A test result analysis system of a flexible direct current control protection system

Abstract

The application discloses a kind of test result analysis systems of flexible direct current control protection system, it is related to power management technical field, including intelligent fault diagnosis and classification module, dynamic simulation and prediction module, abnormal event link tracking module, fault recovery path optimization module and system robustness evaluation module.The test result analysis system of flexible direct current control protection system, by introducing fault classification model, nonlinear state space model and link tracking model, not only effectively overcome the deficiency of traditional system in fault diagnosis, simulation prediction and abnormal event tracking, also significantly improve the intelligent level and dynamic response capability of system.Make the system show higher accuracy and stability when responding to complex power failure and variable dynamic operating environment, provide solid guarantee for improving the reliability and security of power grid.

Description

Technical Field

[0001] This invention relates to the field of power management technology, specifically to a test result analysis system for a flexible DC control and protection system. Background Technology

[0002] Flexible DC control and protection systems play a crucial role in modern power systems, primarily ensuring the stability and reliability of the power grid. However, with the increasing complexity of power systems, traditional control and protection systems face numerous challenges in handling rapidly changing dynamic environments, nonlinear fault detection, and big data analysis.

[0003] The existing technology has the following shortcomings:

[0004] Limitations of intelligent fault diagnosis: Traditional fault diagnosis systems rely on preset rules and human knowledge bases, which cannot adapt to the complex dynamic behavior of flexible DC systems and have significant limitations in identifying diverse fault types.

[0005] Insufficient accuracy in dynamic simulation and prediction: Most existing simulation tools only support linear or approximately nonlinear models, lacking real-time performance and accurate description of the complex nonlinear characteristics of the system, resulting in a large deviation between simulation results and the actual system state.

[0006] Inefficiency of anomaly event tracing: Current anomaly event tracing methods mostly use static analysis techniques, which cannot effectively capture the dynamic process of event development, easily overlook potential causal relationships in the system, and reduce the accuracy of fault location and analysis. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a test result analysis system for a flexible DC control and protection system, thereby resolving the problems mentioned in the background section.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a test result analysis system for a flexible DC control and protection system, comprising an intelligent fault diagnosis and classification module, a dynamic simulation and prediction module, an abnormal event link tracing module, a fault recovery path optimization module, and a system robustness assessment module;

[0009] The intelligent fault diagnosis and classification module automatically identifies and classifies various faults occurring in flexible DC systems. Through big data-driven methods, it trains fault modes using actual operating data to quickly and accurately detect different types of faults, including short-circuit faults, insulation faults, and communication faults.

[0010] The dynamic simulation and prediction module is used to simulate the behavior of flexible DC systems under different operating conditions in order to predict possible changes in the system in a specific future time period. By simulating different operating scenarios, such as load changes, equipment failures and system switching, the stability and reliability of the system under various conditions are evaluated.

[0011] The abnormal event link tracing module performs link tracing and analysis on the development process of abnormal events, quickly locates the root cause of the problem, and reproduces the entire process of the abnormal event based on time series analysis and association rule mining technology, including the event triggering conditions, the behavior of related equipment, and the scope of the final failure impact.

[0012] The fault recovery path optimization module intelligently optimizes the recovery path after a fault occurs, restoring the system to normal operation in the shortest time and with the least resources. It uses the shortest path algorithm and resource scheduling algorithm, combined with the network topology and device health status, to generate the optimal recovery plan.

[0013] The system robustness assessment module is used for stability and reliability analysis of flexible DC control and protection systems.

[0014] To further optimize this technical solution, the intelligent fault diagnosis and classification module, based on the flexible DC system, combines the advantages of deep learning and time series analysis to construct a fault classification model, which is used to handle the nonlinear and dynamic characteristics of the flexible DC system.

[0015] The fault classification model includes the convolutional layer feature extraction formula, the LSTM temporal feature integration formula, and the fault classification output formula.

[0016] The fault classification model periodically feeds the latest fault data into the model and uses an incremental learning algorithm to fine-tune the model parameters, continuously improving the diagnostic accuracy of the fault classification model and enhancing the robustness of the model under different operating conditions.

[0017] To further optimize this technical solution, in the fault classification model:

[0018] The formula for feature extraction from convolutional layers is shown below:

[0019] ;

[0020] in,

[0021] Indicates at time step Feature map;

[0022] It is the convolution kernel, i.e., the weight matrix;

[0023] The input signal at time step The original data;

[0024] It is the bias term of the convolutional layer;

[0025] It is an activation function, such as ReLU;

[0026] The formula for integrating LSTM temporal features is shown below:

[0027] ;

[0028] in,

[0029] In time step The hidden state;

[0030] It is the weight matrix of the hidden layer;

[0031] It is a unit state;

[0032] , , These are the forget gate, the input gate, and the candidate state;

[0033] It is the joint input of the current feature and the previous hidden state;

[0034] It is the hyperbolic tangent function, used to introduce nonlinear characteristics;

[0035] The fault classification output formula is shown below:

[0036] ;

[0037] in,

[0038] It is the predicted probability distribution for fault classification;

[0039] It is the weight matrix of the output layer;

[0040] It is a hidden state from the LSTM;

[0041] It is the bias term of the output layer.

[0042] To further optimize this technical solution, the fault classification model, when used, includes the following specific processes:

[0043] Data preprocessing and feature extraction: The raw data of the flexible DC system is preprocessed, including normalization, noise removal and segmentation into input segments with fixed time steps. Then, the processed data is input into the convolutional layer, and local features at the time step are extracted by the convolutional layer feature extraction formula.

[0044] Temporal modeling and dynamic feature integration: Next, the features extracted by the convolutional layer are input into the LSTM network. The LSTM network uses the LSTM temporal feature integration formula to integrate time series information and predict future states based on the current and past system states.

[0045] Fault Classification and Output: Finally, the output of the LSTM network is fed into the softmax layer, and is converted into a probability distribution of fault types using the fault classification output formula.

[0046] To further optimize this technical solution, the dynamic simulation and prediction module predicts the behavior of the flexible DC system in dynamic simulation. It incorporates a nonlinear state-space model (NL-SSM) and combines it with an adaptive predictive control algorithm (APC) to adjust the system simulation and prediction parameters in real time. The formula for the nonlinear state-space model is shown below:

[0047] ;

[0048] in,

[0049] Is the system in time State vectors, such as voltage, current, and device status;

[0050] These are control input vectors, such as load changes and scheduling strategies;

[0051] It is a nonlinear function of state transition, describing the change of system state over time;

[0052] It is a nonlinear function of the observation equation, used to map the state to the observation data;

[0053] and These are system process noise and observation noise, respectively, which are modeled as random variables;

[0054] Nonlinear state-space model (NL-SSM) is used to describe the state changes of flexible DC systems at different time steps, through nonlinear functions. and The model captures the dynamic behavior of the system and the impact of inputs on the system state.

[0055] To further optimize this technical solution, the adaptive predictive control algorithm incorporates an adjustment formula, which is shown below:

[0056] ;

[0057] in,

[0058] It is time The control input vector;

[0059] It is a gain matrix, used to adjust the amplitude of the control input;

[0060] It represents the difference between observations, used to adaptively adjust the control strategy;

[0061] Adaptive predictive control algorithms continuously adjust the control input vector To optimize the dynamic simulation process of the system, the control strategy is adjusted in real time based on the changes between the current observation and the previous observation, so that the system can respond quickly in the dynamic environment and maintain the high accuracy and relevance of the simulation results.

[0062] To further optimize this technical solution, the dynamic simulation and prediction module, when in use, includes the following process:

[0063] Initial condition setting and state-space modeling: Before conducting the simulation, the initial conditions are first set according to the operating conditions of the flexible DC system, including the initial state of the system. and control input Substituting these initial conditions into the formula of the nonlinear state-space model, through and The function calculates the system's state and observations at the initial time point;

[0064] Adaptive predictive control in dynamic simulation: updating the state vector based on real-time dynamic data input. and control input vector The system uses adjustment formulas to adjust the control input, enabling it to adapt quickly to changes in the external environment.

[0065] Multi-scenario simulation and prediction: Dynamically adjust the functions in the nonlinear state-space model according to different system operating scenarios such as load fluctuations and equipment failures. and The model is optimized by training it on historical data under different operating conditions in order to more accurately predict the system's behavior in specific situations.

[0066] Simulation Result Analysis and Adjustment Feedback: Finally, analyze key indicators in the simulation results, such as system stability, response time, and energy loss, and feed these results back into the control input adjustment formula. If the simulation results deviate from the expected values, adjust the gain matrix accordingly. The size of the variable will be used to further optimize the adaptive predictive control strategy, thereby improving the accuracy of future simulations and predictions.

[0067] To further optimize this technical solution, the abnormal event link tracing module has a built-in link tracing model. The link tracing model treats the event development process as a multi-node causal network and uses association rule mining technology to analyze the relationship between different events, and to trace and reconstruct the event link.

[0068] The link tracing model includes formulas for node relationships in an Event Causal Network (ECN) and formulas for mining time series association rules;

[0069] The event causal network node relationship formula determines whether there is a direct causal relationship between events by analyzing the conditional probability of different events. Only when the conditional probability exceeds a threshold is the relationship between the two events considered to be related.

[0070] The time series association rule mining formula analyzes the correlation between events under specific conditions by calculating confidence and lift.

[0071] To further optimize this technical solution, in the link tracing model:

[0072] The formula for the causal network node relationship is shown below:

[0073] ;

[0074] in,

[0075] and These represent two event nodes in a flexible DC system;

[0076] Indicates in the event After the incident, The conditional probability of occurrence;

[0077] It is a preset threshold used to determine the causal relationship between events;

[0078] The formula for mining time series association rules is shown below:

[0079] ;

[0080] in,

[0081] It is an association rule. The confidence level represents the confidence level in the event Post-event The probability of it also occurring;

[0082] It's the lift, used to measure the performance of an event. and events The strength of the correlation between them;

[0083] Indicates an event Frequency of occurrence throughout the entire event sequence.

[0084] To further optimize this technical solution, the system robustness assessment module evaluates the system's ability to resist abnormal events and sudden failures by analyzing the system's response behavior under different operating conditions.

[0085] A robustness analysis method based on Monte Carlo simulation is adopted to predict the system's performance under complex operating conditions through a large number of random simulations. Through multi-parameter variation analysis, the weakest link in the system that is most susceptible to the impact is accurately identified, providing a scientific basis for improving and enhancing the system's stability.

[0086] Compared with the prior art, the present invention provides a test result analysis system for a flexible DC control and protection system, which has the following beneficial effects:

[0087] The test result analysis system for this flexible DC control and protection system, by introducing fault classification models, nonlinear state-space models, and link tracing models, not only effectively overcomes the shortcomings of traditional systems in fault diagnosis, simulation prediction, and abnormal event tracking, but also significantly improves the system's intelligence level and dynamic response capability. This enables the system to exhibit higher accuracy and stability when dealing with complex power faults and changing dynamic operating environments, providing a solid guarantee for improving the reliability and security of the power grid. Through the collaborative work of multiple modules, the system can automatically locate the root cause of faults, predict future system changes, and issue early warnings in a very short time, providing engineers with efficient analysis and decision support. Attached Figure Description

[0088] Figure 1 This is a schematic diagram of the structure of the test result analysis system of the flexible DC control and protection system proposed in this invention;

[0089] Figure 2 This is a flowchart illustrating the fault classification model in the test result analysis system of a flexible DC control and protection system proposed in this invention.

[0090] Figure 3This is a flowchart illustrating the dynamic simulation and prediction module in the test result analysis system of a flexible DC control and protection system proposed in this invention.

[0091] Figure 4 This is a flowchart illustrating the link tracing model in the test result analysis system of a flexible DC control and protection system proposed in this invention. Detailed Implementation

[0092] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0093] Example:

[0094] Please see Figure 1 A test result analysis system for a flexible DC control and protection system includes an intelligent fault diagnosis and classification module, a dynamic simulation and prediction module, an abnormal event link tracing module, a fault recovery path optimization module, and a system robustness assessment module.

[0095] The intelligent fault diagnosis and classification module automatically identifies and classifies various faults occurring in flexible DC systems. Through big data-driven methods, it trains fault modes using actual operating data to quickly and accurately detect different types of faults, including short-circuit faults, insulation faults, and communication faults.

[0096] Traditional fault diagnosis systems typically require long-term model training, while this module can update the model in real time based on new data during operation, thus improving the accuracy of diagnosis.

[0097] In this embodiment, the intelligent fault diagnosis and classification module, based on the flexible DC system, combines the advantages of deep learning and time series analysis to construct a fault classification model to handle the nonlinear and dynamic characteristics of the flexible DC system.

[0098] The fault classification model includes the convolutional layer feature extraction formula, the LSTM temporal feature integration formula, and the fault classification output formula.

[0099] The fault classification model periodically feeds the latest fault data into the model and uses an incremental learning algorithm to fine-tune the model parameters, continuously improving the diagnostic accuracy of the fault classification model and enhancing the robustness of the model under different operating conditions.

[0100] Furthermore, in the fault classification model:

[0101] The formula for feature extraction from convolutional layers is shown below:

[0102] ;

[0103] in,

[0104] Indicates at time step Feature map;

[0105] It is the convolution kernel, i.e., the weight matrix;

[0106] The input signal at time step The original data;

[0107] It is the bias term of the convolutional layer;

[0108] It is an activation function, such as ReLU.

[0109] Convolutional layers are used to extract local features from time-series data of flexible DC systems. These features represent the state changes of the system at different time points. The purpose of this step is to extract the spatial characteristics of the data, preparing for subsequent time-series analysis.

[0110] The formula for integrating LSTM temporal features is shown below:

[0111] ;

[0112] in,

[0113] In time step The hidden state;

[0114] It is the weight matrix of the hidden layer;

[0115] It is a unit state;

[0116] , , These are the forget gate, the input gate, and the candidate state;

[0117] It is the joint input of the current feature and the previous hidden state;

[0118] It is the hyperbolic tangent function, used to introduce nonlinear characteristics.

[0119] The spatial features extracted by the convolutional layers are fused with the temporal features in the time series data. The LSTM network utilizes its memory units to capture long-term dependencies, enabling it to accurately model the behavioral changes of the system at different time points.

[0120] The fault classification output formula is shown below:

[0121] ;

[0122] in,

[0123] It is the predicted probability distribution for fault classification;

[0124] It is the weight matrix of the output layer;

[0125] It is a hidden state from the LSTM;

[0126] It is the bias term of the output layer.

[0127] The hidden states are transformed into a probability distribution using the softmax function to predict the type of fault. This classification method can accurately map various faults in flexible DC systems (such as short-circuit faults, insulation faults, communication faults, etc.) to specific categories.

[0128] like Figure 2 As shown, the fault classification model, when used, includes the following specific procedures:

[0129] Data preprocessing and feature extraction: The raw data of the flexible DC system is preprocessed, including normalization, noise removal and segmentation into input segments with fixed time steps. Then, the processed data is input into the convolutional layer, and local features at the time step are extracted by the convolutional layer feature extraction formula.

[0130] Temporal modeling and dynamic feature integration: Next, the features extracted by the convolutional layer are input into the LSTM network. The LSTM network uses the LSTM temporal feature integration formula to integrate time series information and predict future states based on the current and past system states.

[0131] Fault Classification and Output: Finally, the output of the LSTM network is fed into the softmax layer, and is converted into a probability distribution of fault types using the fault classification output formula.

[0132] Traditional models often ignore local spatial features when processing time series data, while this model, by combining convolutional layers with LSTM, can process both spatial and temporal information of the system simultaneously, making fault diagnosis more accurate.

[0133] The model employs incremental learning, automatically updating its learning parameters as new data arrives, thus improving its fault identification capabilities in new environments. This dynamic adjustment method significantly enhances the model's adaptability and long-term performance.

[0134] The dynamic simulation and prediction module is used to simulate the behavior of flexible DC systems under different operating conditions in order to predict possible changes in the system in a specific future time period. By simulating different operating scenarios, such as load changes, equipment failures and system switching, the stability and reliability of the system under various conditions are evaluated.

[0135] In this embodiment, the dynamic simulation and prediction module predicts the behavior of the flexible DC system in dynamic simulation. It incorporates a nonlinear state-space model (NL-SSM) and combines it with an adaptive predictive control algorithm (APC) to adjust the parameters of the system simulation and prediction in real time. The formula for the nonlinear state-space model is shown below:

[0136] ;

[0137] in,

[0138] Is the system in time State vectors, such as voltage, current, and device status;

[0139] These are control input vectors, such as load changes and scheduling strategies;

[0140] It is a nonlinear function of state transition, describing the change of system state over time;

[0141] It is a nonlinear function of the observation equation, used to map the state to the observation data;

[0142] and These are system process noise and observation noise, respectively, which are modeled as random variables;

[0143] Nonlinear state-space model (NL-SSM) is used to describe the state changes of flexible DC systems at different time steps, through nonlinear functions. and The model captures the dynamic behavior of the system and the impact of inputs on the system state.

[0144] Furthermore, the adaptive predictive control algorithm has a built-in adjustment formula, which is shown below:

[0145] ;

[0146] in,

[0147] It is time The control input vector;

[0148] It is a gain matrix, used to adjust the amplitude of the control input;

[0149] It represents the difference between observations, used to adaptively adjust the control strategy;

[0150] Adaptive predictive control algorithms continuously adjust the control input vector To optimize the dynamic simulation process of the system, the control strategy is adjusted in real time based on the changes between the current observation and the previous observation, so that the system can respond quickly in the dynamic environment and maintain the high accuracy and relevance of the simulation results.

[0151] like Figure 3 As shown, the dynamic simulation and prediction module, when used, includes the following process:

[0152] Initial condition setting and state-space modeling: Before conducting the simulation, the initial conditions are first set according to the operating conditions of the flexible DC system, including the initial state of the system. and control input Substituting these initial conditions into the formula of the nonlinear state-space model, through and The function calculates the system's state and observations at the initial time point;

[0153] Adaptive predictive control in dynamic simulation: updating the state vector based on real-time dynamic data input. and control input vector The system uses adjustment formulas to adjust the control input, enabling it to adapt quickly to changes in the external environment.

[0154] Multi-scenario simulation and prediction: Dynamically adjust the functions in the nonlinear state-space model according to different system operating scenarios such as load fluctuations and equipment failures. and The model is optimized by training it on historical data under different operating conditions in order to more accurately predict the system's behavior in specific situations.

[0155] Simulation Result Analysis and Adjustment Feedback: Finally, analyze key indicators in the simulation results, such as system stability, response time, and energy loss, and feed these results back into the control input adjustment formula. If the simulation results deviate from the expected values, adjust the gain matrix accordingly. The size of the variable will be used to further optimize the adaptive predictive control strategy, thereby improving the accuracy of future simulations and predictions.

[0156] This model can capture the nonlinear dynamic characteristics of flexible DC systems and, compared to linear models, can more accurately simulate system behavior under complex operating conditions. This flexibility makes the model perform better when dealing with nonlinear events such as load jumps and equipment failures.

[0157] An adaptive predictive control (APC) algorithm was introduced, enabling the model to dynamically adjust based on real-time feedback from the system. Traditional control methods are typically fixed, while the adaptive mechanism continuously optimizes the simulation strategy as the system state changes, significantly improving the simulation accuracy and real-time performance.

[0158] The abnormal event link tracing module performs link tracing and analysis on the development process of abnormal events, quickly locates the root cause of the problem, and reproduces the entire process of the abnormal event based on time series analysis and association rule mining technology, including the event triggering conditions, the behavior of related devices, and the scope of the final failure impact.

[0159] In this embodiment, the abnormal event link tracing module has a built-in link tracing model. The link tracing model regards the event development process as a multi-node causal network and uses association rule mining technology to analyze the relationship between different events, and to trace and reconstruct the event link.

[0160] The link tracing model includes formulas for node relationships in an Event Causal Network (ECN) and formulas for mining time series association rules;

[0161] The event causal network node relationship formula determines whether there is a direct causal relationship between events by analyzing the conditional probability of different events. Only when the conditional probability exceeds a threshold is the relationship between the two events considered to be related.

[0162] The time series association rule mining formula analyzes the correlation between events under specific conditions by calculating confidence and lift.

[0163] Furthermore, in the link tracing model:

[0164] The formula for the causal network node relationship is shown below:

[0165] ;

[0166] in,

[0167] and These represent two event nodes in a flexible DC system;

[0168] Indicates in the event After the incident, The conditional probability of occurrence;

[0169] It is a preset threshold used to determine the causal relationship between events.

[0170] The formula for mining time series association rules is shown below:

[0171] ;

[0172] in,

[0173] It is an association rule. The confidence level represents the confidence level in the event Post-event The probability of it also occurring;

[0174] It's the lift, used to measure the performance of an event. and events The strength of the correlation between them;

[0175] Indicates an event Frequency of occurrence throughout the entire event sequence.

[0176] like Figure 4 As shown, when using this model:

[0177] Event Data Collection and Preliminary Analysis: First, time-series data containing abnormal events are collected from the flexible DC system. This data includes information such as the event occurrence time, equipment status, and operation records. Through data preprocessing, the raw event data is segmented into time-series segments, and potential key event nodes are identified.

[0178] Constructing an Event Causal Network (ECN): The ECN is constructed using the node relationship formula. Conditional probability analysis is used to determine which event nodes have causal relationships. The ECN graph is then drawn, connecting the nodes and edges to form the initial structure of the event progression chain.

[0179] Association rule mining and dynamic link optimization: Next, the confidence and lift between events are calculated using time series association rule mining formulas. The strength of the association rule is determined by the confidence formula. In this way, the links in the event causal network can be optimized, removing those low-relevance spurious associations and retaining only the paths that are truly likely to cause anomalous events.

[0180] Dynamic link tracing and event reconstruction: After an abnormal event occurs, the cause and development process of the event are traced using a pre-constructed causal network and association rule links. Based on the temporal relationship and causal strength between event nodes, the entire process of the abnormal event is dynamically reconstructed, from the triggering conditions and the behavior of related devices to the scope of the final failure. This dynamic link tracing process helps engineers quickly identify key nodes in the event chain after a system failure, providing important clues for problem localization and resolution.

[0181] Anomaly prediction and early warning mechanism: Utilizing historical data accumulated in the event causal network model, a predictive model is established to predict possible chains of anomalies based on the current system state. Once the initial triggering conditions of a high-risk event are detected, an alert can be sent to operators through the early warning mechanism, prompting them to take necessary preventative measures.

[0182] The fault recovery path optimization module intelligently optimizes the recovery path after a fault occurs, restoring the system to normal operation in the shortest time and with the least resources. It uses the shortest path algorithm and resource scheduling algorithm, combined with the network topology and device health status, to generate the optimal recovery plan.

[0183] In this embodiment, fuzzy logic and intelligent optimization algorithms (such as genetic algorithms or particle swarm optimization algorithms) are introduced, so that the recovery path not only considers physical distance, but also comprehensively evaluates the impact of the fault on the overall system performance and various constraints in the recovery process.

[0184] The system robustness assessment module is used for stability and reliability analysis of flexible DC control and protection systems.

[0185] In this embodiment, the system robustness assessment module evaluates the system's ability to resist abnormal events and sudden failures by analyzing the system's response behavior under different operating conditions.

[0186] A robustness analysis method based on Monte Carlo simulation is adopted to predict the system's performance under complex operating conditions through a large number of random simulations. Through multi-parameter variation analysis, the weakest link in the system that is most susceptible to the impact is accurately identified, providing a scientific basis for improving and enhancing the system's stability.

[0187] The beneficial effects of this invention are:

[0188] The test result analysis system for this flexible DC control and protection system, by introducing fault classification models, nonlinear state-space models, and link tracing models, not only effectively overcomes the shortcomings of traditional systems in fault diagnosis, simulation prediction, and abnormal event tracking, but also significantly improves the system's intelligence level and dynamic response capability. This enables the system to exhibit higher accuracy and stability when dealing with complex power faults and changing dynamic operating environments, providing a solid guarantee for improving the reliability and security of the power grid. Through the collaborative work of multiple modules, the system can automatically locate the root cause of faults, predict future system changes, and issue early warnings in a very short time, providing engineers with efficient analysis and decision support.

[0189] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0190] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A test result analysis system for a flexible DC control and protection system, characterized in that, It includes an intelligent fault diagnosis and classification module, a dynamic simulation and prediction module, an abnormal event link tracing module, a fault recovery path optimization module, and a system robustness assessment module; The intelligent fault diagnosis and classification module automatically identifies and classifies various faults occurring in the flexible DC system. Driven by big data, it uses actual operating data to train fault modes and detect different types of faults, including short-circuit faults, insulation faults, and communication faults. The dynamic simulation and prediction module is used to simulate the behavior of the system under different operating conditions in order to predict possible changes in the system in a specific future time period. By simulating different operating scenarios, including load changes and equipment failures, it evaluates the stability and reliability of the system under various conditions. The dynamic simulation and prediction module includes a nonlinear state-space model and an adaptive predictive control algorithm, which adjusts the parameters for system simulation and prediction in real time. The adaptive predictive control algorithm includes adjustment formulas. ;in, It is time The control input vector; It is the gain matrix. It is the difference between the observations; the adaptive predictive control algorithm adjusts... To optimize the dynamic simulation process of the system, the control strategy is adjusted in real time based on the changes between the current and previous observations. In use, the nonlinear function of state transition in the nonlinear state-space model is dynamically adjusted according to different operating scenarios. and the nonlinear function of the observation equation The model is trained and optimized using historical data under different operating conditions. The simulation results are analyzed to assess system stability, response time, and energy loss. These results are then fed back into the control input adjustment formula. When the simulation results deviate from the expected value, the gain matrix is ​​adjusted. Size-optimized adaptive predictive control strategy; The abnormal event link tracing module performs link tracing and analysis on the development process of abnormal events, quickly locating the root cause of the problem. Based on time series analysis and association rule mining techniques, it reconstructs the entire process of the abnormal event, including the event triggering conditions, the behavior of related equipment, and the scope of the final failure impact. It collects time series data containing abnormal events from the system and filters out potential key event nodes. It constructs an event causal network using conditional probability-related event causal network node relationship formulas, calculates the confidence and lift between events using time series association rule mining formulas, determines the strength of association rules through confidence formulas, and optimizes the links in the event causal network. After an abnormal event occurs, it uses the constructed event causal network and association rule links to trace the cause and development process of the event, and uses historical data in the event causal network to build a predictive model to predict possible abnormal event chains based on the current system state. The fault recovery path optimization module intelligently optimizes the recovery path after a fault occurs, restoring the system to normal operation in the shortest time and with the least resources. It uses the shortest path algorithm and resource scheduling algorithm, combined with the network topology and device health status, to generate the optimal recovery plan. The system robustness assessment module is used for stability and reliability analysis of flexible DC control and protection systems.

2. The test result analysis system for a flexible DC control and protection system according to claim 1, characterized in that, The intelligent fault diagnosis and classification module, based on the flexible DC system, combines the advantages of deep learning and time series analysis to construct a fault classification model, which is used to handle the nonlinear and dynamic characteristics of the flexible DC system. The fault classification model includes the convolutional layer feature extraction formula, the LSTM temporal feature integration formula, and the fault classification output formula. The fault classification model periodically feeds the latest fault data into the model and uses an incremental learning algorithm to fine-tune the model parameters, continuously improving the diagnostic accuracy of the fault classification model and enhancing the robustness of the model under different operating conditions.

3. The test result analysis system for a flexible DC control and protection system according to claim 2, characterized in that, In the fault classification model: The formula for feature extraction from convolutional layers is shown below: ； in, Indicates at time step Feature map; It is the convolution kernel, i.e., the weight matrix; The input signal at time step The original data; It is the bias term of the convolutional layer; It is an activation function, such as ReLU; The formula for integrating LSTM temporal features is shown below: ； in, In time step The hidden state; It is the weight matrix of the hidden layer; It is a unit state; , , These are the forget gate, the input gate, and the candidate state; It is the joint input of the current feature and the previous hidden state; It is the hyperbolic tangent function, used to introduce nonlinear characteristics; The fault classification output formula is shown below: ； in, It is the predicted probability distribution for fault classification; It is the weight matrix of the output layer; It is a hidden state from the LSTM; It is the bias term of the output layer.

4. The test result analysis system for a flexible DC control and protection system according to claim 2, characterized in that, The fault classification model, when used, includes the following specific procedures: Data preprocessing and feature extraction: The raw data of the flexible DC system is preprocessed, including normalization, noise removal and segmentation into input segments with fixed time steps. Then, the processed data is input into the convolutional layer, and local features at the time step are extracted by the convolutional layer feature extraction formula. Temporal modeling and dynamic feature integration: Next, the features extracted by the convolutional layer are input into the LSTM network. The LSTM network uses the LSTM temporal feature integration formula to integrate time series information. The LSTM network uses its memory units to capture long-term dependencies and can accurately model the behavior changes of the system at different time points. Fault Classification and Output: Finally, the output of the LSTM network is fed into the softmax layer, and is converted into a probability distribution of fault types using the fault classification output formula.

5. The test result analysis system for a flexible DC control and protection system according to claim 1, characterized in that, The dynamic simulation and prediction module predicts the behavior of the flexible DC system in dynamic simulation. It incorporates a nonlinear state-space model (NL-SSM) and combines it with an adaptive predictive control algorithm (APC) to adjust the parameters of the system simulation and prediction in real time. The formula for the nonlinear state-space model is shown below: ； in, Is the system in time The state vector includes: voltage, current, and device state; yt is the system's observed value at time t; It is the control input vector, including: load changes and scheduling strategies; It is a nonlinear function of state transition, describing the change of system state over time; It is a nonlinear function of the observation equation, used to map the state to the observation data; and These are system process noise and observation noise, respectively, which are modeled as random variables; Nonlinear state-space model (NL-SSM) is used to describe the state changes of flexible DC systems at different time steps, through nonlinear functions. and The model captures the dynamic behavior of the system and the impact of inputs on the system state.

6. The test result analysis system for a flexible DC control and protection system according to claim 5, characterized in that, When the dynamic simulation and prediction module is used, the following process is included: Initial condition setting and state-space modeling: Before conducting the simulation, the initial conditions are first set according to the operating conditions of the flexible DC system, including the initial state of the system. and control input Substituting these initial conditions into the formula of the nonlinear state-space model, and through... and The function calculates the system's state and observations at the initial time point; Adaptive predictive control in dynamic simulation: updating the state vector based on real-time dynamic data input. and control input vector The system uses adjustment formulas to adjust the control input, enabling it to adapt quickly to changes in the external environment. Multi-scenario simulation and prediction: Dynamically adjust the functions in the nonlinear state-space model according to different system operating scenarios. and The model is optimized by training it with historical data under different operating conditions in order to more accurately predict the system's behavior in specific situations. Simulation Result Analysis and Adjustment Feedback: Finally, the key indicators in the simulation results are analyzed, including system stability, response time, and energy loss. These results are then fed back into the control input adjustment formula. If the simulation results deviate from the expected values, the gain matrix is ​​adjusted. The size of the variable will be used to further optimize the adaptive predictive control strategy, thereby improving the accuracy of future simulations and predictions.

7. The test result analysis system for a flexible DC control and protection system according to claim 1, characterized in that, The abnormal event link tracing module has a built-in link tracing model. The link tracing model treats the event development process as a multi-node causal network and uses association rule mining technology to analyze the relationship between different events, and to trace and reconstruct the event link. The link tracing model includes formulas for node relationships in an Event Causal Network (ECN) and formulas for mining time series association rules; The event causal network node relationship formula determines whether there is a direct causal relationship between events by analyzing the conditional probability of different events. Only when the conditional probability exceeds a threshold is the relationship between the two events considered to be related. The time series association rule mining formula analyzes the correlation between events under specific conditions by calculating confidence and lift.

8. The test result analysis system for a flexible DC control and protection system according to claim 7, characterized in that, In the link tracing model: The formula for the causal network node relationship is shown below: ； in, and These represent two event nodes in a flexible DC system; Indicates in the event After the incident, The conditional probability of occurrence; It is a preset threshold used to determine the causal relationship between events; The formula for mining time series association rules is shown below: ； in, It is an association rule. The confidence level represents the confidence level in the event Post-event The probability of it also occurring; It's the lift, used to measure the performance of an event. and events The strength of the correlation between them; Indicates an event Frequency of occurrence throughout the entire event sequence.

9. The test result analysis system for a flexible DC control and protection system according to claim 1, characterized in that, The system robustness assessment module evaluates the system's ability to resist abnormal events and sudden failures by analyzing the system's response behavior under different operating conditions. A robustness analysis method based on Monte Carlo simulation is adopted to predict the system's performance under complex operating conditions through a large number of random simulations. Through multi-parameter variation analysis, the weakest link in the system that is most susceptible to the impact is accurately identified, providing a scientific basis for improving and enhancing the system's stability.

Images