario. The dynamic voltage safety analysis model introduces a detailed model that considers the electromechanical transient process. The power angle, angular velocity and other state variables of the rotating motor are added as inputs to the Long Short-Term Memory (LSTM) network. At the same time, the model structure is optimized to accurately simulate the dynamic interaction process between the rotating motor and the system, quantify the impact of the electromechanical transient coupling effect on dynamic voltage safety, and solve the problem of inaccurate dynamic safety assessment caused by the inability of traditional models to simulate electromechanical coupling.

[0035] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A voltage security analysis method for an integrated source-grid-load-storage scenario in an n-1 scenario, characterized in that, Includes the following steps: S1. Collect equipment parameters, operating data, and n-1 scenario data from integrated power generation, grid, load, and storage projects and public power grids, and perform preprocessing. S2. Construct a static voltage safety model using gradient boosting regression trees to predict the static voltage stability margin in scenario n-1. ; S3. A dynamic voltage safety model is constructed using a long short-term memory network to predict the voltage drop depth in the n-1 scenario. Voltage recovery time and voltage oscillation amplitude ; S4. Calculate the comprehensive risk index by integrating static and dynamic indicators through a collaborative triggering mechanism. Classify voltage safety levels and provide prevention and control recommendations; S5. Verify the model based on measured and predicted data, dynamically optimize model parameters, and adjust the model adaptability for different types of integrated source-grid-load-storage projects.

2. The voltage security analysis method for an integrated source-grid-load-storage scenario (n-1 scenario) according to claim 1, characterized in that, In step S1, the equipment parameters, operating data, and n-1 scenario data of the integrated power generation, grid, load, and storage project and the public power grid are collected and preprocessed, including the following steps: S1.1 Collect basic data on integrated power generation, grid, load and storage projects and public power grids; S1.2 Clean, standardize, and align the data to generate a sample set of n-1 event features and voltage safety indicators.

3. The voltage security analysis method for an integrated source-grid-load-storage scenario (n-1 scenario) according to claim 2, characterized in that, In S1.1, the basic data collected for the integrated power generation, grid, load, and storage project and the public power grid specifically includes: Equipment parameters: capacity and control strategy of distributed power sources; charging and discharging characteristics, SOC constraints, impedance of internal lines / transformers, and rated capacity of energy storage systems; Operational data: Voltage at the control point, time-series data of active / reactive power; output / load data of each unit within the power source, grid, load, and storage system; n-1 scenario data: historical n-1 event records and corresponding voltage response curves.

4. The voltage security analysis method for an integrated source-grid-load-storage scenario (n-1 scenario) according to claim 1, characterized in that, In step S2, a gradient boosting regression tree is used to construct a static voltage safety model to predict the static voltage stability margin in scenario n-1. It includes the following steps: S2.1 Generate multivariate input samples for the n-1 scenario through Monte Carlo sampling, including the distributed power output of the source-grid-load-storage system, the active and reactive power of various loads, the charging and discharging power of the energy storage system, and the types and locations of components that may be taken out of operation. S2.2 For each sample, power flow calculations under the n-1 scenario are performed using a detailed source-grid-load-storage system simulation model to obtain the steady-state voltage at the grid inlet. and the corresponding injected active power reactive power To form a dataset ,in , The number of samples; S2.

3. A static voltage response model is constructed using a gradient boosting regression tree to fit a nonlinear mapping relationship. During model training, the parameters are optimized by minimizing the mean square error between the predicted voltage and the actual voltage. S2.

4. Based on the above static voltage security analysis model, the static voltage stability margin is calculated using the continuous power flow method. .

5. The voltage security analysis method for an integrated source-grid-load-storage scenario (n-1 scenario) according to claim 1, characterized in that, In step S3, a dynamic voltage safety model is constructed using a long short-term memory network to predict the voltage drop depth in the n-1 scenario. Voltage recovery time and voltage oscillation amplitude It includes the following steps: S3.

1. An LSTM model is constructed using a Long Short-Term Memory (LSTM) network to capture the temporal dependencies of voltage. The training data for the LSTM model comes from detailed electromagnetic transient simulations: for each possible n-1 scenario, waveform data of voltage changes over time after a fault are generated through simulation. Synchronously record the system state variables before the fault. The dataset format is ,in , The number of samples is dynamic; S3.2, The LSTM model structure includes an input layer, hidden layers, and an output layer: Input layer: Receives system state variables and fault type identifier; Hidden layer: Employs 3 layers of LSTM units, each containing 64 neurons, and controls the flow of information through a gating mechanism; Output layer: Outputs predicted voltage values ​​for the next 5000 time points. ; S3.3 After obtaining the voltage dynamic response curve through the LSTM model, the dynamic voltage safety status under the n-1 scenario is determined by calculating the dynamic voltage stability evaluation index.

6. The voltage security analysis method for integrated source-grid-load-storage n-1 scenario according to claim 5, characterized in that: In S3.2, the loss function for training the LSTM model is the root mean square error, and the model parameters are iteratively updated through the Adam optimizer until the loss function converges.

7. The voltage security analysis method for integrated source-grid-load-storage n-1 scenario according to claim 5, characterized in that, In S3.3, the dynamic voltage safety assessment index includes voltage sag depth. Voltage recovery time and voltage oscillation amplitude ; Voltage sag depth Minimum voltage after the fault With rated voltage The ratio; Voltage recovery time For the voltage to drop and then recover to the rated voltage 90% of the required time; Voltage oscillation amplitude The peak-to-peak value of the voltage fluctuation after the fault versus the rated voltage The ratio of .

8. The voltage security analysis method for an integrated source-grid-load-storage scenario (n-1 scenario) according to claim 1, characterized in that, In step S4, a comprehensive risk index is calculated by integrating static and dynamic indicators through a collaborative triggering mechanism. The process of classifying voltage safety levels and providing prevention and control recommendations includes the following steps: S4.1, Set the critical value for static voltage stability margin. When the static voltage stability margin... When the voltage is less than the critical value of the static voltage stability margin, the dynamic voltage model is automatically activated to assess the risk of dynamic instability. S4.

2. Integrating static and dynamic characteristics to define a comprehensive risk index. ; S4.3, Based on the comprehensive risk index The voltage safety status is divided into three states: safe, warning, and dangerous, and corresponding prevention and control suggestions are output.

9. The voltage security analysis method for an integrated source-grid-load-storage scenario (n-1 scenario) according to claim 1, characterized in that, In step S5, the model is validated based on measured and predicted data, the model parameters are dynamically optimized, and the model adaptability is adjusted for different types of integrated source-grid-load-storage projects, including the following steps: S5.1 Compare the model predictions with the actual measured data to ensure the static voltage stability margin. The errors of both the dynamic voltage safety assessment index and the dynamic voltage safety assessment index meet the preset accuracy requirements; S5.

2. Incorporate new n-1 event data every quarter and retrain the model; for scenarios where the error exceeds the standard, optimize the model through parameter identification. S5.3 Adapt the model to different types of projects.

10. The voltage security analysis method for an integrated source-grid-load-storage scenario (n-1 scenario) according to claim 9, characterized in that, In S5.3, the models for different types of projects are adapted and adjusted, specifically including the following items: Industrial enterprise projects: The dynamic voltage safety analysis model adds a motor load ratio parameter, and the LSTM input layer of the dynamic voltage safety analysis model incorporates the motor starting current characteristics; Incremental distribution network projects: The gradient boosting regression tree of the static voltage security analysis model is used to add parameters related to the output of distributed power sources. The dynamic voltage security analysis model considers the impact of the inverter's low voltage ride-through strategy and improves the model's adaptability by introducing virtual synchronous machine control parameters. For family workshops and whole-village development projects: the static voltage safety analysis model uses a time-varying parameter gradient boosting regression tree, and the training samples for the dynamic voltage safety analysis model need to cover fault scenarios during peak residential electricity consumption. For manufacturing enterprise projects: The dynamic voltage safety analysis model introduces a detailed model that considers electromechanical transient processes. The state variables of the rotating motor are added as inputs to the Long Short-Term Memory (LSTM) network. At the same time, the model structure is optimized to accurately simulate the dynamic interaction process between the rotating motor and the system, and to quantify the impact of electromechanical transient coupling effects on dynamic voltage safety.

Images