A current transformer error dynamic monitoring method and system

Abstract

The application relates to the technical field of power system measurement, and discloses a current transformer error dynamic monitoring method and system, wherein the current transformer error dynamic monitoring method comprises the following steps: obtaining current transformer time sequence data and system event logs; constructing a time sequence causal diagram to represent the time correlation between events and error changes; applying a counter causal model to identify potential causal links; designing a multiple world simulation engine to generate counterfactual scenarios; quantifying causal effects by comparing actual observations with counterfactual simulation results; establishing a monitoring mechanism to track key trigger events in real time; generating a dynamic causal explanation report and adjusting a compensation strategy; the application breaks through the limitations of traditional correlation analysis, can accurately distinguish between causality and correlation, accurately identifies the real trigger factors of current transformer error changes, reduces the false positive rate and the false negative rate, and realizes accurate dynamic monitoring of current transformer errors.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system measurement, more particularly, it relates to a current transformer error dynamic monitoring method and system. BACKGROUND

[0002] As an important measuring device in the power system, the measurement accuracy of the current transformer directly affects the safe operation and economic benefits of the power system. In actual operation, the measurement error of the current transformer will dynamically change with the change of the system operating condition, including the fluctuation of the ratio error and the phase angle error. These error changes may be caused by a variety of system events, such as load fluctuation, system switching, short-circuit fault, lightning disturbance and other complex factors alone or in combination. The traditional current transformer error monitoring method mainly relies on static calibration and periodic detection, which is difficult to realize real-time tracking and accurate identification of dynamic changes of error.

[0003] The existing dynamic monitoring technology is mainly based on correlation analysis method, which identifies possible influencing factors by statistically analyzing the correlation between system events and error changes. However, the correlation analysis method has significant technical limitations: when facing multiple event combinations, it is difficult to distinguish the true causal relationship from only the correlation, and it is easy to misjudge the accidental time coincidence as causal association; at the same time, this method cannot effectively handle the interaction and time sequence dependence between events, resulting in insufficient accuracy in identifying the root cause of error changes under complex operating conditions. This technical defect makes the existing monitoring system often have false positives and false negatives, affecting the reliability of power system operation and maintenance.

[0004] In addition, the existing technology lacks a deep understanding of the error change mechanism and prediction ability. The traditional method mainly adopts a passive monitoring mode, which can only conduct post-analysis after the error change occurs, and cannot provide prospective early warning and preventive measures.

[0005] Under the development trend of increasing complexity and intelligentization of the power system, there is an urgent need for a current transformer error monitoring technology that can accurately identify the causal relationship of error changes, achieve accurate dynamic monitoring, and have prediction and early warning capabilities, to meet the higher requirements of modern power systems for measurement accuracy and intelligent operation and maintenance. SUMMARY

[0006] The present application provides a current transformer error dynamic monitoring method and system, which solves the technical problem that related technologies cannot accurately identify the true trigger factors of error changes when facing multiple event combinations.

[0007] The present application provides a current transformer error dynamic monitoring method, comprising the following steps:

[0008] Collect and preprocess current transformer time series measurement data and system event logs, including data cleaning, standardization and outlier treatment;

[0009] Feature extraction on preprocessed data, identify key time series features and event features;

[0010] Build a time causal graph to represent the temporal relationship between events and error changes, and calculate the initial weights;

[0011] The time causal graph is a directed graph structure, where nodes represent system events or current transformer error change points, directed edges represent possible causal relationships, and edge weights reflect the strength of causal relationships;

[0012] Apply the counter causal model to analyze the time causal graph, identify potential causal links and correlation strength;

[0013] The counter causal model includes:

[0014] Event counter module, establish event frequency statistics table, record the occurrence frequency, first occurrence time, last occurrence time and average occurrence interval of each system event in the observation period;

[0015] Error change monitoring module, establish error change record table, continuously record the occurrence time, change amplitude, change direction and change duration of current transformer error change points;

[0016] Correlation analysis module, establish event-error co-occurrence count matrix, use hierarchical time window technology to establish short-term, medium-term and long-term time windows of different scales, calculate the co-occurrence count matrix respectively and get the comprehensive correlation strength by weighted summation;

[0017] Probability reasoning module, based on the count matrix and event frequency statistics table, calculate the conditional probability and joint probability, and process the joint influence probability of event sequence on error change through Bayes theorem;

[0018] Causal link screening module, based on statistical significance test to screen potential causal links, calculate the chi-square statistic for each pair of event-error relationship, when the chi-square value is greater than the critical value, it is considered that the causal link has statistical significance;

[0019] Generate counterfactual scenarios through multiple world simulation engine, simulate the occurrence and non-occurrence of specific events, and evaluate the event impact;

[0020] Multiple world simulation engine includes state representation module, event intervention module, state transition calculation unit and uncertainty quantification unit, through event operation function to intervene specific events, create parallel state evolution of actual path and counterfactual path;

[0021] The counterfactual scenario refers to a hypothetical situation different from the actual observation by artificially intervening in the occurrence state of a specific event; the generation of the counterfactual scenario is based on the control variable method: keeping all other conditions unchanged, only changing the state of the target event, and observing the difference in system response;

[0022] The difference between the actual observation and the counterfactual simulation result is compared, and the causal effect of each event on the error change of the current transformer is quantified;

[0023] Based on the strength of the causal effect, the key trigger events affecting the error of the current transformer are monitored and identified in real time, and warning information is generated.

[0024] Further, the preprocessing includes:

[0025] The primary current value and the secondary current value of the current transformer are subjected to amplitude normalization processing;

[0026] The ratio error and the phase angle error are subjected to standardization processing;

[0027] The event types in the system event log are encoded, and the non-numeric event types are converted into numeric classification variables.

[0028] Further, the construction of the time series causal graph with a directed weighted graph structure includes:

[0029] Through the error change point identification algorithm, the original error sequence is subjected to wavelet decomposition and uses the Daubechies-4 wavelet basis function, applies the soft threshold function to shrink the wavelet coefficients of each layer, then calculates the first-order difference of the denoised error sequence and applies the double threshold judgment to identify the significant error change points;

[0030] Through the event sequence standardization algorithm, the system event log is subjected to standardization processing, and each event is assigned with the attributes of time stamp, event type code, event severity and event duration, forming a structured event sequence;

[0031] Through the graph node creation algorithm, a graph node is created for each error change point and system event, a node index table is established, and each node has the attributes of node ID, node type, time stamp and node weight;

[0032] Through the time window constraint algorithm, the maximum time lag parameter is set based on the system physical response characteristics, and the time sequence constraint and time window constraint check are performed through the directed edge creation algorithm to create directed edges that meet the constraint conditions;

[0033] Through the edge weight calculation algorithm, a multi-factor weighted model is used to calculate the weight of each directed edge, including the conditional probability term, the time decay term and the statistical correlation term;

[0034] By edge pruning algorithm, based on weight threshold pruning, statistical saliency pruning and redundant edge pruning, insignificant edges are removed to reduce the complexity of the graph.

[0035] Further, the multi-world simulation engine is composed of the following four core components:

[0036] State representation module: represents the state of the power system at time as a multi-dimensional vector;

[0037] Event intervention module: realizes intervention on specific events through event operation functions;

[0038] State transition calculation unit: used to predict the time evolution of the system state under the condition that a specific event occurs or does not occur. The system state at the next time is calculated based on the state transition function;

[0039] The state transition calculation unit includes:

[0040] Physical constraint layer: constraint condition set is constructed based on the basic laws of power system;

[0041] Historical data learning layer: learns historical state transition patterns through time series analysis algorithms;

[0042] Mixed inference layer: fuses physical constraints and historical learning results;

[0043] State verification layer: checks whether the predicted result meets the system constraint condition. If the predicted state violates the constraint condition, adjust the state value through the least square method to make it meet the constraint requirement;

[0044] Uncertainty quantification unit: evaluates the reliability of the simulation results, and provides statistical reliability guarantee for causal effect analysis; includes four calculation modules:

[0045] Monte Carlo sampling module: assesses the impact of parameter uncertainty on the results through repeated sampling;

[0046] Bayesian estimation module: calculates the posterior probability distribution of the prediction result based on Bayes theorem;

[0047] Sensitivity analysis module: identifies the input parameters that have the greatest impact on the results through partial differentiation method;

[0048] Confidence interval generator: provides confidence interval for the prediction result based on statistical distribution theory.

[0049] Further, the state representation module performs the following preprocessing on the input heterogeneous data:

[0050] Minimum maximum normalization is performed on continuous parameters to limit the value range to a specific interval;

[0051] The periodic parameters are subjected to sine-cosine transformation to avoid period jump;

[0052] The discrete parameters are subjected to one-hot encoding processing.

[0053] Further, the step of quantifying the causal effect of each event on the error change of the current transformer comprises:

[0054] Determining the analysis target error change point and its possible related event set;

[0055] Obtaining the error change prediction value under the actual scenario and counterfactual scenario through the multi-world simulation engine, and creating parallel state evolution of the actual path and counterfactual path;

[0056] Calculating the causal effect of the event on the error change through the average processing effect formula, and the formula is the difference between the error change expectation value under the actual scenario and the error change expectation value under the counterfactual scenario;

[0057] Performing robustness verification, repeatedly calculating multiple times under random initial conditions, and calculating the confidence interval of the causal effect using statistical distribution theory;

[0058] Z-Scre standardization normalization processing is performed on the causal effects of all events to convert the effect values of different scales and units into comparable unified dimensions.

[0059] Further, the step of real-time monitoring and identifying the key trigger event affecting the error of the current transformer comprises:

[0060] Establishing a causal event monitoring priority queue to focus on events with higher causal effect intensity;

[0061] Through an adaptive threshold algorithm, the causal effect intensity threshold for triggering an alarm is dynamically adjusted based on the historical statistical characteristics of the system operating state;

[0062] Through a sliding time window analysis algorithm, the size and sliding step of the time window are determined, the local causal effect value of the data in each sliding window is calculated, and the causal effect values of adjacent windows are compared to detect the dynamic changes of the causal relationship;

[0063] Through an event combination effect evaluation algorithm, a combination recognition algorithm is used to detect frequently co-occurring event combinations, an interaction effect calculation algorithm is used to quantify the synergistic or antagonistic effect between events, and an AHP algorithm is used to decompose the event combination into main effect and interaction effect.

[0064] Further, the sliding time window analysis algorithm comprises:

[0065] Determining the size and sliding step of the time window;

[0066] Calculating the local causal effect value of the data in each sliding window;

[0067] Comparing the causal effect values of adjacent windows, detecting the dynamic changes of the causal relationship;

[0068] When the change exceeds the preset threshold, triggering reevaluation.

[0069] Further, it also includes:

[0070] Generating a dynamic causal explanation report, including a causal link visualization chart of error change;

[0071] Providing error compensation suggestions for different types of events;

[0072] Building an error trend prediction model;

[0073] Building a knowledge base, accumulating causal analysis experience, and continuously optimizing compensation strategies.

[0074] The application discloses a current transformer error dynamic monitoring system, comprising:

[0075] A data acquisition and preprocessing unit is used for acquiring current transformer time series measurement data and system event logs, and performing amplitude normalization processing, standardization processing and event type encoding conversion;

[0076] A time series causal diagram construction unit is used for constructing a directed weighted graph structure, and through error change point identification algorithm, event sequence standardization algorithm, graph node creation algorithm, time window constraint algorithm, directed edge creation algorithm, edge weight calculation algorithm and edge pruning algorithm, the time series causal association between system events and error changes is represented;

[0077] A counter causal model unit is based on event frequency statistics and time series association analysis, and through an event counter module, an error change monitoring module, an association analysis module, a probability reasoning module and a causal link screening module, a potential causal relationship candidate set that passes statistical significance test is identified;

[0078] A multiple world simulation engine is based on state space search and conditional reasoning, and contains a state representation module, an event intervention module, a state transition calculation unit and an uncertainty quantification unit, and through an event operation function, a specific event is intervened to generate parallel state evolution of actual path and counterfactual path;

[0079] A causal effect quantification unit is used for calculating the causal effect of events on error changes through an average processing effect formula, performing robustness verification and calculating a confidence interval, and performing normalization processing on the causal effect;

[0080] A key trigger event monitoring unit is configured to establish a causal event monitoring priority queue, and to monitor and identify key trigger events affecting the error of the current transformer in real time through an adaptive threshold algorithm and a sliding time window analysis algorithm.

[0081] A dynamic report generation unit is configured to generate a causal link visualization chart of error changes, provide error compensation suggestions for different types of events, and build an error trend prediction algorithm and a knowledge base to accumulate causal analysis experience.

[0082] The present application has the following advantages:

[0083] The present application breaks through the limitations of traditional correlation analysis and realizes accurate identification of causal relationships. Traditional current transformer error monitoring methods mainly rely on correlation analysis, which is difficult to distinguish between true causal relationships and accidental correlations in the face of multiple event combinations. The present application introduces counterfactual analysis theory, combined with a multiple world simulation engine and a counter causal model, which can effectively distinguish between only correlation and true causality, and fundamentally solves the problem of inaccurate error change root identification.

[0084] The present application improves the accuracy and reliability of dynamic error monitoring. The present application can accurately identify key trigger events of current transformer error changes through time series causal graph construction and causal effect quantification algorithms, which greatly reduces the false positive rate and false negative rate. The multiple world simulation engine generates counterfactual scenarios to provide a reliable comparison benchmark for causal relationship verification, significantly improving the accuracy of error tracking and providing a solid technical foundation for subsequent precise compensation.

[0085] The present application realizes intelligent monitoring capability that dynamically adapts to complex operating conditions. The adaptive threshold mechanism of the present application can dynamically adjust the monitoring sensitivity according to the system operating state, the sliding time window analysis algorithm can capture the dynamic changes of causal relationships in real time, and the event combination effect evaluation model can handle multiple event interactions. These technical innovations enable the monitoring system to adapt to the complex and variable operating conditions of the power system, maintaining consistent and stable monitoring performance.

[0086] The present application improves the intelligence and forward-looking level of power system operation and maintenance. The present application not only can monitor error changes in real time, but also can provide targeted compensation suggestions through dynamic causal explanation reports and predict future error change trends through error trend prediction models. The accumulation mechanism of the knowledge base enables the system to continuously learn and optimize, realizing the transition from passive monitoring to active prevention, and significantly improving the overall reliability and operation efficiency of power system measurement. BRIEF DESCRIPTION OF DRAWINGS

[0087] Figure 1 is a flowchart of a current transformer error dynamic monitoring method of the present application;

[0088] Figure 2 is a bar chart of the timing causal graph edge weight analysis of the present application;

[0089] Figure 3 is a line chart of the short-circuit fault counterfactual scenario comparison analysis of the present application;

[0090] Figure 4 is a line chart of the short-circuit fault counterfactual scenario phase angle error comparison of the present application;

[0091] Figure 5 is a column chart of the causal effect intensity analysis of different system events of the present application;

[0092] Figure 6 is a grouped column chart of the error trigger factor identification accuracy comparison of the present application;

[0093] Figure 7 is a grouped column chart of the false positive rate and false negative rate comparison of the present application. DETAILED DESCRIPTION

[0094] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is merely meant to provide a better understanding of the subject matter described herein and can include changes, modifications, or additions of elements to the functions and arrangements of the elements discussed without departing from the scope of the present disclosure. Various examples can omit, substitute, or add various procedures or components as appropriate. Also, it should be understood that some features described with respect to one example can be combined in other examples.

[0095] In at least one embodiment of the present application, a current transformer error dynamic monitoring method is disclosed, as shown in Figure 1 includes the following steps:

[0096] Step 1, collect and pre-process current transformer timing measurement data and system event logs, including data cleaning, standardization and outlier processing;

[0097] This step collects timing measurement data during the operation of the current transformer through the monitoring device of the power system, including but not limited to: primary current value, secondary current value, ratio error, phase angle error and other parameters. At the same time, system event logs are collected, including load changes, system switching, fault occurrence and other event information that may affect the performance of the current transformer. The data obtained is synchronized according to the time stamp to ensure the timing consistency of the data analysis.

[0098] The raw data obtained needs to be pre-processed as follows:

[0099] The primary current value and the secondary current value are subjected to amplitude normalization processing to eliminate the influence of the order of magnitude difference;

[0100] The contrast error and phase angle error are standardized to convert them into a standard distribution with a mean of 0 and a standard deviation of 1.

[0101] The event types in the system event log are encoded, and non-numeric event types such as "load change" and "system switching" are converted into numeric classification variables.

[0102] The timestamp data is converted into relative time intervals to facilitate the calculation of time series correlation strength.

[0103] As shown in Figure 2 , the correlation strength between different system events and error changes is shown, and the edge weight value can be used to intuitively compare which events are more strongly associated with a specific type of error change. The edge weight of short-circuit fault and phase angle error is the highest (0.882), indicating that they have the strongest causal association.

[0104] Step 2: Feature extraction is performed on the preprocessed data to identify key time series features and event features.

[0105] Based on the obtained time series data and event log, a time series causal graph is constructed to represent the temporal association between events and error changes. The specific steps of the construction algorithm of the time series causal graph are as follows:

[0106] The current transformer error data is preprocessed by threshold detection and wavelet transform denoising to identify significant error change points , where , , represent the first , , significant error change points. represents the total number of significant error change points identified.

[0107] The threshold detection parameters include:

[0108] Error change rate threshold: when the error change rate of adjacent points exceeds, it is determined as a significant change point;

[0109] Error amplitude threshold: when the absolute value of error exceeds, it is determined as a significant change point.

[0110] The wavelet transform parameters include:

[0111] Wavelet basis function: db4 wavelet is selected;

[0112] Decomposition level: usually take 3-5 layers;

[0113] Threshold function: soft threshold function is used for coefficient shrinkage.

[0114] Standardize system event logs to form event sequences where, , , represent the first , , events respectively; represents the total number of events.

[0115] Create graph nodes for each error change point and system event;

[0116] Set the maximum time lag parameter , usually 5 times the system response time;

[0117] For each pair of events and error change points , if (i.e. the error change occurs within the valid time window after the event), create a directed edge ;

[0118] Calculate the initial weight of the edge, the weight calculation formula is:

[0119] ;

[0120] where, represents the edge weight from event to error change ; represents the frequency of the co-occurrence of event and error change (times / total observation time); represents the frequency of event occurrence (number of event occurrences / total observation time); is a time decay factor, which controls the speed of time decay, the larger the value, the faster the decay; represents the timestamp of the error change ; represents the timestamp of the event ; represents the time interval between error change and event occurrence; is a time decay function, used to exponentially decay the association strength according to the time interval, the larger the time interval, the more significant the decay; represents the conditional probability, i.e. the probability of error change occurring under the condition that event occurs.

[0121] Perform edge pruning to remove edges with weights below a threshold, reducing the complexity of the graph.

[0122] Further, the time-series causal graph construction algorithm supports a dynamic update mechanism. With the acquisition of new data, the structure and edge weights of the graph can be automatically adjusted.

[0123] Further, the time-series causal graph construction algorithm can integrate domain knowledge constraints to exclude physically impossible causal relationships.

[0124] The constructed time-series causal graph is a directed graph structure, where nodes represent system events or current transformer error change points, directed edges represent possible causal associations, and edge weights reflect the strength of causal associations.

[0125] As shown in Figure 3 , the actual scenario and counterfactual scenario (assuming the fault does not occur) are shown. Through this comparison, the actual impact of the short-circuit fault on the current transformer ratio error can be visually observed, verifying the effectiveness of the counterfactual analysis method.

[0126] Step 3, construct a time-series causal graph to represent the temporal association between events and error changes, and perform initial weight calculation;

[0127] Based on the constructed time-series causal graph, this step applies the counter model to identify potential causal links.

[0128] The counter model is a causal reasoning algorithm based on event frequency statistics and time-series association analysis. It records the co-occurrence patterns between events by establishing multiple counter variables and identifies causal relationships by combining probability reasoning methods. The core is to transform complex causal reasoning problems into computable statistical problems.

[0129] The counter model consists of the following five functional modules:

[0130] Event counter module: establish an event frequency statistics table , which records the occurrence frequency of each system event in the observation time period , the first occurrence time , the last occurrence time , and the average occurrence interval . The specific calculation process is as follows: for each event type , traverse the entire time series, and when a type event is detected, the counter increases by 1, and the timestamp is recorded.

[0131] Error change monitoring module: establish an error change record table , which continuously records the occurrence time of current transformer error change points ​ , magnitude of change , direction of change (increase or decrease) and duration of change . The identification of error change points is based on the moving average difference algorithm: calculate the difference sequence of error values at consecutive time points, and mark as significant change points when the absolute value of the difference value exceeds the preset threshold .

[0132] Correlation analysis module: establish event-error co-occurrence count matrix , where is the number of event types, is the number of error change points. Matrix elements record the number of co-occurrences of events and error changes within a specified time window . The calculation process is: for each error change point , search for all events occurring within its forward time window and update the corresponding count value .

[0133] The correlation analysis module uses hierarchical time window technology to establish three different scales of time windows:

[0134] Short-term window 1-10 seconds): capture transient causal relationships;

[0135] Medium-term window (10-60 seconds): capture system response processes;

[0136] Long-term window (1-10 minutes): capture cumulative effects;

[0137] For each time scale, calculate the short-term, medium-term, and long-term co-occurrence count matrix , , respectively, and then obtain the comprehensive correlation strength by weighted summation:

[0138]

[0139] where , , , represent the short-term, medium-term, and long-term count weight coefficients, respectively. The default values are 0.4, 0.3, and 0.3, respectively. For example, if transient causal relationships are emphasized, the .

[0140] Probability inference module: based on the count matrix and event frequency table, the conditional probability is calculated and joint probability The conditional probability calculation formula is:

[0141]

[0142] Wherein is the event and the co-occurrence number of error change , is the total number of events occurred.

[0143] The probability reasoning module handles the time sequence dependence through Bayes theorem: for the event sequence , the joint influence probability of error change is calculated:

[0144]

[0145] Causal link screening module: screening potential causal links based on statistical significance test. For each pair of event-error relationship , the statistical significance index value is calculated:

[0146]

[0147] Wherein is the total number of observations, is the number of times when event occurred but error change did not occur, is the number of times when error change occurred but event did not occur, is the number of times when both event and error change did not occur.

[0148] When value is greater than the critical value (typically the critical value corresponding to the significance level ), it is considered that the causal link has statistical significance, and is retained in the candidate set.

[0149] The output of the application counter causal model is a set of potential causal relationship candidates that have passed statistical significance test, wherein is the association strength, is the statistical significance index;

[0150] Further, the probability inference module of the counterfactual causal model can handle temporal dependencies and consider potential interactions between events, thereby improving the accuracy of causal link identification.

[0151] The output of the counterfactual causal model is a set of candidate causal relationships, which includes the possible causal relationships between system events and error changes.

[0152] As shown in Figure 4 , the comparison of the current transformer phase angle error over time under the actual scenario and the counterfactual scenario (assuming the fault did not occur) is shown. The chart clearly shows that the short-circuit fault causes a significant peak in the phase angle error, while in the counterfactual scenario the phase angle error remains stable and low, further proving that the short-circuit fault is the main cause of the phase angle error change.

[0153] Step 4: Apply the counterfactual causal model to analyze the temporal causal graph and identify potential causal links and correlation strengths;

[0154] This step designs and implements a multi-world simulation engine to generate counterfactual scenarios. The technical meaning and specific implementation of the multi-world simulation engine: The multi-world simulation engine is a computational system based on state space search and conditional inference, which simulates the behavior of the system under different event intervention conditions by constructing multiple parallel system state evolution paths. Its core technical idea is to expand the single timeline of actual observation into multiple hypothetical timelines, each corresponding to a specific event intervention situation.

[0155] Technical meaning of counterfactual scenario: Counterfactual scenario refers to a hypothetical situation that is different from the actual observation by artificially intervening in the occurrence state of a specific event (assuming that the event that has occurred has not occurred, or assuming that the event that has not occurred has occurred). The generation of counterfactual scenarios is based on the control variable method: keep all other conditions unchanged, only change the state of the target event, and observe the differences in system response.

[0156] The multi-world simulation engine consists of the following four core components:

[0157] 1. State representation module: represents the state of the power system at time as a multi-dimensional vector , where is the state dimension. The state vector includes current transformer parameters (primary current, secondary current, ratio error, phase angle error), system load state (load size, load type, load change rate), environmental factors (temperature, humidity, electromagnetic interference intensity), and operating parameters (voltage level, frequency, phase) and other parameters.

[0158] This module performs standardization preprocessing on the input heterogeneous data:

[0159] For continuous parameters Min-max normalization is performed: The value range is limited to the interval [0, 1];

[0160] For periodic parameters Sine-cosine transformation is performed: where is the period length, avoiding the problem of period boundary jump;

[0161] For discrete parameters One-hot encoding is performed: the discrete value is converted to a dimensional binary vector, where only one element is 1 and the rest are 0.

[0162] 2. Event intervention module: implement intervention on specific events through event operation functions The event intervention module maintains an event intervention table , which records the intervention state of each event , where:

[0163] Indicates forced removal of the event (even if it actually occurs, it is set to not occurred);

[0164] Indicates maintaining the actual state of the event;

[0165] Indicates forced addition of the event (even if it does not actually occur, it is set to occur);

[0166] Specific implementation process of event intervention: given the original event sequence and intervention operation , generate the intervened event sequence . When (remove event), delete all events of type from the original sequence; when (add event), insert event at the specified time.

[0167] 3. State transition calculation unit: used to predict the time evolution of system state under the condition of occurrence or non-occurrence of specific events. This unit calculates the next time system state based on state transition function :

[0168]

[0169] where is​ intervention event vector at time t, is the system parameter vector.

[0170] The state transition calculation unit contains four processing levels:

[0171] Physical constraint level: build constraint condition set based on power system basic law (Ohm's law, Kirchhoff's law, etc.) to ensure that the state transition meets the physical feasibility. The constraint conditions include power balance constraints, voltage limit constraints, current limit constraints, etc.

[0172] Historical data learning level: learn historical state transition patterns through time series analysis algorithm. This layer uses a sliding time window method to predict the change trend based on historical state sequence .

[0173] Mixed inference level: fuse physical constraints and historical learning results. The fusion formula is: where is the fusion weight.

[0174] State verification level: check whether the predicted result meets the system constraint condition. If the predicted state violates the constraint condition, adjust the state value through the least squares method to make it meet the constraint requirement.

[0175] 4. Uncertainty quantification unit: evaluate the reliability of the simulation results to provide statistical reliability guarantee for causal effect analysis. This unit contains four calculation modules:

[0176] Monte Carlo sampling module: evaluate the impact of parameter uncertainty on the results by repeated sampling. For each uncertain parameter , assume it follows a known distribution , generate random samples , run the state transition calculation for each sample to get the result distribution.

[0177] Bayesian estimation module: calculate the posterior probability distribution of the predicted result based on Bayes' theorem. Given the observation data and the prior distribution , calculate the posterior distribution of the parameter: .

[0178] Sensitivity analysis module: identify the input parameters that have the greatest impact on the results through partial differentiation method. For output and input parameter , calculate the sensitivity index , The larger the value of parameter , the more significant the impact on the result.​

[0179] Confidence interval generator: provides confidence intervals for prediction results based on statistical distribution theory. For a prediction value , its standard error is calculated, and then a confidence interval with confidence level is constructed: where is the critical value of distribution.

[0180] Running process of the multiple world simulation engine: for each error change point of interest , the engine creates two parallel state evolution paths: the actual path (maintaining the original state of all events) and the counterfactual path (performing intervention operations on a specific event or . Both paths start from the same initial state and evolve at the same time step, eventually resulting in different error change prediction results, thereby quantifying the causal effect of the event on the error change .

[0181] The multiple world simulation engine generates multiple possible counterfactual scenarios for each error change point of interest through parallel computing, simulating the situation of “if a certain event did not occur, would the error change still occur?”

[0182] As shown in Figure 5 , the causal effect intensity of different system events on current transformer error change is shown, and each event is sorted by the normalized causal effect intensity value. The chart clearly shows that short-circuit faults, system switching, and lightning disturbances have the highest causal effect intensity, which provides an important basis for event prioritization in current transformer error monitoring.

[0183] Step 5, generate counterfactual scenarios through the multiple world simulation engine to simulate the occurrence and non-occurrence of specific events and evaluate event impact;

[0184] This step quantifies the causal effect of each event on the current transformer error change by comparing the differences between actual observations and counterfactual simulation results. The execution steps of the causal effect quantification algorithm are as follows:

[0185] Determine the target error change point and its possible related event set , where , , represent the , , an event; denotes the total number of events.

[0186] For each event , the error change prediction value under the actual scenario (event occurs) and counterfactual scenario (event does not occur) is obtained from the multiple world simulation engine;

[0187] Calculate the average treatment effect (ATE):

[0188] ;

[0189] where, denotes the event The average treatment effect of error change; denotes the i-th system event; denotes the j-th error change point; denotes the operation of making the event occur through intervention; denotes the operation of making the event not occur through intervention; is a conditional expectation function, used to calculate the average or expected value of error change under a specific intervention condition; denotes the direction of causality, from cause to result.

[0190] Perform robustness verification by repeating steps 2-3 multiple times under random initial conditions, and calculate the confidence interval of the causal effect;

[0191] Normalize the causal effect of all events, convert the effect values of different scales and units to comparable unified dimensions through Z-Score standardization, and then construct an attribution distribution graph;

[0192] According to the preset threshold, filter out key events with significant causal effects.

[0193] Further, the causal effect quantification algorithm can process the composite causal effect of multiple events in a recursive manner, considering the sequential dependence relationship between events.

[0194] Further, the causal effect quantification algorithm adopts time decay weight when processing long-term causal effects, so that recent events have higher weight values.

[0195] By quantifying the causal effect, this method can distinguish between only correlation and true causality, and thus identify the true trigger factors of current transformer error change.

[0196] For example Figure 6 ​As shown, the identification accuracy of the traditional correlation analysis method and the counterfactual analysis method proposed in this application under different error types and event type combinations is compared. The chart directly shows that the accuracy of this method is significantly higher than that of the traditional method in all test scenarios, especially in the slow influencing factors such as environmental temperature changes, which are difficult to identify.

[0197] Step 6, compare the difference between actual observation and counterfactual simulation result, quantify the causal effect of each event on the error change of current transformer;

[0198] Based on the quantified causal effect strength, this step establishes a continuously running monitoring mechanism to identify and track key trigger events affecting the error of current transformer in real time. The mechanism includes:

[0199] Establish a causal event monitoring priority queue, and pay more attention to events with higher causal effect strength;

[0200] Implement an adaptive threshold mechanism to dynamically adjust the causal effect strength threshold for triggering alarms according to the system running state;

[0201] Implement a sliding time window analysis algorithm, the specific steps are:

[0202] Determine the size of the time window and the sliding step, usually 1-3 times the system response time;

[0203] Calculate the local causal effect value for the data in each sliding window;

[0204] Compare the causal effect values of adjacent windows to detect the dynamic changes of causal relationship;

[0205] Trigger re-evaluation when the change exceeds the preset threshold;

[0206] Integrate the event combination effect evaluation model, which includes:

[0207] Combination identification module: detect frequently co-occurring event combinations;

[0208] Interaction effect calculation unit: quantify the synergistic or antagonistic effect between events, and standardize and cross-process the interaction features of different event types;

[0209] Analytic hierarchy module: decompose event combinations into main effects and interaction effects;

[0210] Visualization interface: convert complex interaction relationships into intuitive visual expressions.

[0211] The key trigger event monitoring mechanism realizes real-time and accurate tracking of the error change of the current transformer, providing a reliable basis for subsequent error compensation.

[0212] As Figure 7 shown, the traditional correlation analysis method and the counterfactual analysis method are compared in terms of false positive rate, false negative rate, alarm accuracy and overall monitoring effectiveness. The chart clearly shows the significant advantage of the method in reducing false positive rate and false negative rate, proving that counterfactual analysis can more accurately identify the true trigger factors of current transformer error changes, reduce unnecessary maintenance costs, and not miss important issues.

[0213] Step 7, based on the strength of the causal effect, real-time monitoring and identification of key trigger events affecting the current transformer error, and generation of early warning information;

[0214] This step generates a dynamic causal explanation report based on the previous analysis and adjusts the compensation strategy for the root cause in real time. Specifically, it includes:

[0215] Automatically generate a causal link visualization chart of error changes, which intuitively displays the causal relationship between events and errors;

[0216] Provide error compensation suggestions for different types of events, including parameter adjustment strategies and algorithm correction methods;

[0217] Build an error trend prediction model, which consists of the following components:

[0218] Time series feature extractor: captures the time pattern and periodicity of historical error changes, and performs seasonal adjustment and trend decomposition preprocessing on the input error time series data;

[0219] Causal enhancement prediction unit: integrates the identified causal relationships to improve prediction accuracy;

[0220] Multi-step predictor: capable of predicting error change trends on different time scales;

[0221] Confidence assessment module: provides reliability estimates for each prediction to facilitate decision-making;

[0222] Build a knowledge base to accumulate causal analysis experience and continuously optimize compensation strategies.

[0223] Through the dynamic causal explanation report, system operation and maintenance personnel can clearly understand the root cause of the current transformer error change and take more accurate compensation measures.

[0224] Application examples of the present embodiment:

[0225] Application scenarios:

[0226] This example is based on a current transformer error dynamic monitoring system in a certain 500 kV substation. The substation is equipped with multiple current transformers for measuring the current of main transformers, transmission lines, busbars, and other equipment. In actual operation, the substation often faces complex events such as load fluctuations, lightning strikes, short-circuit faults, and system switching, which can cause dynamic changes in current transformer measurement errors. Traditional monitoring methods cannot accurately identify which events are the true triggering factors of error changes under these complex events.

[0227] This example selects a current transformer (CT-L231) installed on a 500 kV transmission line for monitoring, with a 30-day observation period. During this period, various system events and current transformer error change data are recorded.

[0228] Implementation process instance:

[0229] Data acquisition and preprocessing:

[0230] This example uses the automatic measurement system in the substation to collect the primary and secondary side current values of current transformer CT-L231, and calculates the ratio error and phase angle error. At the same time, system event logs are collected from the substation event management system. Table 1 shows an example (partial) of the collected raw data:

[0231] Table 1: Example of current transformer time series measurement data and system event logs;

[0232]

[0233] Preprocess the collected data: normalize the primary and secondary current values, standardize the ratio error and phase angle error, and encode the system event types into numerical categorical variables.

[0234] Time causal graph construction and counter causal model application:

[0235] Based on the preprocessed data, a time causal graph is constructed to represent the temporal relationship between system events and error changes. By identifying significant error change points and analyzing event sequences, directed edges are established and edge weights are calculated. In practical applications, the maximum time lag parameter is set to 30 seconds (5 times the system response time). Table 2 shows an example of edge weight calculation for the time causal graph:

[0236] Table 2: Example of edge weight calculation for the time causal graph;

[0237]

[0238] After calculating the edge weights, perform edge pruning to remove edges with weights below a threshold value (0.5) to reduce the complexity of the graph.

[0239] The application counter causal model is used to analyze the time series causal graph and identify potential causal links. The process includes: the event counter module records and tracks the number of system event occurrences and timestamps; the error change monitoring module continuously records the error change points and amplitudes; the correlation analysis module calculates the spatio-temporal correlation degree of events and error changes; the probability reasoning module realizes conditional probability calculation based on Bayesian network; the causal link screening module screens potential causal links according to the statistical significance threshold. Through analysis, the potential causal relationships with high statistical significance are screened out as follows:

[0240] Short circuit fault→rapid increase of ratio error→rapid increase of phase angle error;

[0241] System switching→phase angle error increase;

[0242] Load increase→slow increase of ratio error.

[0243] The multiple world simulation engine generates counterfactual scenarios:

[0244] For the screened potential causal relationships, the multiple world simulation engine generates counterfactual scenarios. The engine is composed of four key components working together: the state representation module represents the power system state as a multi-dimensional vector (including current transformer parameters, load state, etc.); the event intervention module creates hypothetical scenarios that "remove" or "add" specific events; the state transition model (including physical constraint layer, historical data learning layer, hybrid reasoning layer, and state verification layer) predicts system state evolution; the uncertainty quantification unit (including Monte Carlo sampling module, Bayesian estimation module, sensitivity analysis module, and confidence interval generator) evaluates the reliability of simulation results.

[0245] For example, for the causal relationship "short circuit fault→rapid increase of ratio error", the system state evolution under the conditions of short circuit fault occurrence and non-occurrence is generated. Table 3 shows the counterfactual scenario simulation results for a short circuit fault:

[0246] Table 3: Comparison of counterfactual scenario simulation results for short circuit fault;

[0247]

[0248] As can be seen from Table 3, in the actual scenario of short circuit fault occurrence, both the ratio error and the phase angle error increase significantly; while in the counterfactual scenario (assuming short circuit fault does not occur), the error change is very small, consistent with the slow change trend under normal operating conditions.

[0249] Quantifying causal effects and monitoring key trigger events:

[0250] The causal effect of each event on the error change of the current transformer is quantified by comparing the difference between the actual observation and the counterfactual simulation result. The average treatment effect (ATE) is calculated, and the causal effects of all events are normalized to screen out key events with significant causal effects. Table 4 shows the quantitative results of the causal effect intensity of different events:

[0251] Table 4: Quantitative results of the causal effect intensity of different system events

[0252]

[0253] Based on the quantified causal effect intensity, a monitoring mechanism is established to identify and track key triggering events that affect the error of the current transformer in real time. The monitoring mechanism includes: establishing a causal event monitoring priority queue, focusing on events with higher causal effect intensity; implementing an adaptive threshold mechanism, dynamically adjusting the threshold for triggering alarms according to the system operating state (currently set to a causal effect intensity of 0.6); implementing a sliding time window analysis algorithm, setting the window size to 15 seconds (3 times the system response time), calculating the local causal effect value and comparing adjacent windows; integrating an event combination effect evaluation model, quantifying the synergistic or antagonistic effects between events through a combination of identification modules, interaction effect calculation units, hierarchical analysis modules, and visualization interfaces.

[0254] The system automatically establishes a causal event monitoring priority queue, focusing on short-circuit faults, system switching, and lightning disturbance, which have higher causal effect intensity. When these events occur, the system will immediately increase the monitoring frequency of the current transformer error and prepare the corresponding compensation strategy.

[0255] Technical effect verification:

[0256] This embodiment verifies the two main technical effects of the method through 30 days of substation current transformer operation monitoring: the accuracy of error trigger factor identification is improved, and the false alarm rate is reduced.

[0257] The accuracy of error trigger factor identification is significantly improved:

[0258] Table 5 shows the comparison of the accuracy of error trigger factor identification between the proposed method and the traditional correlation analysis method:

[0259] Table 5: Comparison of error trigger factor identification accuracy

[0260]

[0261] The data shows that the average accuracy of identifying the real trigger of current transformer error change is 90.0%, which is 29.5 percentage points higher than the traditional correlation analysis method. Especially in the identification of slow influencing factors such as environmental temperature changes, the accuracy has increased by 37.4 percentage points.

[0262] False positive rate and false negative rate are reduced:

[0263] In the 30-day operation monitoring, the system recorded 278 current transformer error change events, of which 126 were significant changes. Table 6 shows the comparison of false positive rate and false negative rate between the proposed method and the traditional method:

[0264] Table 6: Comparison of false positive rate and false negative rate;

[0265]

[0266] The data shows that after using the proposed method, the false positive rate is reduced from 23.8% to 6.5%, and the false negative rate is reduced from 18.7% to 4.2%, with an overall monitoring performance improvement of 22.7 percentage points. This means that the system can more accurately identify error change events that need attention, avoid unnecessary maintenance costs, and ensure that important issues are not missed.

[0267] Generate dynamic causal explanation report:

[0268] In this embodiment, the system also generates a dynamic causal explanation report based on the aforementioned analysis and adjusts the compensation strategy for the root cause in real time.

[0269] The system also provides error compensation suggestions for the identified key events, such as parameter recalibration after fault recovery for short-circuit fault-induced errors, and adaptive angle compensation algorithm for phase angle errors caused by system switching.

[0270] The system also builds an error trend prediction model, which consists of a time series feature extractor (seasonal adjustment and trend decomposition of input error time series data), a causal enhancement prediction unit, a multi-step predictor, and a confidence assessment module. Through this model, the system can predict the error change trend caused by various events within the next 24 hours and provide a 95% confidence interval.

[0271] The system accumulates causal analysis experience in the knowledge base, continuously optimizes the compensation strategy, and improves the accuracy and reliability of power system measurement.

[0272] The above describes the embodiments of the present application, but the embodiments are not limited to the above specific embodiments, and the above specific embodiments are only illustrative but not restrictive, and the ordinary skilled in the art can make more equivalent embodiments under the inspiration of the embodiments, which are all within the protection scope of the embodiments.

Claims

1. A method of current transformer error dynamic monitoring, characterized by, The method comprises the following steps: Collect and pre-process the current transformer time series measurement data and system event logs, including data cleaning, standardization and outlier processing; Feature extraction is performed on the pre-processed data to identify key time series features and event features; A time causal graph is constructed to represent the temporal relationship between events and error changes, and the initial weight calculation is performed; The time causal graph is a directed graph structure, where nodes represent system events or current transformer error change points, directed edges represent possible causal relationships, and edge weights reflect the strength of causal relationships; The counter causal model is applied to analyze the time causal graph to identify potential causal links and association strength; The counter causal model includes: An event counter module establishes an event frequency table to record the occurrence frequency, first occurrence time, last occurrence time and average occurrence interval of each system event within the observation period; An error change monitoring module establishes an error change record table to continuously record the occurrence time, change amplitude, change direction and change duration of the current transformer error change point; An association analysis module establishes an event-error co-occurrence count matrix, uses a hierarchical time window technique to establish short-term, medium-term and long-term time windows of different scales, calculates the co-occurrence count matrix respectively, and obtains the comprehensive association strength through weighted summation; A probability reasoning module calculates the conditional probability and joint probability based on the count matrix and event frequency table, and processes the joint influence probability of event sequence on error change through Bayes theorem; A causal link screening module screens potential causal links based on statistical significance test, calculates the chi-square statistic for each event-error relationship, and considers that the causal link has statistical significance when the chi-square value is greater than the critical value; Generate counterfactual scenarios through a multiple world simulation engine to simulate the occurrence and non-occurrence of specific events and evaluate event impact; The multiple world simulation engine includes a state representation module, an event intervention module, a state transition calculation unit and an uncertainty quantification unit, which intervene in specific events through event operation functions to create parallel state evolution of actual path and counterfactual path; Counterfactual scenarios refer to hypothetical situations different from actual observations constructed by artificially intervening in the occurrence state of specific events; the generation of counterfactual scenarios is based on the control variable method: keeping all other conditions unchanged, only changing the state of the target event, and observing the difference in system response; Compare the differences between actual observations and counterfactual simulation results to quantify the causal effects of each event on current transformer error changes; Based on the strength of causal effects, real-time monitoring and identification of key trigger events affecting current transformer errors are performed, and warning information is generated.

2. The method of claim 1, wherein, Preprocessing includes: Amplitude normalization of primary current and secondary current values of the current transformer; Standardization of ratio error and phase angle error; Encoding event types in system event logs to convert non-numeric event types into numeric categorical variables.

3. The method of claim 1, wherein, The construction of the time causal graph with a directed and weighted graph structure includes: The error change point identification algorithm is used to identify the error change points by wavelet decomposition of the original error sequence, application of a soft threshold function to the wavelet coefficients of each layer using a Daubechies-4 wavelet basis function, calculation of the first-order difference of the denoised error sequence, and application of a double-threshold judgment; The event sequence standardization algorithm is used to standardize the system event log by assigning timestamps, event type codes, event severity, and event duration attributes to each event, forming a structured event sequence; The graph node creation algorithm is used to create graph nodes for each error change point and system event, establish a node index table, and assign node ID, node type, timestamp, and node weight attributes to each node; The time window constraint algorithm is used to set the maximum time lag parameter based on the system physical response characteristics, and the directed edge creation algorithm is used to perform time sequence constraint and time window constraint checks to create directed edges that meet the constraint conditions; The edge weight calculation algorithm is used to calculate the weight of each directed edge using a multi-factor weighted model, including a conditional probability term, a time decay term, and a statistical correlation term; The edge pruning algorithm is used to remove insignificant edges based on weight threshold pruning, statistical significance pruning, and redundant edge pruning to reduce the complexity of the graph.

4. The method of claim 1, wherein, The multi-world simulation engine consists of the following four core components: state representation module: represents the state of the power system at time t as a multi-dimensional vector; Event intervention module: implements intervention on specific events through event operation functions; State transition calculation unit: used to predict the time evolution of the system state under the condition that a specific event occurs or does not occur; calculates the next time system state based on state transition function; The state transition calculation unit includes: Physical constraint layer: constructs a set of constraint conditions based on the basic laws of power systems; Historical data learning layer: learns historical state transition patterns through time series analysis algorithms; Mixed reasoning layer: combines physical constraints and historical learning results; State verification layer: checks whether the predicted results meet the system constraint conditions. If the predicted state violates the constraint conditions, adjust the state value through the least squares method to meet the constraint requirements; Uncertainty quantification unit: evaluates the reliability of the simulation results to provide statistical reliability guarantee for causal effect analysis; includes four calculation modules: Monte Carlo sampling module: assesses the impact of parameter uncertainty on the results through repeated sampling; Bayesian estimation module: calculates the posterior probability distribution of the prediction results based on Bayes' theorem; Sensitivity analysis module: identifies the input parameters that have the greatest impact on the results through partial differentiation; Confidence interval generator: provides confidence intervals for the prediction results based on statistical distribution theory.

5. The method of claim 1, wherein, The state representation module performs the following preprocessing on the input heterogeneous data: Minimum-maximum normalization is performed on continuous parameters to limit the value range to a specific interval; Sine-cosine transformation is performed on periodic parameters to avoid period jumps; Discrete parameters are processed using one-hot encoding.

6. The method of claim 1, wherein, The steps to quantify the causal effects of each event on the error change of the current transformer include: Determine the analysis target error change point and its possible related event set; Obtain error change prediction values in actual and counterfactual scenarios through a multiple-world simulation engine, and create parallel state evolutions of actual and counterfactual paths; Calculate the causal effect of events on error changes through an average treatment effect formula, which is the difference between the expected value of error changes in the actual scenario and the expected value of error changes in the counterfactual scenario; Perform robustness verification by repeatedly calculating multiple times under random initial conditions, and calculate the confidence interval of the causal effect using statistical distribution theory; Standardize and normalize the causal effects of all events using the Z-Score method, converting effect values of different scales and units into comparable unified dimensions.

7. The method of claim 1, wherein, The steps for real-time monitoring and identifying key trigger events affecting the error of current transformers include: Establish a causal event monitoring priority queue, focusing on events with higher causal effect strength; Adjust the causal effect strength threshold for triggering alarms based on the historical statistical characteristics of system operation states through an adaptive threshold algorithm; Determine the size and sliding step of the time window, calculate the local causal effect value for the data in each sliding window, and compare the causal effect values of adjacent windows to detect the dynamic changes of causal relationships through a sliding time window analysis algorithm; Through an event combination effect evaluation algorithm, including a combination recognition algorithm to detect frequently co-occurring event combinations, an interaction effect calculation algorithm to quantify the synergistic or antagonistic effects between events, and a hierarchical analysis algorithm to decompose event combinations into main effects and interaction effects.

8. The method of claim 7, wherein the step of determining the error of the current transformer comprises the steps of: The sliding time window analysis algorithm includes: Determine the size and sliding step of the time window; Calculate the local causal effect value for the data in each sliding window; Compare the causal effect values of adjacent windows to detect the dynamic changes of causal relationships; Trigger a re-evaluation when the changes exceed the preset threshold.

9. The method of claim 1, wherein, Also includes: Generate a dynamic causal explanation report, including a visual chart of the causal chain of error changes; Provide error compensation suggestions for different types of events; Build an error trend prediction model; Build a knowledge base to accumulate causal analysis experience and continuously optimize compensation strategies.

10. A current transformer error dynamic monitoring system for performing the method of any one of claims 1-9, wherein, Includes: A data acquisition and preprocessing unit for acquiring current transformer time series measurement data and system event logs, and performing amplitude normalization, standardization, and event type encoding conversion; A time series causal graph construction unit for constructing a directed and weighted graph structure, representing the time series causal relationship between system events and error changes through error change point identification algorithms, event sequence standardization algorithms, graph node creation algorithms, time window constraint algorithms, directed edge creation algorithms, edge weight calculation algorithms, and edge pruning algorithms; A counter causal model unit based on event frequency statistics and time series correlation analysis, identifying a set of potential causal relationship candidates that have passed statistical significance tests through event counter modules, error change monitoring modules, correlation analysis modules, probability reasoning modules, and causal link filtering modules; A multiple-world simulation engine based on state space search and conditional reasoning, containing state representation modules, event intervention modules, state transition calculation units, and uncertainty quantification units, which intervenes in specific events through event operation functions to generate parallel state evolutions of actual and counterfactual paths; A causal effect quantification unit is configured to calculate the causal effect of events on error changes by averaging the effect formula, perform robustness verification and calculate a confidence interval, and normalize the causal effect; A key trigger event monitoring unit is configured to establish a causal event monitoring priority queue, and monitor and identify key trigger events affecting the error of the current transformer in real time by using an adaptive threshold algorithm and a sliding time window analysis algorithm; A dynamic report generation unit is configured to generate a visual chart of the causal link of error changes, provide error compensation suggestions for different types of events, and build an error trend prediction algorithm and a knowledge base to accumulate causal analysis experience.

Images