Intelligent integrated service method and system for anti-overflow and anti-static controller

Abstract

The application provides an intelligent integrated service method and system for an anti-overflow and anti-static controller, relates to the technical field of risk prediction, and comprises the following steps: according to a model ID of a target controller, a fault time sequence tracking data set of a historical sample controller is called, and the target controller is an anti-overflow and anti-static controller; index fluctuation correlation analysis is performed, and a fault fluctuation correlation tree is constructed; a cross-level risk prediction network is constructed; associated sensing equipment polling backtracking is performed, and an intersection index monitoring time sequence array is obtained; a scene matching probability distribution is output, risk prediction level verification is performed, and an integrated risk prediction result is output; a fault propagation path atlas is matched, and source parameter suppression regulation and control are performed. The application solves the technical problem that, in the prior art, the anti-overflow and anti-static controller mostly determines whether a fault occurs by setting a fixed threshold, lacks flexibility and adaptability, is prone to false alarms and missed alarms, and thus affects the operation stability and safety of the equipment.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of risk prediction, in particular to an intelligent integrated service method and system for an anti-overflow and anti-static controller. BACKGROUND

[0002] The anti-overflow and anti-static controller is widely used in industrial equipment and electrical systems, and its main function is to prevent the influence of current, voltage and mechanical failure on the equipment, especially in high sensitivity environments. Such a controller can ensure the stable operation of the equipment by monitoring various physical parameters such as current, voltage and pressure in real time. Most of the existing technologies determine whether a fault occurs by setting fixed thresholds. However, the operating environment and working conditions of the anti-overflow and anti-static controller change greatly, and the fixed thresholds are difficult to adapt to the behavior of the equipment in different situations, leading to misjudgment or missed judgment, and thus failing to discover potential faults and risks in time, affecting the stability and safety of the equipment. SUMMARY

[0003] The application provides an intelligent integrated service method and system for an anti-overflow and anti-static controller, aiming to solve the technical problem that most of the existing anti-overflow and anti-static controllers determine whether a fault occurs by setting fixed thresholds, lack flexibility and adaptability, and are prone to false alarms and missed alarms, thereby affecting the stability and safety of the equipment.

[0004] The first aspect of the application provides an intelligent integrated service method for an anti-overflow and anti-static controller, which comprises the following steps: according to the model ID of a target controller, a historical sample controller's fault time sequence tracking dataset is called, wherein the target controller is an anti-overflow and anti-static controller; based on the fault time sequence tracking dataset, an index fluctuation correlation analysis is performed to construct a fault fluctuation correlation tree; according to the node level linkage attribute of the fault fluctuation correlation tree, a cross-level risk prediction network is constructed; according to the root node monitoring index composition of the fault fluctuation correlation tree, an associated sensing device polling backtracking is performed to obtain an intersection index monitoring time sequence array; after inputting the intersection index monitoring time sequence array into a benchmark risk prediction model of the cross-level risk prediction network, according to the scene matching probability distribution output by the benchmark risk prediction model, a plurality of fault risk prediction models connected in parallel in the cross-level risk prediction network are locally activated to perform risk prediction level verification, and an integrated risk prediction result is output; a fault propagation path atlas matching the integrated risk prediction result is matched to perform source parameter suppression regulation of the target controller.

[0005] In a second aspect, the application discloses an intelligent integrated service system for an anti-overflow and anti-static controller, which is used for the intelligent integrated service method for the anti-overflow and anti-static controller, and comprises a tracking data calling module, a fluctuation correlation analysis module, a prediction network construction module, a polling backtracking module, a level checking module and a suppression and regulation module.

[0006] The one or more technical solutions provided in the application have at least the following beneficial effects:

[0007] Through historical sample data calling and index fluctuation correlation analysis, key monitoring index fluctuation patterns can be extracted based on the fault time sequence tracking dataset, which enables the system to identify potential fault risks in advance, and through the fault fluctuation correlation tree, the triggering factors and propagation paths of different fault types are determined, thereby providing more accurate early warning information for equipment maintenance personnel; based on the node level linkage properties of the fault fluctuation correlation tree, a cross-level risk prediction network is constructed, which can integrate monitoring indexes and fault prediction models at different levels to form a global risk assessment framework; through this integration, multiple complex fault scenarios can be considered simultaneously, and the interactive effects of different fault factors can be comprehensively evaluated, thereby realizing cross-level and cross-domain fault prediction and control; through local activation of multiple fault risk prediction models based on the scene matching probability distribution, different risk levels can be dynamically checked to ensure the reliability and accuracy of the prediction results; through matching the fault propagation path atlas of the integrated risk prediction results, the source index type can be accurately identified, and customized dynamic parameter hierarchical suppression can be implemented, thereby effectively preventing further expansion of the fault; this adaptive regulation strategy can dynamically optimize the operating state of the controller according to real-time data and risk assessment, thereby improving the overall reliability and safety of the equipment.

[0008] The above description is only a summary of the technical solutions of the present application. In order to enable the technical means of the present application to be more clearly understood and implemented according to the content of the specification, and in order to enable the above and other purposes, characteristics and advantages of the present application to be more apparent and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 The intelligent integrated service method for the anti-overflow and anti-static controller provided by the embodiments of the present application is shown in the flowchart.

[0010] Figure 2 The intelligent integrated service system structure diagram for the anti-overflow and anti-static controller provided by the embodiments of the present application is shown in the flowchart.

[0011] Explanation of reference signs: tracking data retrieval module 10, fluctuation correlation analysis module 20, prediction network construction module 30, polling backtracking module 40, hierarchical verification module 50, and suppression regulation module 60. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide an intelligent integrated service method and system for an anti-overflow and anti-static controller, which solves the technical problem that the anti-overflow and anti-static controllers in the prior art mostly determine whether a fault occurs by setting a fixed threshold, lack flexibility and adaptability, are prone to false alarms and missed alarms, and thus affect the operation stability and safety of the equipment.

[0013] After introducing the basic principles of the present application, the various non-limiting embodiments of the present application will be specifically introduced in combination with the drawings of the specification. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0014] Embodiment one, as shown in the drawings, the embodiments of the present application provide an intelligent integrated service method for an anti-overflow and anti-static controller, the method comprising: Figure 1

[0015] According to the model ID of the target controller, the fault time sequence tracking data set of the historical sample controller is retrieved, wherein the target controller is an anti-overflow and anti-static controller.

[0016] ​The model ID of the target controller is the key to identify a specific anti-overflow and anti-static controller, and each controller has a unique model ID to distinguish different controller types and specifications. According to the model ID, the relevant historical sample controller failure time tracking data set is retrieved from the database or historical record system, which is a set of time series data recording the failures of the historical sample controller at different time points and their performance, including the running state of the historical sample controller, monitoring indicators (such as temperature, voltage, flow, etc.), time points of failure, duration, etc. These data match the target controller.

[0017] Based on the failure time tracking data set, an index fluctuation correlation analysis is performed to construct a failure fluctuation correlation tree.

[0018] The index fluctuation correlation analysis is an analysis of multiple monitoring indicators such as temperature, pressure, current, etc. obtained from the failure time tracking data set, to find the fluctuation and correlation between different monitoring indicators. For example, using time series analysis, etc. to identify whether there is a correlation between the changes of each monitoring indicator. On the basis of correlation analysis, a failure fluctuation correlation tree is generated, where each node represents a monitoring indicator, and the edges between each node represent the correlation strength and fluctuation relationship between them. The failure fluctuation correlation tree shows the dependency relationship between each monitoring indicator in a tree structure, providing structured information for subsequent risk prediction and fault diagnosis.

[0019] According to the node level linkage properties of the failure fluctuation correlation tree, a cross-level risk prediction network is constructed.

[0020] The node level linkage property refers to how nodes affect each other, for example, the fluctuation of a certain monitoring indicator directly affects the running state or fault type of the upper layer controller. Based on the node level linkage properties of the failure fluctuation correlation tree, a cross-level risk prediction network is constructed, aiming to predict the probability or risk of failure according to the relationship between different levels. In the construction process, algorithms such as neural networks, decision trees, support vector machines, etc. are used for machine learning methods. These algorithms are trained based on historical data to provide accurate failure risk prediction.

[0021] According to the root node monitoring indicator composition of the failure fluctuation correlation tree, an associated sensor device polling backtracking is performed to obtain an intersection indicator monitoring time series array.

[0022] In the fault fluctuation correlation tree, the root node represents the most critical monitoring indicators in the system, and the monitoring indicators of the root node constitute the basis for all subsequent related analysis, which is the starting point of the entire monitoring and prediction. According to the monitoring indicators of the root node, the associated sensor devices are polled. Polling is to obtain data points closely related to the monitoring indicators of the root node from the associated sensor devices. For example, if the root node is temperature, the associated sensor devices polled can be temperature sensors, pressure sensors, flow sensors, etc. After obtaining the data through polling, a backtracking operation is performed. The backtracking process is to identify the similarity between the current monitoring data and the historical samples through the comparison of historical data, focusing on finding the time points that have similar fluctuation patterns in the historical data. Through backtracking, the intersection indicator monitoring time series array is obtained, which contains the monitoring data related to the root node that has appeared in the historical data and meets the specific fluctuation characteristics. These data represent the time series of various sensor indicators in a certain period of time and are the key input for subsequent risk prediction.

[0023] After inputting the intersection indicator monitoring time series array into the benchmark risk prediction model in the cross-level risk prediction network, according to the scene matching probability distribution output by the benchmark risk prediction model, the multiple fault risk prediction models in parallel in the cross-level risk prediction network are locally activated to perform risk prediction level verification, and the integrated risk prediction result is output.

[0024] The intersection indicator monitoring time series array is input into the benchmark risk prediction model in the cross-level risk prediction network. This model has been trained and can predict the risk of device failure based on the input time series data. The output of the benchmark risk prediction model is a scene matching probability distribution, which represents the probability of different failure scenarios of the controller under the given monitoring data. For example, the probability of a certain type of failure (such as electrical failure, temperature overload failure, etc.).

[0025] According to the scene matching probability distribution, the parallel models in the cross-level risk prediction network are locally activated. These parallel fault risk prediction models are prediction models for different types of failures. For example, if the benchmark risk prediction model outputs a high probability of electrical failure, the related electrical failure risk prediction model will be activated for in-depth analysis. These parallel models perform risk prediction level verification, which is cross-verification and correction of the prediction results of multiple models to ensure the accuracy of the prediction. The prediction results of each model are compared with other models to verify consistency, and the final risk assessment is adjusted according to the verification result. The integrated risk prediction result is the final result obtained by combining the prediction outputs of all models, which includes a comprehensive assessment of the current failure risk of the target controller and the specific type and probability of future possible failures, ensuring the comprehensiveness and accuracy of the prediction.

[0026] The fault propagation path atlas matching the integrated risk prediction result is used to perform source parameter suppression regulation of the target controller.

[0027] The fault propagation path atlas is a systematic chart used to describe the propagation process from the source indicator to the occurrence of the fault. By analyzing the influence and linkage of different nodes, the fault propagation path atlas is drawn to show how a certain fault source affects other monitoring indicators and eventually leads to the occurrence of the fault. According to the fault propagation path atlas, different suppression regulation methods are adopted according to the type of the source indicator and the structure of the fault propagation path. The source indicator is the key factor leading to the fault propagation, and suppressing the fluctuations of these source indicators can effectively slow down or avoid the expansion of the fault.

[0028] Further, the method comprises the following steps:

[0029] Based on the fault time series tracking dataset, monitoring indicator fault fluctuation correlation analysis is performed, and a plurality of associated indicator groups of a plurality of fault correlation scenarios are output. The plurality of associated indicator groups are subjected to set intersection operation to obtain an intersection indicator group. The fluctuation coupling coefficients of the intersection indicator group and the plurality of associated indicator groups are quantified as a plurality of scenario branch weights. The intersection indicator group is taken as a root node, and the plurality of associated indicator groups are taken as a plurality of child nodes. Based on the plurality of scenario branch weights, a weighted connection edge of the root node and the plurality of child nodes is established to complete the construction of the fault fluctuation correlation tree.

[0030] Based on the fault time series tracking dataset, fault fluctuation correlation analysis is performed on each monitoring indicator. The goal is to analyze the change trend and fluctuation mode of each monitoring indicator and identify whether there is a certain temporal correlation or mutual influence between different monitoring indicators. For example, whether the fluctuation of a certain monitoring indicator will trigger the fluctuation of another monitoring indicator. The analysis methods include statistical correlation analysis, time series analysis (such as correlation calculation, covariance, mutual information, etc.), and association rule mining, etc. to find out the fluctuation relationship of each monitoring indicator. Through correlation analysis, a plurality of fault correlation scenarios are identified, each of which represents a potential fault type or system state. These fault correlation scenarios are based on the combination of different indicators and their fluctuation modes, and correspond to a plurality of fault correlation scenarios. A plurality of associated indicator groups are output, which include the monitoring indicators corresponding to the fault correlation scenarios. These monitoring indicators show obvious fluctuation characteristics when the fault occurs.

[0031] The set intersection operation is performed on multiple correlation indicator groups. This process is to calculate the intersection of the correlation indicator set in each fault correlation scenario, and obtain the intersection indicator group shared by all scenarios. For example, assuming that there are three correlation indicator groups, which correspond to three fault scenarios respectively. In each correlation indicator group, multiple relevant monitoring indicators are included. By performing the intersection operation on these correlation indicator groups, the intersection indicator group obtained is the indicator that appears in all scenarios. The intersection indicator group reflects the common monitoring features in different fault scenarios, and is an important input in subsequent risk prediction and model training.

[0032] The fluctuation coupling coefficient between the intersection indicator group and each correlation indicator group is quantified, which can quantify the correlation strength between these intersection indicator groups. The fluctuation coupling coefficient refers to the correlation degree of the fluctuation between two intersection indicator groups. For example, if the fluctuation of the intersection indicator group and the fluctuation of a certain correlation indicator group are very synchronous, then their fluctuation coupling coefficient will be higher. Conversely, if the fluctuations between them are not very relevant, the fluctuation coupling coefficient will be lower. Quantifying the fluctuation coupling coefficient can use statistical methods such as Pearson correlation coefficient, mutual information, and covariance to capture the complex relationship between indicators. The calculated fluctuation coupling coefficient is used as the weight of multiple scenario branches. The larger the fluctuation coupling coefficient, the stronger the correlation between the scenario and the intersection indicator group, and the more important the scenario is to the overall prediction. Therefore, the weight of the scenario should be increased accordingly.

[0033] The intersection indicator group is used as the root node because it represents the core monitoring indicators shared by all scenarios and embodies the main failure modes of the system. Multiple correlation indicator groups are used as multiple child nodes, each of which represents a specific fault scenario and its associated monitoring indicators. The connection between the root node and each child node is established through a weighted edge. The weight of the weighted edge is set according to the scenario branch weight calculated in the previous step. The scenario with a larger weight has a larger connection edge weight, indicating that the scenario has a greater impact on the overall system. The final fault fluctuation correlation tree presents a hierarchical structure, which clearly shows the dependency relationship between monitoring indicators and the fault propagation path.

[0034] Further, according to the node level linkage attribute of the fault fluctuation correlation tree, a cross-level risk prediction network is constructed, and the method comprises:

[0035] The plurality of fault correlation scenarios are taken as screening labels, and a plurality of fault correlation tracking records are extracted from the fault time sequence tracking dataset; a plurality of fault risk prediction models are constructed based on the plurality of fault correlation tracking records; the intersection indicator group is taken as an information elimination rule, and non-intersection indicator column elimination is performed on the fault time sequence tracking dataset to obtain an intersection indicator monitoring time sequence array; the intersection indicator monitoring time sequence array is taken as training data, and the benchmark risk prediction model is constructed; and the benchmark risk prediction model and the plurality of fault risk prediction models are integrated according to the node level linkage of the fault fluctuation correlation tree to obtain the cross-level risk prediction network.

[0036] The plurality of fault correlation scenarios are taken as screening labels, that is, the basis for screening data, and a plurality of corresponding fault correlation tracking records are extracted from the fault time sequence tracking dataset according to the screening labels. These fault correlation tracking records are data segments reflecting specific fault modes in historical data and contain time sequence trends, fluctuation modes, abnormal values and the like.

[0037] The plurality of fault correlation tracking records are used to train a fault risk prediction model for each fault correlation scenario. These models can be constructed using convolutional neural networks, long short-term memory networks, multilayer perceptrons and the like. Different features are used to train corresponding fault risk prediction models for different fault correlation scenarios. This helps to capture different types of faults and risks and provides more accurate predictions in different fault correlation scenarios.

[0038] The intersection indicator group is a set of core monitoring indicators shared by the plurality of fault correlation scenarios. The intersection indicator group is taken as an information elimination rule to eliminate those non-core monitoring indicators unrelated to the intersection indicators from the fault time sequence tracking dataset. Only the intersection indicators that have a direct impact on fault risk prediction are retained. The dataset after elimination of non-intersection indicators is referred to as an intersection indicator monitoring time sequence array. This array only contains time sequence data of key indicators and is used for subsequent model training and prediction.

[0039] The intersection indicator monitoring time sequence array is taken as training data to construct a benchmark risk prediction model. This model can also be constructed using convolutional neural networks, long short-term memory networks, multilayer perceptrons and the like. The goal is to train using the time sequence characteristics of the intersection indicators to predict the occurrence probability or risk level of different fault correlation scenarios.

[0040] The node level linkage of the fault fluctuation correlation tree means that the relationship between nodes is hierarchical, and information flow and hierarchical decision can be performed from the root node to the child node in turn. Based on the node level structure of the fault fluctuation correlation tree, the training and output of the model are combined to integrate the benchmark risk prediction model and multiple fault risk prediction models. Through this integration mode, the cross-level risk prediction network obtained can perform fault prediction at different levels. The benchmark risk prediction model provides the overall fault prediction probability, and the multiple fault risk prediction models further refine and specify the prediction results of different fault correlation scenarios.

[0041] Further, the intersection index monitoring time series array is taken as training data to construct the benchmark risk prediction model, and the method comprises:

[0042] Based on the fault node, the intersection index monitoring time series array is subjected to sliding window feature extraction to generate multiple backtracking time series feature vectors of multiple historical fault types. A pre-constructed risk probability distribution prediction double-channel architecture is generated, wherein the risk probability distribution prediction double-channel architecture comprises a 1D-CNN branch and an LSTM branch in parallel. The multiple historical fault types and multiple backtracking time series feature vectors are taken as training data to perform weighted multi-scene probability distribution training on the risk probability distribution prediction double-channel architecture. The risk probability distribution prediction double-channel architecture after training is subjected to gradient quantization compression to generate a lightweight benchmark risk prediction model.

[0043] The sliding window is a time series processing technology that extracts features step by step along the time axis by applying a fixed-size window on time series data. The size of the sliding window depends on the duration of the fault mode and the length of the historical information that is expected to be captured. For each fault node, the corresponding features are extracted from the intersection index monitoring time series array using the sliding window to generate multiple backtracking time series feature vectors of multiple historical fault types. Historical fault types refer to different fault types that have occurred in the past. Each fault type has different time series features. The backtracking time series feature vector is composed of fluctuation intensity features, trend steepness features, and cross-index coupling features, which are used to describe the fluctuation pattern and trend of historical faults.

[0044] The 1D-CNN (1D convolutional neural network) branch is used to process time series data, which can effectively capture local fluctuation patterns. Through convolution operations, CNN can identify local patterns in time series, such as rapid fluctuations, mutations, or other short-term trend changes. In fault prediction, local fluctuation patterns are used to capture sudden failures and short-term fluctuations. Therefore, 1D-CNN is used to specifically capture these details. The LSTM (Long Short-Term Memory) branch focuses on long-term dependencies in time series data. LSTM can retain memory over a long period of time and capture long-term trends and changes, so it is suitable for analyzing long-term trends of fault patterns over time. Through LSTM, the long-term evolution of faults can be modeled, and the potential risks of faults can be predicted, especially complex and nonlinear relationships. In the risk probability distribution prediction dual-channel architecture, the 1D-CNN branch and the LSTM branch run in parallel, respectively processing local fluctuation patterns and long-term trend dependencies. In this way, time series data can be analyzed from multiple dimensions and fault risks can be predicted comprehensively.

[0045] The risk probability distribution prediction dual-channel architecture is trained to predict the probability distribution of fault occurrence based on different types of historical fault data. During training, different weights are assigned to different fault correlation scenarios. The weight values can be adjusted according to the severity, frequency of occurrence, or other key factors of the scenario. This weighting method helps the model pay more attention to certain high-risk or important fault scenarios, improving prediction accuracy. Through weighted training, the model can simultaneously predict multiple fault correlation scenarios and optimize them according to their weights. Through training, the model can output a probability distribution representing the probability of occurrence of different fault scenarios.

[0046] Gradient quantization compression is a model compression technique that aims to reduce the computational complexity and memory usage of the model while minimizing the loss of prediction performance. The specific approach is to reduce the size of the model by quantizing the gradient (i.e., using lower precision data representation). This helps to deploy the model in situations where hardware resources are limited. After the compression process is complete, the resulting lightweight baseline risk prediction model has smaller size and lower computational requirements compared to the original model, making it suitable for real-time monitoring and fast prediction scenarios.

[0047] Further, based on the plurality of fault correlation tracking records, a plurality of fault risk prediction models are constructed, and the method comprises:

[0048] The first fault correlation tracking record is decomposed based on the fault level attribute to obtain a plurality of sample fault levels and a plurality of sample correlation fluctuation feature vector groups, wherein the sample correlation fluctuation feature vector group is composed of a plurality of sample correlation fluctuation feature vectors of a plurality of correlation indicators in the first correlation indicator group; the plurality of sample fault levels and the plurality of sample correlation fluctuation feature vector groups are input as training data into a standard CNN risk prediction model, hierarchical parameter adjustment training is performed, and a first fault risk prediction model is generated.

[0049] The fault level attribute represents the severity of the fault, such as a minor fault, a moderate fault, a serious fault, etc., the first fault correlation tracking record is any one of a plurality of fault correlation tracking records as a current analysis object, the first fault correlation tracking record contains detailed time series data of the fault event, by decomposing these first fault correlation tracking records, different fault level information can be extracted according to the fault occurrence and influence, the sample fault level refers to the fault severity corresponding to each fault, and the sample correlation fluctuation feature vector group refers to a plurality of sample correlation fluctuation feature vectors extracted from the first fault correlation tracking record based on a plurality of correlation indicators, which describe the fluctuation patterns of various indicators within a certain time window.

[0050] The plurality of sample fault levels and the plurality of sample correlation fluctuation feature vector groups are combined as training data, wherein the input features are the plurality of sample correlation fluctuation feature vector groups, and the training labels are the plurality of sample fault levels. A standard CNN (Convolutional Neural Network) is used for training, CNN is used to automatically extract local features in data in time series prediction, and is particularly suitable for processing time series data with local fluctuation features. CNN can effectively learn local fluctuation patterns, and the abnormality of local fluctuation is an important signal for determining faults. In the training process, the model is optimized through hierarchical parameter adjustment, including adjusting network layers, convolution kernel size and other hyperparameters to optimize the performance of the model, adjusting learning rate, optimizer and other training parameters to ensure that the model can effectively converge and avoid overfitting, designing different loss functions or weights for different fault levels to enable the model to better learn the prediction of serious faults or the sensitivity adjustment of minor faults. Through this series of training process, the first fault risk prediction model is obtained, which can predict the sample fault level according to the input sample correlation fluctuation feature vector group.

[0051] Further, the backtracking time sequence feature vector is composed of fluctuation intensity features, trend steepness features and cross-indicator coupling features.

[0052] The fluctuation intensity feature is used to measure the amplitude of fluctuation in time series data, which can be calculated by standard deviation, variance or other volatility indicators. Fluctuation intensity is an important signal in failure prediction, especially in the early stage of failure, where fluctuation usually changes significantly. The trend steepness feature is used to measure the speed of trend change. Failure is often accompanied by a dramatic change in system performance, so the steepness of the trend change indicates potential failure risk. Multiple monitoring indicators may have mutual dependence or resonance effect, and the cross-indicator coupling feature can capture these correlations through methods such as covariance and mutual information, revealing the complex relationship between different indicators. Through these feature extraction methods, local fluctuations and long-term trends in time series can be fully displayed, and different types of failure risks can be effectively captured.

[0053] Further, after inputting the intersection indicator monitoring time series array into the baseline risk prediction model of the cross-level risk prediction network, according to the scene matching probability distribution output by the baseline risk prediction model, locally activate multiple failure risk prediction models in parallel in the cross-level risk prediction network to perform risk prediction level verification, and output an integrated risk prediction result, the method comprises:

[0054] The intersection indicator monitoring time series array is subjected to time series fluctuation feature extraction to obtain a real-time associated fluctuation feature vector group; the real-time associated fluctuation feature vector group is loaded into the baseline risk prediction model for scene probability reasoning calculation, and a scene matching probability distribution is output; according to the scene matching probability distribution and multiple scene branch weights, dynamic computing power distribution of the multiple failure risk prediction models is performed; according to the start and end time stamps of the intersection indicator monitoring time series array, time series data of the multiple associated indicator groups are backtracked and aligned to obtain multiple associated indicator monitoring time series arrays; the multiple associated indicator monitoring time series arrays are mapped and loaded into the multiple failure risk prediction models, and risk level prediction is performed in parallel to output multiple failure risk level prediction results; the scene matching probability distribution and the multiple failure risk level prediction results are integrated as the integrated risk prediction result output.

[0055] The intersection indicator monitoring time series array is processed to extract time series fluctuation features therefrom. The time series fluctuation features refer to dynamic information reflecting changes in indicators in time series, including fluctuation intensity, trend steepness, periodicity and extreme fluctuation. The extracted real-time associated fluctuation feature vector group contains time series fluctuation features of each monitoring indicator within a specific time window. These features not only involve the fluctuation pattern of a single monitoring indicator, but also include cross-indicator coupling relationships between multiple associated indicators.

[0056] The extracted real-time correlation fluctuation feature vector set is loaded into the pre-constructed benchmark risk prediction model as input. The scene probability inference calculation refers to the model calculating based on the input features to infer the probability distribution of the current scene belonging to a certain preset fault scene. This probability distribution reflects the possibility of each fault scene occurring and is the basis for the next step of fault risk prediction. The output scene matching probability distribution is a probability vector, where each value corresponds to the probability of a specific fault scene occurring, providing the confidence level of the current system state belonging to each fault scene.

[0057] According to the scene matching probability distribution and the multiple scene branch weights, dynamic computing power allocation is performed on the multiple fault risk prediction models. This means that for high-probability fault scenes, more computing resources and model complexity are allocated, such as using multiple parallel high-precision models for prediction; for low-risk scenes, less computing resources are allocated, and lightweight models are used for prediction. This dynamic computing power allocation ensures efficient use of computing resources while providing more prediction guarantees in high-risk situations.

[0058] The intersection index monitoring time series array is obtained from the intersection index time series data array. Each time series data has its collection timestamp, which records the specific time of the data. The backtracking alignment process is to ensure that all correlation index data can be aligned on the time axis, especially when the data comes from different sensors or different time windows. Specifically, according to the collection start and end timestamps of the intersection index monitoring time series array, the time series data of other related indicators are backtracked and aligned to ensure that all data are compared on the same time dimension. If the time collection intervals of some correlation indicators are inconsistent, they need to be aligned in time windows to ensure that the data of each indicator are compared within the same time range. After backtracking and alignment, multiple correlation index monitoring time series arrays are obtained, which contain the monitoring data of all correlation indicators within the same time period.

[0059] The obtained multiple correlation index monitoring time series arrays are input into the pre-trained multiple fault risk prediction models. Each fault risk prediction model independently calculates the input correlation index monitoring time series array. Through parallel execution, multiple fault prediction tasks can be processed simultaneously without relying on serial calculation. The output of these models is the fault risk level prediction result for each fault scene, i.e., the probability distribution of each fault type, and the corresponding risk level is given.

[0060] The scene matching probability distribution and the plurality of fault risk level prediction results are integrated, for example, by a weighted manner, the scene matching probability distribution and the fault risk level prediction results are combined, the risk of each fault scene is comprehensively evaluated, and the basis for weighting includes probability values of different scenes, accuracies of different fault models, and experience weights of historical data, etc. Finally, the integrated risk prediction result provides a more accurate risk evaluation, reflects the fault risk state of the whole system, can provide support for the decision system, and then triggers the corresponding preventive measures or repair scheme.

[0061] Further, based on the fault time sequence tracking dataset, monitoring index fault fluctuation correlation analysis is performed, and a plurality of correlation index groups of a plurality of fault correlation scenes are output. The method comprises:

[0062] Based on fault scene type decomposition of the fault time sequence tracking dataset, a plurality of scene-specific time sequence subsets corresponding to the plurality of fault correlation scenes are obtained; P index safety fluctuation intervals of P monitoring indexes in the global monitoring indexes are interactively obtained; the P index safety fluctuation intervals are used to traverse the plurality of scene-specific time sequence subsets, and a plurality of abnormal fluctuation event sets are screened; the recurrence frequency of the P monitoring indexes is counted in the plurality of abnormal fluctuation event sets, and a plurality of groups of fault index recurrence frequencies are obtained; based on a preset correlation frequency threshold, the plurality of groups of fault index recurrence frequencies are screened, and the plurality of correlation index groups are output.

[0063] Each fault scene type represents a specific fault or fault mode, such as voltage fluctuation, temperature anomaly, mechanical wear, etc. According to different fault scene types, the fault time sequence tracking dataset is decomposed, so as to cut the fault time sequence tracking dataset into a plurality of scene-specific time sequence subsets. Each scene-specific time sequence subset contains time sequence data in the corresponding scene. For example, if a fault scene type is caused by high temperature, the corresponding scene-specific time sequence subset only contains data of temperature-related monitoring indexes in the high temperature fault scene.

[0064] The P monitoring indexes are key indexes selected from the global monitoring indexes. These indexes usually have strong correlation with the fault mode of the system. P is a positive integer. By analyzing historical data and fault events, an index safety fluctuation interval is defined for each selected monitoring index. The interval includes the upper and lower limits of the normal fluctuation of the monitoring index, indicating the expected fluctuation range of the monitoring index in the normal working state.

[0065] For each monitoring indicator in the scenario-specific time subset, check if the data is outside the defined indicator safety fluctuation range. If the value of a certain monitoring indicator is outside its indicator safety fluctuation range, it is marked as an abnormal fluctuation event, which indicates that the monitoring indicator has fluctuated at certain times, which is inconsistent with the normal operation state, and usually indicates the existence of a fault risk or unstable factors in the system.

[0066] For all sets of multiple abnormal fluctuation events marked, count the recurrence frequency of each monitoring indicator in all abnormal fluctuation events, i.e., whether a certain monitoring indicator frequently appears in multiple abnormal events or it is associated with the abnormal fluctuation of other indicators. The recurrence frequency count can be obtained by analyzing the abnormal fluctuation events in each scenario subset. For example, if temperature and current often fluctuate abnormally together, the recurrence frequency of these two indicators is high. The resulting multiple sets of fault indicator recurrence frequency record the abnormal fluctuation recurrence frequency of all monitoring indicators, and a high fault indicator recurrence frequency means that the probability of its failure is high.

[0067] According to historical data analysis or expert experience, set an association frequency threshold. Only indicators with a fault indicator recurrence frequency greater than the association frequency threshold are considered to have sufficient relevance in the fault scenario and are worthy of being used as input for subsequent risk prediction models. Select indicators with a fault indicator recurrence frequency greater than the association frequency threshold to form multiple association indicator groups. These association indicator groups represent which monitoring indicators have a higher occurrence frequency when fluctuating abnormally in multiple fault scenarios and have strong relevance. Therefore, they will be used as key indicators for further analysis and prediction.

[0068] Further, the dynamic parameter hierarchical suppression of the target controller is performed according to a source indicator type of the fault propagation path graph.

[0069] The source indicator type includes a current-type source indicator, a voltage-type source indicator, a mechanical-type source indicator, and a composite-type source indicator. The current-type source indicator can cause device failure due to current over-limiting or fluctuation. By limiting the current value within a safe range through amplitude adjustment, the current-type source indicator can avoid device failure caused by excessively high current. The formula is: wherein, is the current value, is the suppressed current value, k is the conduction intensity, which represents the linkage relationship between current and other factors in the fault propagation path, and 0.2 is the suppression proportion coefficient, which is adjusted according to system design.

[0070] The voltage-type source indicator can cause component damage or circuit failure due to excessively large voltage fluctuation. By limiting the voltage change rate, the voltage-type source indicator can ensure that the voltage is within the allowed gradient range and avoid failure caused by excessively fast voltage fluctuation. The formula is: wherein, is the rate of voltage change, is the maximum threshold of voltage change, which decreases with the increase of risk level, that is, the higher the risk level, the smaller the allowable rate of voltage change.

[0071] The mechanical source index is represented by excessive mechanical vibration or pulse, which can accelerate the wear of mechanical equipment and even cause equipment failure. By adjusting the attenuation coefficient of the pulse and slowing down the impact of mechanical vibration, the formula is: wherein, is the frequency of the suppressed mechanical vibration, is the current mechanical vibration frequency, and k is the conduction intensity.

[0072] The composite source index involves multiple monitoring indexes, usually the combined effect of multiple physical or electrical quantities. In this case, the suppression of a single index is not enough to effectively control the risk, so multiple related indexes need to be adjusted simultaneously through a cross-domain collaborative suppression matrix. The cross-domain collaborative suppression matrix is a multi-dimensional matrix, each dimension representing a source index. By collaboratively adjusting these indexes, a comprehensive suppression strategy is formed to reduce the probability of system failure.

[0073] In summary, the intelligent integrated service method for the anti-overflow and anti-static controller provided by the embodiments of the present application has the following technical effects:

[0074] Through historical sample data retrieval and index fluctuation correlation analysis, key monitoring index fluctuation patterns can be extracted based on fault timing tracking data sets, which enables the system to identify potential fault risks in advance and determine the trigger factors and propagation paths of different fault types through the fault fluctuation correlation tree, providing more accurate early warning information for equipment maintenance personnel. Based on the node level linkage properties of the fault fluctuation correlation tree, a cross-level risk prediction network is constructed, which can integrate monitoring indexes and fault prediction models at different levels to form a global risk assessment framework. Through this integration, multiple complex fault scenarios can be considered simultaneously, and the interactive effects of different fault factors can be comprehensively evaluated to achieve cross-level and cross-domain fault prediction and control. By locally activating multiple fault risk prediction models through scenario matching probability distribution, different risk levels can be dynamically verified to ensure the reliability and accuracy of the prediction results. Through the fault propagation path map of the matching integrated risk prediction results, the source index type can be accurately identified, and customized dynamic parameter grading suppression can be implemented to effectively prevent further expansion of the fault. This adaptive control strategy can dynamically optimize the operating state of the controller based on real-time data and risk assessment, improving the overall reliability and safety of the equipment.

[0075] Embodiment two, based on the same inventive concept as the intelligent integrated service method for the anti-overflow and anti-static controller in the preceding embodiments, asFigure 2 As shown, the embodiments of the present application provide an intelligent integrated service system for anti-overflow and anti-static controller, which comprises:

[0076] The tracking data calling module 10 is configured to call a fault timing tracking data set of a historical sample controller according to a model ID of a target controller, wherein the target controller is an anti-overflow and anti-static controller; the fluctuation correlation analysis module 20 is configured to perform index fluctuation correlation analysis based on the fault timing tracking data set and construct a fault fluctuation correlation tree; the prediction network construction module 30 is configured to construct a cross-level risk prediction network according to a node level linkage attribute of the fault fluctuation correlation tree; the polling backtracking module 40 is configured to perform associated sensor device polling backtracking according to a root node monitoring index composition of the fault fluctuation correlation tree to obtain an intersection index monitoring timing array; the level checking module 50 is configured to input the intersection index monitoring timing array into a benchmark risk prediction model of the cross-level risk prediction network, and perform risk prediction level checking on multiple fault risk prediction models in parallel in the cross-level risk prediction network according to a scene matching probability distribution output by the benchmark risk prediction model to output an integrated risk prediction result; and the suppression and regulation module 60 is configured to match a fault propagation path atlas of the integrated risk prediction result to perform source parameter suppression and regulation of the target controller.

[0077] Further, the fluctuation correlation analysis module 20 is configured to perform the following operation steps:

[0078] Perform monitoring index fault fluctuation correlation analysis based on the fault timing tracking data set, and output multiple associated index groups of multiple fault correlation scenes; perform set intersection operation on the multiple associated index groups to obtain an intersection index group; quantify fluctuation coupling coefficients of the intersection index group and the multiple associated index groups as multiple scene branch weights; take the intersection index group as a root node, take the multiple associated index groups as multiple child nodes, establish weighted connection edges of the root node and the multiple child nodes based on the multiple scene branch weights, and complete construction of the fault fluctuation correlation tree.

[0079] Further, the prediction network construction module 30 is configured to perform the following operation steps:

[0080] extract a plurality of fault-associated tracking records from the fault time series tracking dataset as the plurality of fault-associated scenarios are screening labels; construct a plurality of fault risk prediction models based on the plurality of fault-associated tracking records; perform non-intersection index column elimination on the fault time series tracking dataset based on the intersection indicator group as information elimination rules to obtain an intersection indicator monitoring time series array; construct the baseline risk prediction model by taking the intersection indicator monitoring time series array as training data; and integrate the baseline risk prediction model and the plurality of fault risk prediction models according to the node level linkage of the fault fluctuation-associated tree to obtain the cross-level risk prediction network.

[0081] Further, the prediction network construction module 30 is configured to perform the following operation steps:

[0082] extract a plurality of historical fault type backtracking time series feature vectors based on the fault node and the intersection indicator monitoring time series array by sliding window feature extraction; pre-construct a risk probability distribution prediction double-channel architecture, wherein the risk probability distribution prediction double-channel architecture includes a 1D-CNN branch and an LSTM branch in parallel; perform weighted multi-scene probability distribution training on the risk probability distribution prediction double-channel architecture by taking the plurality of historical fault types and the plurality of backtracking time series feature vectors as training data; and generate the lightweight baseline risk prediction model by gradient quantization compression training of the risk probability distribution prediction double-channel architecture.

[0083] Further, the prediction network construction module 30 is configured to perform the following operation steps:

[0084] decompose the first fault-associated tracking record based on the fault level attribute to obtain a plurality of sample fault levels and a plurality of sample associated fluctuation feature vector groups, wherein the sample associated fluctuation feature vector group is composed of a plurality of sample associated fluctuation feature vectors of a plurality of associated indicators in the first associated indicator group; and input the plurality of sample fault levels and the plurality of sample associated fluctuation feature vector groups as training data into a standard CNN risk prediction model to perform hierarchical parameter tuning training to generate a first fault risk prediction model.

[0085] Further, the backtracking time series feature vector is composed of a fluctuation intensity feature, a trend steepness feature, and a cross-indicator coupling feature.

[0086] Further, the level verification module 50 is configured to perform the following operation steps:

[0087] The intersection index monitoring time series array is subjected to time series fluctuation feature extraction to obtain a real-time correlation fluctuation feature vector group; the real-time correlation fluctuation feature vector group is loaded to the benchmark risk prediction model to perform scene probability reasoning calculation, and a scene matching probability distribution is output; according to the scene matching probability distribution and a plurality of scene branch weights, dynamic computing power distribution of the plurality of fault risk prediction models is performed; according to the start and end time stamps of the intersection index monitoring time series array, time series data backtracking alignment of the plurality of correlation index groups is performed to obtain a plurality of correlation index monitoring time series arrays; the plurality of correlation index monitoring time series arrays are mapped and loaded to the plurality of fault risk prediction models, and risk level prediction is performed in parallel to output a plurality of fault risk level prediction results; the scene matching probability distribution and the plurality of fault risk level prediction results are integrated as the integrated risk prediction result output.

[0088] Further, the fluctuation correlation analysis module 20 is configured to perform the following operation steps:

[0089] The fault time series tracking data set is decomposed based on a fault scene type to obtain a plurality of scene-specific time series subsets corresponding to the plurality of fault correlation scenes; P index safety fluctuation intervals of P monitoring indexes in the global monitoring indexes are obtained interactively; the P index safety fluctuation intervals are used to traverse the plurality of scene-specific time series subsets to filter a plurality of abnormal fluctuation event sets; the P monitoring indexes are subjected to recurrence frequency counting in the plurality of abnormal fluctuation event sets to obtain a plurality of sets of fault index recurrence frequencies; the plurality of sets of fault index recurrence frequencies are filtered based on a preset correlation frequency threshold to output the plurality of correlation index groups.

[0090] Further, the dynamic parameter hierarchical suppression of the target controller is performed according to a source index type of the fault propagation path graph.

[0091] The foregoing detailed description of the intelligent integrated service method for the anti-overflow and anti-static controller enables those skilled in the art to clearly understand the intelligent integrated service system for the anti-overflow and anti-static controller in the embodiments. Since the system corresponds to the disclosed method, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0092] The foregoing description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent integrated services oriented to an anti-overflow and anti-static controller, characterized in that, The method comprises: According to the model ID of the target controller, the fault time sequence tracking data set of the historical sample controller is called, wherein the target controller is an anti-overflow and anti-static controller; Based on the fault time sequence tracking data set, index fluctuation correlation analysis is carried out, and a fault fluctuation correlation tree is constructed; According to the node level linkage attribute of the fault fluctuation correlation tree, a cross-level risk prediction network is constructed; According to the root node monitoring index composition of the fault fluctuation correlation tree, the associated sensing device polling backtracking is carried out, and an intersection index monitoring time sequence array is obtained; After inputting the intersection index monitoring time sequence array into the benchmark risk prediction model of the cross-level risk prediction network, according to the scene matching probability distribution output by the benchmark risk prediction model, the cross-level risk prediction network is locally activated, and a plurality of fault risk prediction models in parallel are executed to perform risk prediction level verification, and an integrated risk prediction result is output; The fault propagation path atlas matching the integrated risk prediction result is matched, and the source parameter suppression regulation of the target controller is carried out; Based on the fault time sequence tracking data set, index fluctuation correlation analysis is carried out, and a fault fluctuation correlation tree is constructed, the method comprising: Based on the fault time sequence tracking data set, monitoring index fault fluctuation correlation analysis is carried out, and a plurality of associated index groups of a plurality of fault correlation scenes are output; The plurality of associated index groups are subjected to set intersection operation to obtain an intersection index group; Quantify the fluctuation coupling coefficients of the intersection index group and the plurality of associated index groups as the scene branch weights; The intersection index group is taken as a root node, and the plurality of associated index groups are taken as a plurality of child nodes, a weighted connection edge of the root node and the plurality of child nodes is established based on the plurality of scene branch weights, and the construction of the fault fluctuation correlation tree is completed; According to the node level linkage attribute of the fault fluctuation correlation tree, a cross-level risk prediction network is constructed, the method comprising: The plurality of fault correlation scenes are taken as screening labels, and a plurality of fault correlation tracking records are extracted from the fault time sequence tracking data set; Based on the plurality of fault correlation tracking records, a plurality of fault risk prediction models are constructed; The intersection index group is taken as an information elimination rule, and non-intersection index column elimination is carried out on the fault time sequence tracking data set to obtain an intersection index monitoring time sequence array; The intersection index monitoring time sequence array is taken as training data, and the benchmark risk prediction model is constructed; According to the node level linkage of the fault fluctuation correlation tree, the benchmark risk prediction model and the plurality of fault risk prediction models are integrated to obtain the cross-level risk prediction network.

2. The intelligent integrated service method for the anti-overflow and anti-static controller according to claim 1, wherein, The intersection index monitoring time sequence array is taken as training data, and the benchmark risk prediction model is constructed, the method comprising: Based on the fault node, the intersection index monitoring time sequence array is subjected to sliding window feature extraction to generate a plurality of backtracking time sequence feature vectors of a plurality of historical fault types; A pre-constructed risk probability distribution prediction double-channel architecture is constructed, wherein the risk probability distribution prediction double-channel architecture comprises a 1D-CNN branch and an LSTM branch in parallel; The plurality of historical fault types and the plurality of backtracking time sequence feature vectors are taken as training data to perform weighted multi-scene probability distribution training on the risk probability distribution prediction double-channel architecture; The trained risk probability distribution prediction double-channel architecture is compressed by gradient quantization to generate a lightweight benchmark risk prediction model.

3. The intelligent integrated service method for anti-overflow and anti-static controller of claim 1, wherein, Based on the plurality of fault correlation tracking records, a plurality of fault risk prediction models are constructed, and the method comprises: Based on the fault level attribute, the first fault correlation tracking record is decomposed to obtain a plurality of sample fault levels and a plurality of sample correlation fluctuation feature vector groups, wherein the sample correlation fluctuation feature vector group is composed of a plurality of sample correlation fluctuation feature vectors of a plurality of correlation indicators in the first correlation indicator group; The plurality of sample fault levels and the plurality of sample correlation fluctuation feature vector groups are taken as training data to input a standard CNN risk prediction model to perform hierarchical parameter tuning training to generate a first fault risk prediction model.

4. The intelligent integrated service method for anti-overflow and anti-static controller of claim 2, wherein, The backtracking time sequence feature vector is composed of fluctuation intensity features, trend steepness features, and cross-indicator coupling features.

5. The intelligent integrated service method for anti-overflow and anti-static controller of claim 3, wherein, After the intersection indicator monitoring time sequence array is input into the benchmark risk prediction model of the cross-level risk prediction network, according to the scene matching probability distribution output by the benchmark risk prediction model, the plurality of fault risk prediction models in parallel in the cross-level risk prediction network are locally activated to perform risk prediction level verification, and an integrated risk prediction result is output, and the method comprises: The intersection indicator monitoring time sequence array is subjected to time sequence fluctuation feature extraction to obtain a real-time correlation fluctuation feature vector group; The real-time correlation fluctuation feature vector group is loaded into the benchmark risk prediction model for scene probability reasoning calculation to output the scene matching probability distribution; According to the scene matching probability distribution and a plurality of scene branch weights, dynamic computing power distribution of the plurality of fault risk prediction models is performed; According to the collection start and end time stamps of the intersection indicator monitoring time sequence array, time sequence data backtracking alignment of the plurality of correlation indicator groups is performed to obtain a plurality of correlation indicator monitoring time sequence arrays; The plurality of correlation indicator monitoring time sequence arrays are mapped and loaded into the plurality of fault risk prediction models to perform risk level prediction in parallel to output a plurality of fault risk level prediction results; The scene matching probability distribution and the plurality of fault risk level prediction results are integrated as the integrated risk prediction result output.

6. The intelligent integrated service method for anti-overflow and anti-static controller of claim 1, wherein, Based on the fault time sequence tracking data set, monitoring indicator fault fluctuation correlation analysis is performed to output a plurality of correlation indicator groups of a plurality of fault correlation scenes, and the method comprises: Based on fault scene type decomposition, the fault time sequence tracking data set is decomposed to obtain a plurality of scene-specific time sequence subsets corresponding to the plurality of fault correlation scenes; P indicator safety fluctuation intervals of P monitoring indicators in the global monitoring indicators are interactively obtained; The P indicator safety fluctuation intervals are used to traverse the plurality of scene-specific time sequence subsets to filter a plurality of abnormal fluctuation event sets; In the plurality of abnormal fluctuation event sets, the recurrence frequency of the P monitoring indicators is counted to obtain a plurality of fault indicator recurrence frequencies. Filter the recurrence frequencies of the plurality of groups of fault indicators based on a preset correlation frequency threshold, and output the plurality of groups of correlation indicators.

7. The intelligent integrated service method for anti-overflow and anti-static controller of claim 1, wherein, Perform dynamic parameter hierarchical suppression of the target controller according to the source indicator type of the fault propagation path graph.

8. A smart integrated service system for an anti-overflow and anti-static controller, characterized in that, The intelligent integrated service method for implementing the anti-overflow and anti-static controller according to any one of claims 1-7, the system comprises: A tracking data calling module is configured to call a fault timing tracking data set of a historical sample controller according to a model ID of a target controller, wherein the target controller is an anti-overflow and anti-static controller; A fluctuation correlation analysis module is configured to perform indicator fluctuation correlation analysis based on the fault timing tracking data set and construct a fault fluctuation correlation tree; A prediction network construction module is configured to construct a cross-level risk prediction network according to the node level linkage attribute of the fault fluctuation correlation tree; A polling backtracking module is configured to perform associated sensor equipment polling backtracking according to the root node monitoring indicator composition of the fault fluctuation correlation tree, and obtain an intersection indicator monitoring timing array; A level verification module is configured to input the intersection indicator monitoring timing array into a benchmark risk prediction model of the cross-level risk prediction network, and locally activate a plurality of fault risk prediction models in parallel in the cross-level risk prediction network to perform risk prediction level verification according to a scene matching probability distribution output by the benchmark risk prediction model, and output an integrated risk prediction result; An inhibition regulation module is configured to match a fault propagation path graph of the integrated risk prediction result and perform source parameter inhibition regulation of the target controller.

Images